Research Article A QoS Aware Resource Allocation Strategy for …downloads.hindawi.com/journals/tswj/2014/419236.pdf · 2019-07-31 · Proposed Resource Allocation Strategy is section
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Research ArticleA QoS Aware Resource Allocation Strategy for 3D AVStreaming in OFDMA Based Wireless Systems
Young-uk Chung Yong-Hoon Choi Suwon Park and Hyukjoon Lee
College of Electronics and Information Engineering Kwangwoon University Seoul 139-701 Republic of Korea
Correspondence should be addressed to Hyukjoon Lee hleekwackr
Received 26 May 2014 Accepted 2 August 2014 Published 28 August 2014
Academic Editor Chin-Chia Wu
Copyright copy 2014 Young-uk Chung et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited
Three-dimensional (3D) video is expected to be a ldquokiller apprdquo for OFDMA-based broadband wireless systemsThe main limitationof 3D video streaming over a wireless system is the shortage of radio resources due to the large size of the 3D traffic This paperpresents a novel resource allocation strategy to address this problem In the paper the video-plus-depth 3D traffic type is consideredThe proposed resource allocation strategy focuses on the relationship between 2D video and the depth map handling them withdifferent priorities It is formulated as an optimization problem and is solved using a suboptimal heuristic algorithm Numericalresults show that the proposed scheme provides a better quality of service compared to conventional schemes
1 Introduction
Recent advances in video technology have made three-dimensional audiovideo (3D AV) services technically fea-sible 3D AV has emerged to provide more immersive expe-riences compared to conventional two-dimensional (2D)video It allows viewers to feel a compelling sense of physicallyreal spaces Therefore it is expected to be a ldquokiller apprdquofor smart-phones and corresponding broadband wirelesssystems
3D AV is generated by multiview images which isobtained in two ways through multiview cameras [1 2] andwith a depth-image-based rendering (DIBR) technique [3ndash7] Multiview cameras involve a number of cameras to obtainimages according to usersrsquo viewpoints However these typesof cameras require complicated coding and careful trans-mission techniques because the amount of data increasesproportionally to the number of cameras
Instead of using raw data of multiviewed images theDIBR techniquemakes use of two types of images 2D textureimages and synchronized depth maps This is known asthe video-plus-depth concept which was introduced by theadvanced three-dimensional television system technologies(ATTEST) project [8] These two types of image streams areseparately encoded into the base layer and the enhancement
layer The DIBR technique can reconstruct and render 3Dvideo from these two data streams Because the required datasize and the degree of processing complexity are relativelylow the DIBR technique is considered as suitable for a 3DAV system
As mentioned above the 3D AV service is expected tobecome widely popular on smart-phones and correspondingbroadband wireless systems based on the orthogonal fre-quency division multiple access (OFDMA) OFDMA tech-nology is known to be suitable for next-generation broadbandwireless systems because high-speed data transmission ispossible under limited radio resources However the sizeproblem is still severe in OFDMA-based wireless systemsbecause the limited radio resources represent a fundamentalproblem of current wireless systems Though the video-plus-depth concept based on the DIBR technique can reduce theamount of data compared to the multiview camera methodthe amount of 3D AV traffic is still large enough such that itcan hardly be transmitted over a wireless system in real-timeTherefore the efficient resource allocation of radio resourcesis important to guarantee the QoS of a 3D AV service
In OFDMA-based wireless systems resource allocationis performed by means of a scheduling procedure Therehave been many scheduling algorithms for wireless systemsThe maximum carrier-to-interference ratio (Max CI) [9]
Hindawi Publishing Corporatione Scientific World JournalVolume 2014 Article ID 419236 11 pageshttpdxdoiorg1011552014419236
2 The Scientific World Journal
round-robin (RR) [10] proportionally fair (PF) [11] andfast fair-throughput (FFTH) [12] scheduling algorithms arethe most commonly used types The Max CI algorithmprovides throughput maximization and the RR algorithmachieves the optimal level of fairness The PF and theFFTH algorithms provide a good balance between systemthroughput and fairness but they do not take the QoS intoaccount Several scheduling algorithms have been studied toguarantee QoS for end users The modified largest weighteddelay first (MLWDF) [13] and the exponential rule (EXP) [14]algorithms are the most popular types They consider boththe maximum allowable delay and the instantaneous channelrate However little work has been carried out to find anoptimal resource allocation scheme that considers the QoSof a 3D AV service
In this paper we propose a resource allocation scheme forwireless transmissions of 3D AV traffic We focus on a QoSguarantee for 3D AV services in the proposed scheme Todo this we adopt a concept which uses a base layer and anenhancement layer for 3DAV traffic as introduced by video-plus-depth concept This concept processes 3D traffic moreefficiently
The rest of this paper is organized as follows first wegive an overview of the system environment in which theproposed scheme is adopted in Section 2 In Section 3we introduce the proposed resource allocation scheme andformulate it as an optimization problem Then we give adetailed explanation of a suboptimal heuristic algorithm tosolve the optimization problem In Section 4 we describe thesimulation environment and evaluate the performance of theproposed scheme based on several numerical results Finallywe conclude the paper in Section 5
2 System Description
This section gives an overview of the system environment inwhich the proposed resource allocation scheme is adopted Itbegins with an introduction of the overall system architectureof 3D video over wireless systems Also a detailed descriptionof video-plus-depth 3D video is discussed
21 3D Video over Wireless Systems Figure 1 shows the over-all system architecture of 3D video over wireless systemsas considered in this paper The system consists of fourparts denoted here as a 3D media server a packet datanetwork gateway (PDN GW)serving GW an enhancednode-B (eNB)enhanced universal terrestrial radio accessnetwork (E-UTRAN) and the user equipment (UE) The 2Dvideo and its associated depthmap are obtained from a depthcamera system or with a 2D-to-3D conversionmethodThesedata are separately encoded as the base and the enhancementlayersThey are transmitted as separate streams of media datathrough a single connection Before transmitting over the IPnetwork thesemedia data are packetized into individual RTPpackets
The media server sends RTP packets to the wirelesssystem through an IP network In this paper we considerwireless systems based on the orthogonal frequency divisionmultiple access (OFDMA) such as the long-term evolution
(LTE) LTE-advanced (LTE-A) IEEE 80216e IEEE 80216mand other technologies In Figure 1 we show an exampleof the system architecture which adopts the radio accessnetwork (RAN) of the LTE-A systems As shown in the figureLTE-A system uses the PDN GWserving GW eNB and UEcomponents This system is connected to a 3D media serverthrough IP networks
The PDN GW or serving GW connects the IP networkand the LTE-A system The RTP packets of the base layerand the enhancement layer data come into the LTE-A systemthrough the general packet radio service (GPRS) tunnelingprotocol (GTP) GTP is an IP-based protocol which is usedin the universal mobile telecommunications system (UMTS)network GTP-U in Figure 1 is used to carry user data withinthe GPRS core network and between the RAN and the corenetwork The carried user data has usually IP or PPP formatIn our system the user data has IP format
eNB receives user data throughGTP-UUDPIP and thensends the data to the UE through the RAN Each instance ofdata is passed through the packet data convergence protocol(PDCP) and the radio link control (RLC) after which itis segmented and encapsulated as one or more mediumaccess control (MAC) frames and inserted into MAC bufferThe base layer traffic and the enhancement layer traffic areinserted into logically separated MAC buffers In the eNBMAC layer radio resources are allocated by schedulingAfter resource allocation by scheduling eNB repeatedly readsMAC frames from theMAC buffer and sends them to the UEthrough the RAN
Upon receiving the MAC frames the UE reassemblesand decapsulates them as RTP packets throughRLCPDCPIPUDP layers The RTP packets are combinedinto the base layer and the enhancement layer data Eachlayer of data is decoded so as to recover the depth map andthe 2D video respectively Finally they are converted into a3D video stream and displayed by the display screen of theUE
22 Video-Plus-Depth 3D Video In this paper we focus onthe video-plus-depth representation of multiview 3D videowhich renders a 3D video using a 2D video stream and itsassociated depth map This type is widely used because itprovides a flexible representation of 3D and because it is com-patible with existing coding and transmission technologies[3ndash8]
In this format the depth map includes 256 leveled greyimages It also contains depth information about the pixelpositions of the associated 2D video It can be acquired bya depth camera directly or can be extracted by a multiviewimage The size of the depth map is related to the numberof viewpoints of the 3D video Because the resource require-ments of 3D video applications based on the video-plus-depth scheme are high compared to those of 2D video appli-cations efficient compression techniques are required for 3Dvideo The layered coding approach can be effective withregard to this requirement In the layered coding approachshown in Figure 2 the 2D video and the depth imagesequence are encoded as the base layer and the enhancementlayer respectively Existing 2D compression techniques are
The Scientific World Journal 3
Wired environment(IP network)
Wireless environment
(RAN)
3D video
encoder
3D recording
2D to 3D conversion
Wired environment(IP network)
2D video ( base layer)
Depth map
(enhancementlayer)
RTP
RTCP
3D media server
RTP
RTCP
3D
video
decoder
DIBR
2D video
( base layer)
Depth map
(enhancement layer)
UE
UD
P
IP L2 L1 L2 IP
UD
P
L1
IP UD
P
L2L1
UD
PIP
UD
PIP
GTP
-U
PDCP L1L2
RLCMACL1
Display screen
PDN GWserving GW
IPL1 RLC
PDCP
MAC
RTP
RTCP
UD
P
eNB (E-UTRAN )
GTP
-U
Figure 1 System architecture of 3D AV over wireless systems
3D media server
3D traffic
DepthEnhancement
layerencoder
Base layerencoder2D video
Bit-stream
Figure 2 Layered coding approach of video-plus-depth 3D video
used to encode both the 2D video and the depth mapsequence After delivery the base and enhancement layers aredecoded as the 2D video and the depth image sequence on thereceiver side Before displaying 3D video on the display the
supplied 2D video and depth image sequences are convertedinto 3D video sequences using an image-warping techniqueknown as DIBR
3 Proposed Resource Allocation Strategy
This section gives a detailed description of the proposedresource allocation strategy adaptive to 3D video over wire-less systems First we describe the key idea of the proposedstrategy to guarantee the QoS formulating it as an opti-mization problem Next we present a suboptimal heuristicapproach to solve this problem
31 Description of the Proposed Strategy There have beenseveral definitions and measures pertaining to the QoS ofa 3D AV service [1ndash8] Among them we use continuityof service as the QoS measure of the 3D AV serviceAccordingly a consecutive frame transmission in the MAClayer is required to guarantee the QoS Moreover an efficient
4 The Scientific World Journal
resource allocation method is required due to the limitedresources in wireless networks
The key idea of the proposed resource allocation strategyis that the base layer traffic is given priority when assigningresources to guarantee the QoS In this paper we focuson 3D video traffic based on the video-plus-depth methodwhich consists of a 2D video stream and its associateddepth map As noted in Section 22 the 2D video and thedepth map are separately encoded as the base layer andthe enhancement layer respectively These two layers ofdata are transmitted together through a single connectionGiven that the enhancement layer data contains valuableinformation with which to implement the 3D video it shouldbe considered as important However the base layer datais more important than the enhancement layer data withregard to the QoS because the enhancement layer data playsa supporting role to convert the base layer data into 3D videoAssuming that eNB can transmit to a UE either the baselayer data or the enhancement layer data on account of ashortage of resource in such a case if the UE receives onlythe enhancement layer data it can no longer be providedwith video streaming service However if the UE receivesonly the base layer data it can avoid interruptions of its videostreaming service though 2D video is provided instead of3D video Therefore we give priority to the base layer trafficwhich can provide a 2D service by itself This also helps toguarantee the QoS
This idea is adopted in the scheduling procedure whichworks in the MAC layer of the eNB As described inSection 21 the base layer data and the enhancement data aredelivered from a 3D media server to the MAC layer buffer inthe eNB These two types of traffic are inserted into logicallyseparated buffersThey are transmitted to the destination (theMAC layer of the UE) when resources are allocated by thescheduler at every slot time The scheduling algorithm aimsto achieve optimal resource allocation which maximizes thethroughput of the 3DAV traffic while guaranteeing the QoSIt is formulated as described below
We assume that traffic whose destination is UE 119894 isdelivered to eNB at an average rate of packetslot time Thepackets are stored in aMAC buffer 119861
119894 For the 3DAV traffic
two buffers are assigned to the enhancement layer traffic andthe base layer traffic respectively In this paper we assumethat the buffers are large enough to not to overflow Westipulate that the amount of data in buffer 119861
119894at the beginning
of the kth slot is 119909119894119896 From the result of scheduling 119906
119894119896is
transmitted during kth slot 119906119894119896depends on the allocated data
rate at buffer 119861119894 Then the buffer is updated as
is the input traffic size in 119861119894during the kth slot
Let119863119894be the average queuing delay for 119861
119894119863119894is related to the
average buffer length via Littlersquos theorem [15] and is describedas
119863119894=1
120582119864 [119909119894119896] (2)
where 120582 = 119864[119886119894119896] is the average packet arrival rate Because
0 le 119906119894119896le 119909119894119896 the smallest average delay of the kth slot
is achieved when 119906119894119896= 119909119894119896
and the average queuing delaybecomes119863
119894= 1
Let 119866119894119899
be the channel gain 119873119894119899
the total noise powerspectral density and 119901
119894119899the allocated power for user 119894 to
subcarrier 119899 In this formulation we consider an OFDMAbased wireless system We assume that M-QAM modulationis applied with a BER requirement The signal-to-noise ratio(SNR) SNR
where Γ = minus ln(5 sdot BER)15 [16] In addition the capacity ofuser 119894 on subcarrier 119899 is normalized by
119903119894119899= ln (1 + 119901
119894119899sdot SNR
119894119899) (4)
The instantaneous data rate of user 119894 can then be described as
119877119894=
119873
sum
119899=1
119908119894119899ln (1 + 119901
119894119899SNR119894119899) (5)
and the number of resource elements (RErsquos) required tosupport 119877
119894while transmitting on subcarrier 119899 is 119904
119894 119908119894119899is the
subcarrier allocation index it has a value of 1 when subcarrier119899 is allocated to user 119894 Otherwise it has a value of 0
Because our goal is throughput maximization whileguaranteeing the QoS the scheduling problem can be writtenas shown below
Maximize OP (119908 119901) = max119908119894119899 119901119894119899
119868
sum
119894=1
ln119877119894
(6)
subject to 119863119887
119894minus 1 le 119863th 119863
119887
119894ge 1 forall119894 (7)
119868
sum
119894=1
119873
sum
119899=1
119901119894119899le 119875119879
119901119894119899ge 0 forall119894 119899 (8)
119868
sum
119894=1
119908119894119899le 1 119908
119894119899ge 0 forall119894 119899 (9)
119868
sum
119894=1
119904119894le 119878max (10)
Here 119868 is the total number of UEs and 119863119887119894is the average
queuing delay of the buffer which contains the base layertraffic for the UE i 119863th is the threshold delay and 119878max isthe total number of RErsquos within a slot 119908 = [119908
119894119899]119868 times119873
119901 =[119901119894119899]119868 times119873
and 119875119879are the total available transmission power
for eNB Note that OP(119908 119901) is neither convex nor concavewith respect to (119908 119901) Although 119908
119894119899is defined to obtain a
value of either 0 or 1 it is permitted to be a real numberbetween 0 and 1 to make the problem tractable
Equation (6) represents the total throughput computedby summing up the logarithmic user data rate assigned toeach UE There are four constraints Equation (7) indicatesthat the queuing delay for the base layer traffic should notexceed the threshold delay bound as the data in the buffer
The Scientific World Journal 5
(I) INITIALIZATION(1) 119878
119894larr 0 forall119894 = 1 119872
(2) 119891119894larr 0 forall119894 = 1 119872
(3) 119878119886larr 0 forall119894 = 1 119872
(II) PHASE1 forall user 119894 from highest to lowest average SNR(4) 119882ℎ119894119897119890 (119878
119886le 119878max) 119889119900
(5) 119899lowast= argmax
119899119903119894119899
(6) 119908119894119899lowast larr 1 119878
119894= 119878119894cup 119899lowast
(7) 119904119894larr lceil
119879119894
119903119894119899
rceil
(8) 119878119886larr 119878119886+ 119904119894
(9) 119868119865 119889119887
119894ge 119863th minus Δ
(10) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(11) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(12) 119864119871119878119864
(13) 119868119865 119889119890
119894ge 119863th minus Δ
(14) 119909119890
119894larr 119909119890
119894minusmin (119903
119894119899 119909119890
119894)
(15) 119909119887
119894larr 119909119887
119894minusmax (119903
119894119899minus 119909119890
119894 0)
(16) 119864119871119878119864
(17) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(18) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(19) 119891119894= 119891119894+119875119879
119873+
1
119878119873119877119894119899lowast
(20) 119864119899119889(III) PHASE2 forall available subcarrier 119899 from 1 to119873(21) 119868119865 119908
is discarded if it fails to be transmitted before the thresholddelay bound Consequently this constraint implies that thebase layer traffic is given priority in when assigning resourcesThis constraint is employed to guarantee the QoS
Equation (8) indicates that the sum of the power allocatedto eachUE should not exceed the total available power Equa-tion (9) indicates that only one subcarrier can be allocated toa UE Equation (10) indicates that the sum of the allocated REof each UE should not exceed the total number of RErsquosThesethree constraints are used for throughput maximization
To maximize OP(119908 119901) subcarrier 119899 should be allocatedto user 119894lowast This is expressed as follows
119894lowast= argmax
119894
119903119894119899
119877119894
(11)
In addition the allocated power of user 119894lowast with subcarrier 119899is given as
119901119894lowast119899= max119891lowast
119894minus
1
SNR119894lowast119899
0 (12)
where 119891119894
lowast is the water-filling level of user 119894lowast [17]
32 Heuristic Approach Given that the scheduling problemformulated in Section 31 is a NP-hard problem we adopt a
suboptimal heuristic approach to solve this problem prac-tically The details of the heuristic algorithm are shown inAlgorithm 1
In this heuristic algorithm 119909119887119894indicates the amount of
base layer trafficwithin the buffer and119909119890119894indicates the amount
of enhancement layer traffic within the buffer 119879119894is the
requested traffic size for user 119894 and 119889119887119894and 119889119890
119894indicate the
queuing delay of base layer packets and enhancement layerpackets respectively 119878
119886is the total number of assigned RErsquos
at one time instance and Δ is the time margin for priorityadjustments
A base layer packet and an enhancement layer packet fora 3D video traffic are delivered from a 3D media server tologically separatedMAC layer buffers in eNBWhen a packetis generated and stored in the buffer the index value of thebuffer increases by one They are transmitted to the MAClayer of the destination UE when resources are allocated bythe scheduler at every slot time The heuristic algorithm inAlgorithm 1 is involved in the scheduling process
In the heuristic algorithm a user whose average SNR ishigher has priority in resource allocation So the user whohas the highest average SNR gets resource firstThe schedulerfinds and allocates the best subcarrier which maximizes 119903
119894119899
According to the userrsquos requested traffic size and allocatedsubcarrierrsquos capacity RErsquos are allocated Then the heuristic
6 The Scientific World Journal
Table 1 MCS level and data rate
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
round-robin (RR) [10] proportionally fair (PF) [11] andfast fair-throughput (FFTH) [12] scheduling algorithms arethe most commonly used types The Max CI algorithmprovides throughput maximization and the RR algorithmachieves the optimal level of fairness The PF and theFFTH algorithms provide a good balance between systemthroughput and fairness but they do not take the QoS intoaccount Several scheduling algorithms have been studied toguarantee QoS for end users The modified largest weighteddelay first (MLWDF) [13] and the exponential rule (EXP) [14]algorithms are the most popular types They consider boththe maximum allowable delay and the instantaneous channelrate However little work has been carried out to find anoptimal resource allocation scheme that considers the QoSof a 3D AV service
In this paper we propose a resource allocation scheme forwireless transmissions of 3D AV traffic We focus on a QoSguarantee for 3D AV services in the proposed scheme Todo this we adopt a concept which uses a base layer and anenhancement layer for 3DAV traffic as introduced by video-plus-depth concept This concept processes 3D traffic moreefficiently
The rest of this paper is organized as follows first wegive an overview of the system environment in which theproposed scheme is adopted in Section 2 In Section 3we introduce the proposed resource allocation scheme andformulate it as an optimization problem Then we give adetailed explanation of a suboptimal heuristic algorithm tosolve the optimization problem In Section 4 we describe thesimulation environment and evaluate the performance of theproposed scheme based on several numerical results Finallywe conclude the paper in Section 5
2 System Description
This section gives an overview of the system environment inwhich the proposed resource allocation scheme is adopted Itbegins with an introduction of the overall system architectureof 3D video over wireless systems Also a detailed descriptionof video-plus-depth 3D video is discussed
21 3D Video over Wireless Systems Figure 1 shows the over-all system architecture of 3D video over wireless systemsas considered in this paper The system consists of fourparts denoted here as a 3D media server a packet datanetwork gateway (PDN GW)serving GW an enhancednode-B (eNB)enhanced universal terrestrial radio accessnetwork (E-UTRAN) and the user equipment (UE) The 2Dvideo and its associated depthmap are obtained from a depthcamera system or with a 2D-to-3D conversionmethodThesedata are separately encoded as the base and the enhancementlayersThey are transmitted as separate streams of media datathrough a single connection Before transmitting over the IPnetwork thesemedia data are packetized into individual RTPpackets
The media server sends RTP packets to the wirelesssystem through an IP network In this paper we considerwireless systems based on the orthogonal frequency divisionmultiple access (OFDMA) such as the long-term evolution
(LTE) LTE-advanced (LTE-A) IEEE 80216e IEEE 80216mand other technologies In Figure 1 we show an exampleof the system architecture which adopts the radio accessnetwork (RAN) of the LTE-A systems As shown in the figureLTE-A system uses the PDN GWserving GW eNB and UEcomponents This system is connected to a 3D media serverthrough IP networks
The PDN GW or serving GW connects the IP networkand the LTE-A system The RTP packets of the base layerand the enhancement layer data come into the LTE-A systemthrough the general packet radio service (GPRS) tunnelingprotocol (GTP) GTP is an IP-based protocol which is usedin the universal mobile telecommunications system (UMTS)network GTP-U in Figure 1 is used to carry user data withinthe GPRS core network and between the RAN and the corenetwork The carried user data has usually IP or PPP formatIn our system the user data has IP format
eNB receives user data throughGTP-UUDPIP and thensends the data to the UE through the RAN Each instance ofdata is passed through the packet data convergence protocol(PDCP) and the radio link control (RLC) after which itis segmented and encapsulated as one or more mediumaccess control (MAC) frames and inserted into MAC bufferThe base layer traffic and the enhancement layer traffic areinserted into logically separated MAC buffers In the eNBMAC layer radio resources are allocated by schedulingAfter resource allocation by scheduling eNB repeatedly readsMAC frames from theMAC buffer and sends them to the UEthrough the RAN
Upon receiving the MAC frames the UE reassemblesand decapsulates them as RTP packets throughRLCPDCPIPUDP layers The RTP packets are combinedinto the base layer and the enhancement layer data Eachlayer of data is decoded so as to recover the depth map andthe 2D video respectively Finally they are converted into a3D video stream and displayed by the display screen of theUE
22 Video-Plus-Depth 3D Video In this paper we focus onthe video-plus-depth representation of multiview 3D videowhich renders a 3D video using a 2D video stream and itsassociated depth map This type is widely used because itprovides a flexible representation of 3D and because it is com-patible with existing coding and transmission technologies[3ndash8]
In this format the depth map includes 256 leveled greyimages It also contains depth information about the pixelpositions of the associated 2D video It can be acquired bya depth camera directly or can be extracted by a multiviewimage The size of the depth map is related to the numberof viewpoints of the 3D video Because the resource require-ments of 3D video applications based on the video-plus-depth scheme are high compared to those of 2D video appli-cations efficient compression techniques are required for 3Dvideo The layered coding approach can be effective withregard to this requirement In the layered coding approachshown in Figure 2 the 2D video and the depth imagesequence are encoded as the base layer and the enhancementlayer respectively Existing 2D compression techniques are
The Scientific World Journal 3
Wired environment(IP network)
Wireless environment
(RAN)
3D video
encoder
3D recording
2D to 3D conversion
Wired environment(IP network)
2D video ( base layer)
Depth map
(enhancementlayer)
RTP
RTCP
3D media server
RTP
RTCP
3D
video
decoder
DIBR
2D video
( base layer)
Depth map
(enhancement layer)
UE
UD
P
IP L2 L1 L2 IP
UD
P
L1
IP UD
P
L2L1
UD
PIP
UD
PIP
GTP
-U
PDCP L1L2
RLCMACL1
Display screen
PDN GWserving GW
IPL1 RLC
PDCP
MAC
RTP
RTCP
UD
P
eNB (E-UTRAN )
GTP
-U
Figure 1 System architecture of 3D AV over wireless systems
3D media server
3D traffic
DepthEnhancement
layerencoder
Base layerencoder2D video
Bit-stream
Figure 2 Layered coding approach of video-plus-depth 3D video
used to encode both the 2D video and the depth mapsequence After delivery the base and enhancement layers aredecoded as the 2D video and the depth image sequence on thereceiver side Before displaying 3D video on the display the
supplied 2D video and depth image sequences are convertedinto 3D video sequences using an image-warping techniqueknown as DIBR
3 Proposed Resource Allocation Strategy
This section gives a detailed description of the proposedresource allocation strategy adaptive to 3D video over wire-less systems First we describe the key idea of the proposedstrategy to guarantee the QoS formulating it as an opti-mization problem Next we present a suboptimal heuristicapproach to solve this problem
31 Description of the Proposed Strategy There have beenseveral definitions and measures pertaining to the QoS ofa 3D AV service [1ndash8] Among them we use continuityof service as the QoS measure of the 3D AV serviceAccordingly a consecutive frame transmission in the MAClayer is required to guarantee the QoS Moreover an efficient
4 The Scientific World Journal
resource allocation method is required due to the limitedresources in wireless networks
The key idea of the proposed resource allocation strategyis that the base layer traffic is given priority when assigningresources to guarantee the QoS In this paper we focuson 3D video traffic based on the video-plus-depth methodwhich consists of a 2D video stream and its associateddepth map As noted in Section 22 the 2D video and thedepth map are separately encoded as the base layer andthe enhancement layer respectively These two layers ofdata are transmitted together through a single connectionGiven that the enhancement layer data contains valuableinformation with which to implement the 3D video it shouldbe considered as important However the base layer datais more important than the enhancement layer data withregard to the QoS because the enhancement layer data playsa supporting role to convert the base layer data into 3D videoAssuming that eNB can transmit to a UE either the baselayer data or the enhancement layer data on account of ashortage of resource in such a case if the UE receives onlythe enhancement layer data it can no longer be providedwith video streaming service However if the UE receivesonly the base layer data it can avoid interruptions of its videostreaming service though 2D video is provided instead of3D video Therefore we give priority to the base layer trafficwhich can provide a 2D service by itself This also helps toguarantee the QoS
This idea is adopted in the scheduling procedure whichworks in the MAC layer of the eNB As described inSection 21 the base layer data and the enhancement data aredelivered from a 3D media server to the MAC layer buffer inthe eNB These two types of traffic are inserted into logicallyseparated buffersThey are transmitted to the destination (theMAC layer of the UE) when resources are allocated by thescheduler at every slot time The scheduling algorithm aimsto achieve optimal resource allocation which maximizes thethroughput of the 3DAV traffic while guaranteeing the QoSIt is formulated as described below
We assume that traffic whose destination is UE 119894 isdelivered to eNB at an average rate of packetslot time Thepackets are stored in aMAC buffer 119861
119894 For the 3DAV traffic
two buffers are assigned to the enhancement layer traffic andthe base layer traffic respectively In this paper we assumethat the buffers are large enough to not to overflow Westipulate that the amount of data in buffer 119861
119894at the beginning
of the kth slot is 119909119894119896 From the result of scheduling 119906
119894119896is
transmitted during kth slot 119906119894119896depends on the allocated data
rate at buffer 119861119894 Then the buffer is updated as
is the input traffic size in 119861119894during the kth slot
Let119863119894be the average queuing delay for 119861
119894119863119894is related to the
average buffer length via Littlersquos theorem [15] and is describedas
119863119894=1
120582119864 [119909119894119896] (2)
where 120582 = 119864[119886119894119896] is the average packet arrival rate Because
0 le 119906119894119896le 119909119894119896 the smallest average delay of the kth slot
is achieved when 119906119894119896= 119909119894119896
and the average queuing delaybecomes119863
119894= 1
Let 119866119894119899
be the channel gain 119873119894119899
the total noise powerspectral density and 119901
119894119899the allocated power for user 119894 to
subcarrier 119899 In this formulation we consider an OFDMAbased wireless system We assume that M-QAM modulationis applied with a BER requirement The signal-to-noise ratio(SNR) SNR
where Γ = minus ln(5 sdot BER)15 [16] In addition the capacity ofuser 119894 on subcarrier 119899 is normalized by
119903119894119899= ln (1 + 119901
119894119899sdot SNR
119894119899) (4)
The instantaneous data rate of user 119894 can then be described as
119877119894=
119873
sum
119899=1
119908119894119899ln (1 + 119901
119894119899SNR119894119899) (5)
and the number of resource elements (RErsquos) required tosupport 119877
119894while transmitting on subcarrier 119899 is 119904
119894 119908119894119899is the
subcarrier allocation index it has a value of 1 when subcarrier119899 is allocated to user 119894 Otherwise it has a value of 0
Because our goal is throughput maximization whileguaranteeing the QoS the scheduling problem can be writtenas shown below
Maximize OP (119908 119901) = max119908119894119899 119901119894119899
119868
sum
119894=1
ln119877119894
(6)
subject to 119863119887
119894minus 1 le 119863th 119863
119887
119894ge 1 forall119894 (7)
119868
sum
119894=1
119873
sum
119899=1
119901119894119899le 119875119879
119901119894119899ge 0 forall119894 119899 (8)
119868
sum
119894=1
119908119894119899le 1 119908
119894119899ge 0 forall119894 119899 (9)
119868
sum
119894=1
119904119894le 119878max (10)
Here 119868 is the total number of UEs and 119863119887119894is the average
queuing delay of the buffer which contains the base layertraffic for the UE i 119863th is the threshold delay and 119878max isthe total number of RErsquos within a slot 119908 = [119908
119894119899]119868 times119873
119901 =[119901119894119899]119868 times119873
and 119875119879are the total available transmission power
for eNB Note that OP(119908 119901) is neither convex nor concavewith respect to (119908 119901) Although 119908
119894119899is defined to obtain a
value of either 0 or 1 it is permitted to be a real numberbetween 0 and 1 to make the problem tractable
Equation (6) represents the total throughput computedby summing up the logarithmic user data rate assigned toeach UE There are four constraints Equation (7) indicatesthat the queuing delay for the base layer traffic should notexceed the threshold delay bound as the data in the buffer
The Scientific World Journal 5
(I) INITIALIZATION(1) 119878
119894larr 0 forall119894 = 1 119872
(2) 119891119894larr 0 forall119894 = 1 119872
(3) 119878119886larr 0 forall119894 = 1 119872
(II) PHASE1 forall user 119894 from highest to lowest average SNR(4) 119882ℎ119894119897119890 (119878
119886le 119878max) 119889119900
(5) 119899lowast= argmax
119899119903119894119899
(6) 119908119894119899lowast larr 1 119878
119894= 119878119894cup 119899lowast
(7) 119904119894larr lceil
119879119894
119903119894119899
rceil
(8) 119878119886larr 119878119886+ 119904119894
(9) 119868119865 119889119887
119894ge 119863th minus Δ
(10) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(11) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(12) 119864119871119878119864
(13) 119868119865 119889119890
119894ge 119863th minus Δ
(14) 119909119890
119894larr 119909119890
119894minusmin (119903
119894119899 119909119890
119894)
(15) 119909119887
119894larr 119909119887
119894minusmax (119903
119894119899minus 119909119890
119894 0)
(16) 119864119871119878119864
(17) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(18) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(19) 119891119894= 119891119894+119875119879
119873+
1
119878119873119877119894119899lowast
(20) 119864119899119889(III) PHASE2 forall available subcarrier 119899 from 1 to119873(21) 119868119865 119908
is discarded if it fails to be transmitted before the thresholddelay bound Consequently this constraint implies that thebase layer traffic is given priority in when assigning resourcesThis constraint is employed to guarantee the QoS
Equation (8) indicates that the sum of the power allocatedto eachUE should not exceed the total available power Equa-tion (9) indicates that only one subcarrier can be allocated toa UE Equation (10) indicates that the sum of the allocated REof each UE should not exceed the total number of RErsquosThesethree constraints are used for throughput maximization
To maximize OP(119908 119901) subcarrier 119899 should be allocatedto user 119894lowast This is expressed as follows
119894lowast= argmax
119894
119903119894119899
119877119894
(11)
In addition the allocated power of user 119894lowast with subcarrier 119899is given as
119901119894lowast119899= max119891lowast
119894minus
1
SNR119894lowast119899
0 (12)
where 119891119894
lowast is the water-filling level of user 119894lowast [17]
32 Heuristic Approach Given that the scheduling problemformulated in Section 31 is a NP-hard problem we adopt a
suboptimal heuristic approach to solve this problem prac-tically The details of the heuristic algorithm are shown inAlgorithm 1
In this heuristic algorithm 119909119887119894indicates the amount of
base layer trafficwithin the buffer and119909119890119894indicates the amount
of enhancement layer traffic within the buffer 119879119894is the
requested traffic size for user 119894 and 119889119887119894and 119889119890
119894indicate the
queuing delay of base layer packets and enhancement layerpackets respectively 119878
119886is the total number of assigned RErsquos
at one time instance and Δ is the time margin for priorityadjustments
A base layer packet and an enhancement layer packet fora 3D video traffic are delivered from a 3D media server tologically separatedMAC layer buffers in eNBWhen a packetis generated and stored in the buffer the index value of thebuffer increases by one They are transmitted to the MAClayer of the destination UE when resources are allocated bythe scheduler at every slot time The heuristic algorithm inAlgorithm 1 is involved in the scheduling process
In the heuristic algorithm a user whose average SNR ishigher has priority in resource allocation So the user whohas the highest average SNR gets resource firstThe schedulerfinds and allocates the best subcarrier which maximizes 119903
119894119899
According to the userrsquos requested traffic size and allocatedsubcarrierrsquos capacity RErsquos are allocated Then the heuristic
6 The Scientific World Journal
Table 1 MCS level and data rate
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
Figure 1 System architecture of 3D AV over wireless systems
3D media server
3D traffic
DepthEnhancement
layerencoder
Base layerencoder2D video
Bit-stream
Figure 2 Layered coding approach of video-plus-depth 3D video
used to encode both the 2D video and the depth mapsequence After delivery the base and enhancement layers aredecoded as the 2D video and the depth image sequence on thereceiver side Before displaying 3D video on the display the
supplied 2D video and depth image sequences are convertedinto 3D video sequences using an image-warping techniqueknown as DIBR
3 Proposed Resource Allocation Strategy
This section gives a detailed description of the proposedresource allocation strategy adaptive to 3D video over wire-less systems First we describe the key idea of the proposedstrategy to guarantee the QoS formulating it as an opti-mization problem Next we present a suboptimal heuristicapproach to solve this problem
31 Description of the Proposed Strategy There have beenseveral definitions and measures pertaining to the QoS ofa 3D AV service [1ndash8] Among them we use continuityof service as the QoS measure of the 3D AV serviceAccordingly a consecutive frame transmission in the MAClayer is required to guarantee the QoS Moreover an efficient
4 The Scientific World Journal
resource allocation method is required due to the limitedresources in wireless networks
The key idea of the proposed resource allocation strategyis that the base layer traffic is given priority when assigningresources to guarantee the QoS In this paper we focuson 3D video traffic based on the video-plus-depth methodwhich consists of a 2D video stream and its associateddepth map As noted in Section 22 the 2D video and thedepth map are separately encoded as the base layer andthe enhancement layer respectively These two layers ofdata are transmitted together through a single connectionGiven that the enhancement layer data contains valuableinformation with which to implement the 3D video it shouldbe considered as important However the base layer datais more important than the enhancement layer data withregard to the QoS because the enhancement layer data playsa supporting role to convert the base layer data into 3D videoAssuming that eNB can transmit to a UE either the baselayer data or the enhancement layer data on account of ashortage of resource in such a case if the UE receives onlythe enhancement layer data it can no longer be providedwith video streaming service However if the UE receivesonly the base layer data it can avoid interruptions of its videostreaming service though 2D video is provided instead of3D video Therefore we give priority to the base layer trafficwhich can provide a 2D service by itself This also helps toguarantee the QoS
This idea is adopted in the scheduling procedure whichworks in the MAC layer of the eNB As described inSection 21 the base layer data and the enhancement data aredelivered from a 3D media server to the MAC layer buffer inthe eNB These two types of traffic are inserted into logicallyseparated buffersThey are transmitted to the destination (theMAC layer of the UE) when resources are allocated by thescheduler at every slot time The scheduling algorithm aimsto achieve optimal resource allocation which maximizes thethroughput of the 3DAV traffic while guaranteeing the QoSIt is formulated as described below
We assume that traffic whose destination is UE 119894 isdelivered to eNB at an average rate of packetslot time Thepackets are stored in aMAC buffer 119861
119894 For the 3DAV traffic
two buffers are assigned to the enhancement layer traffic andthe base layer traffic respectively In this paper we assumethat the buffers are large enough to not to overflow Westipulate that the amount of data in buffer 119861
119894at the beginning
of the kth slot is 119909119894119896 From the result of scheduling 119906
119894119896is
transmitted during kth slot 119906119894119896depends on the allocated data
rate at buffer 119861119894 Then the buffer is updated as
is the input traffic size in 119861119894during the kth slot
Let119863119894be the average queuing delay for 119861
119894119863119894is related to the
average buffer length via Littlersquos theorem [15] and is describedas
119863119894=1
120582119864 [119909119894119896] (2)
where 120582 = 119864[119886119894119896] is the average packet arrival rate Because
0 le 119906119894119896le 119909119894119896 the smallest average delay of the kth slot
is achieved when 119906119894119896= 119909119894119896
and the average queuing delaybecomes119863
119894= 1
Let 119866119894119899
be the channel gain 119873119894119899
the total noise powerspectral density and 119901
119894119899the allocated power for user 119894 to
subcarrier 119899 In this formulation we consider an OFDMAbased wireless system We assume that M-QAM modulationis applied with a BER requirement The signal-to-noise ratio(SNR) SNR
where Γ = minus ln(5 sdot BER)15 [16] In addition the capacity ofuser 119894 on subcarrier 119899 is normalized by
119903119894119899= ln (1 + 119901
119894119899sdot SNR
119894119899) (4)
The instantaneous data rate of user 119894 can then be described as
119877119894=
119873
sum
119899=1
119908119894119899ln (1 + 119901
119894119899SNR119894119899) (5)
and the number of resource elements (RErsquos) required tosupport 119877
119894while transmitting on subcarrier 119899 is 119904
119894 119908119894119899is the
subcarrier allocation index it has a value of 1 when subcarrier119899 is allocated to user 119894 Otherwise it has a value of 0
Because our goal is throughput maximization whileguaranteeing the QoS the scheduling problem can be writtenas shown below
Maximize OP (119908 119901) = max119908119894119899 119901119894119899
119868
sum
119894=1
ln119877119894
(6)
subject to 119863119887
119894minus 1 le 119863th 119863
119887
119894ge 1 forall119894 (7)
119868
sum
119894=1
119873
sum
119899=1
119901119894119899le 119875119879
119901119894119899ge 0 forall119894 119899 (8)
119868
sum
119894=1
119908119894119899le 1 119908
119894119899ge 0 forall119894 119899 (9)
119868
sum
119894=1
119904119894le 119878max (10)
Here 119868 is the total number of UEs and 119863119887119894is the average
queuing delay of the buffer which contains the base layertraffic for the UE i 119863th is the threshold delay and 119878max isthe total number of RErsquos within a slot 119908 = [119908
119894119899]119868 times119873
119901 =[119901119894119899]119868 times119873
and 119875119879are the total available transmission power
for eNB Note that OP(119908 119901) is neither convex nor concavewith respect to (119908 119901) Although 119908
119894119899is defined to obtain a
value of either 0 or 1 it is permitted to be a real numberbetween 0 and 1 to make the problem tractable
Equation (6) represents the total throughput computedby summing up the logarithmic user data rate assigned toeach UE There are four constraints Equation (7) indicatesthat the queuing delay for the base layer traffic should notexceed the threshold delay bound as the data in the buffer
The Scientific World Journal 5
(I) INITIALIZATION(1) 119878
119894larr 0 forall119894 = 1 119872
(2) 119891119894larr 0 forall119894 = 1 119872
(3) 119878119886larr 0 forall119894 = 1 119872
(II) PHASE1 forall user 119894 from highest to lowest average SNR(4) 119882ℎ119894119897119890 (119878
119886le 119878max) 119889119900
(5) 119899lowast= argmax
119899119903119894119899
(6) 119908119894119899lowast larr 1 119878
119894= 119878119894cup 119899lowast
(7) 119904119894larr lceil
119879119894
119903119894119899
rceil
(8) 119878119886larr 119878119886+ 119904119894
(9) 119868119865 119889119887
119894ge 119863th minus Δ
(10) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(11) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(12) 119864119871119878119864
(13) 119868119865 119889119890
119894ge 119863th minus Δ
(14) 119909119890
119894larr 119909119890
119894minusmin (119903
119894119899 119909119890
119894)
(15) 119909119887
119894larr 119909119887
119894minusmax (119903
119894119899minus 119909119890
119894 0)
(16) 119864119871119878119864
(17) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(18) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(19) 119891119894= 119891119894+119875119879
119873+
1
119878119873119877119894119899lowast
(20) 119864119899119889(III) PHASE2 forall available subcarrier 119899 from 1 to119873(21) 119868119865 119908
is discarded if it fails to be transmitted before the thresholddelay bound Consequently this constraint implies that thebase layer traffic is given priority in when assigning resourcesThis constraint is employed to guarantee the QoS
Equation (8) indicates that the sum of the power allocatedto eachUE should not exceed the total available power Equa-tion (9) indicates that only one subcarrier can be allocated toa UE Equation (10) indicates that the sum of the allocated REof each UE should not exceed the total number of RErsquosThesethree constraints are used for throughput maximization
To maximize OP(119908 119901) subcarrier 119899 should be allocatedto user 119894lowast This is expressed as follows
119894lowast= argmax
119894
119903119894119899
119877119894
(11)
In addition the allocated power of user 119894lowast with subcarrier 119899is given as
119901119894lowast119899= max119891lowast
119894minus
1
SNR119894lowast119899
0 (12)
where 119891119894
lowast is the water-filling level of user 119894lowast [17]
32 Heuristic Approach Given that the scheduling problemformulated in Section 31 is a NP-hard problem we adopt a
suboptimal heuristic approach to solve this problem prac-tically The details of the heuristic algorithm are shown inAlgorithm 1
In this heuristic algorithm 119909119887119894indicates the amount of
base layer trafficwithin the buffer and119909119890119894indicates the amount
of enhancement layer traffic within the buffer 119879119894is the
requested traffic size for user 119894 and 119889119887119894and 119889119890
119894indicate the
queuing delay of base layer packets and enhancement layerpackets respectively 119878
119886is the total number of assigned RErsquos
at one time instance and Δ is the time margin for priorityadjustments
A base layer packet and an enhancement layer packet fora 3D video traffic are delivered from a 3D media server tologically separatedMAC layer buffers in eNBWhen a packetis generated and stored in the buffer the index value of thebuffer increases by one They are transmitted to the MAClayer of the destination UE when resources are allocated bythe scheduler at every slot time The heuristic algorithm inAlgorithm 1 is involved in the scheduling process
In the heuristic algorithm a user whose average SNR ishigher has priority in resource allocation So the user whohas the highest average SNR gets resource firstThe schedulerfinds and allocates the best subcarrier which maximizes 119903
119894119899
According to the userrsquos requested traffic size and allocatedsubcarrierrsquos capacity RErsquos are allocated Then the heuristic
6 The Scientific World Journal
Table 1 MCS level and data rate
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
resource allocation method is required due to the limitedresources in wireless networks
The key idea of the proposed resource allocation strategyis that the base layer traffic is given priority when assigningresources to guarantee the QoS In this paper we focuson 3D video traffic based on the video-plus-depth methodwhich consists of a 2D video stream and its associateddepth map As noted in Section 22 the 2D video and thedepth map are separately encoded as the base layer andthe enhancement layer respectively These two layers ofdata are transmitted together through a single connectionGiven that the enhancement layer data contains valuableinformation with which to implement the 3D video it shouldbe considered as important However the base layer datais more important than the enhancement layer data withregard to the QoS because the enhancement layer data playsa supporting role to convert the base layer data into 3D videoAssuming that eNB can transmit to a UE either the baselayer data or the enhancement layer data on account of ashortage of resource in such a case if the UE receives onlythe enhancement layer data it can no longer be providedwith video streaming service However if the UE receivesonly the base layer data it can avoid interruptions of its videostreaming service though 2D video is provided instead of3D video Therefore we give priority to the base layer trafficwhich can provide a 2D service by itself This also helps toguarantee the QoS
This idea is adopted in the scheduling procedure whichworks in the MAC layer of the eNB As described inSection 21 the base layer data and the enhancement data aredelivered from a 3D media server to the MAC layer buffer inthe eNB These two types of traffic are inserted into logicallyseparated buffersThey are transmitted to the destination (theMAC layer of the UE) when resources are allocated by thescheduler at every slot time The scheduling algorithm aimsto achieve optimal resource allocation which maximizes thethroughput of the 3DAV traffic while guaranteeing the QoSIt is formulated as described below
We assume that traffic whose destination is UE 119894 isdelivered to eNB at an average rate of packetslot time Thepackets are stored in aMAC buffer 119861
119894 For the 3DAV traffic
two buffers are assigned to the enhancement layer traffic andthe base layer traffic respectively In this paper we assumethat the buffers are large enough to not to overflow Westipulate that the amount of data in buffer 119861
119894at the beginning
of the kth slot is 119909119894119896 From the result of scheduling 119906
119894119896is
transmitted during kth slot 119906119894119896depends on the allocated data
rate at buffer 119861119894 Then the buffer is updated as
is the input traffic size in 119861119894during the kth slot
Let119863119894be the average queuing delay for 119861
119894119863119894is related to the
average buffer length via Littlersquos theorem [15] and is describedas
119863119894=1
120582119864 [119909119894119896] (2)
where 120582 = 119864[119886119894119896] is the average packet arrival rate Because
0 le 119906119894119896le 119909119894119896 the smallest average delay of the kth slot
is achieved when 119906119894119896= 119909119894119896
and the average queuing delaybecomes119863
119894= 1
Let 119866119894119899
be the channel gain 119873119894119899
the total noise powerspectral density and 119901
119894119899the allocated power for user 119894 to
subcarrier 119899 In this formulation we consider an OFDMAbased wireless system We assume that M-QAM modulationis applied with a BER requirement The signal-to-noise ratio(SNR) SNR
where Γ = minus ln(5 sdot BER)15 [16] In addition the capacity ofuser 119894 on subcarrier 119899 is normalized by
119903119894119899= ln (1 + 119901
119894119899sdot SNR
119894119899) (4)
The instantaneous data rate of user 119894 can then be described as
119877119894=
119873
sum
119899=1
119908119894119899ln (1 + 119901
119894119899SNR119894119899) (5)
and the number of resource elements (RErsquos) required tosupport 119877
119894while transmitting on subcarrier 119899 is 119904
119894 119908119894119899is the
subcarrier allocation index it has a value of 1 when subcarrier119899 is allocated to user 119894 Otherwise it has a value of 0
Because our goal is throughput maximization whileguaranteeing the QoS the scheduling problem can be writtenas shown below
Maximize OP (119908 119901) = max119908119894119899 119901119894119899
119868
sum
119894=1
ln119877119894
(6)
subject to 119863119887
119894minus 1 le 119863th 119863
119887
119894ge 1 forall119894 (7)
119868
sum
119894=1
119873
sum
119899=1
119901119894119899le 119875119879
119901119894119899ge 0 forall119894 119899 (8)
119868
sum
119894=1
119908119894119899le 1 119908
119894119899ge 0 forall119894 119899 (9)
119868
sum
119894=1
119904119894le 119878max (10)
Here 119868 is the total number of UEs and 119863119887119894is the average
queuing delay of the buffer which contains the base layertraffic for the UE i 119863th is the threshold delay and 119878max isthe total number of RErsquos within a slot 119908 = [119908
119894119899]119868 times119873
119901 =[119901119894119899]119868 times119873
and 119875119879are the total available transmission power
for eNB Note that OP(119908 119901) is neither convex nor concavewith respect to (119908 119901) Although 119908
119894119899is defined to obtain a
value of either 0 or 1 it is permitted to be a real numberbetween 0 and 1 to make the problem tractable
Equation (6) represents the total throughput computedby summing up the logarithmic user data rate assigned toeach UE There are four constraints Equation (7) indicatesthat the queuing delay for the base layer traffic should notexceed the threshold delay bound as the data in the buffer
The Scientific World Journal 5
(I) INITIALIZATION(1) 119878
119894larr 0 forall119894 = 1 119872
(2) 119891119894larr 0 forall119894 = 1 119872
(3) 119878119886larr 0 forall119894 = 1 119872
(II) PHASE1 forall user 119894 from highest to lowest average SNR(4) 119882ℎ119894119897119890 (119878
119886le 119878max) 119889119900
(5) 119899lowast= argmax
119899119903119894119899
(6) 119908119894119899lowast larr 1 119878
119894= 119878119894cup 119899lowast
(7) 119904119894larr lceil
119879119894
119903119894119899
rceil
(8) 119878119886larr 119878119886+ 119904119894
(9) 119868119865 119889119887
119894ge 119863th minus Δ
(10) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(11) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(12) 119864119871119878119864
(13) 119868119865 119889119890
119894ge 119863th minus Δ
(14) 119909119890
119894larr 119909119890
119894minusmin (119903
119894119899 119909119890
119894)
(15) 119909119887
119894larr 119909119887
119894minusmax (119903
119894119899minus 119909119890
119894 0)
(16) 119864119871119878119864
(17) 119909119887
119894larr 119909119887
119894minusmin (119903
119894119899 119909119887
119894)
(18) 119909119890
119894larr 119909119890
119894minusmax (119903
119894119899minus 119909119887
119894 0)
(19) 119891119894= 119891119894+119875119879
119873+
1
119878119873119877119894119899lowast
(20) 119864119899119889(III) PHASE2 forall available subcarrier 119899 from 1 to119873(21) 119868119865 119908
is discarded if it fails to be transmitted before the thresholddelay bound Consequently this constraint implies that thebase layer traffic is given priority in when assigning resourcesThis constraint is employed to guarantee the QoS
Equation (8) indicates that the sum of the power allocatedto eachUE should not exceed the total available power Equa-tion (9) indicates that only one subcarrier can be allocated toa UE Equation (10) indicates that the sum of the allocated REof each UE should not exceed the total number of RErsquosThesethree constraints are used for throughput maximization
To maximize OP(119908 119901) subcarrier 119899 should be allocatedto user 119894lowast This is expressed as follows
119894lowast= argmax
119894
119903119894119899
119877119894
(11)
In addition the allocated power of user 119894lowast with subcarrier 119899is given as
119901119894lowast119899= max119891lowast
119894minus
1
SNR119894lowast119899
0 (12)
where 119891119894
lowast is the water-filling level of user 119894lowast [17]
32 Heuristic Approach Given that the scheduling problemformulated in Section 31 is a NP-hard problem we adopt a
suboptimal heuristic approach to solve this problem prac-tically The details of the heuristic algorithm are shown inAlgorithm 1
In this heuristic algorithm 119909119887119894indicates the amount of
base layer trafficwithin the buffer and119909119890119894indicates the amount
of enhancement layer traffic within the buffer 119879119894is the
requested traffic size for user 119894 and 119889119887119894and 119889119890
119894indicate the
queuing delay of base layer packets and enhancement layerpackets respectively 119878
119886is the total number of assigned RErsquos
at one time instance and Δ is the time margin for priorityadjustments
A base layer packet and an enhancement layer packet fora 3D video traffic are delivered from a 3D media server tologically separatedMAC layer buffers in eNBWhen a packetis generated and stored in the buffer the index value of thebuffer increases by one They are transmitted to the MAClayer of the destination UE when resources are allocated bythe scheduler at every slot time The heuristic algorithm inAlgorithm 1 is involved in the scheduling process
In the heuristic algorithm a user whose average SNR ishigher has priority in resource allocation So the user whohas the highest average SNR gets resource firstThe schedulerfinds and allocates the best subcarrier which maximizes 119903
119894119899
According to the userrsquos requested traffic size and allocatedsubcarrierrsquos capacity RErsquos are allocated Then the heuristic
6 The Scientific World Journal
Table 1 MCS level and data rate
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
is discarded if it fails to be transmitted before the thresholddelay bound Consequently this constraint implies that thebase layer traffic is given priority in when assigning resourcesThis constraint is employed to guarantee the QoS
Equation (8) indicates that the sum of the power allocatedto eachUE should not exceed the total available power Equa-tion (9) indicates that only one subcarrier can be allocated toa UE Equation (10) indicates that the sum of the allocated REof each UE should not exceed the total number of RErsquosThesethree constraints are used for throughput maximization
To maximize OP(119908 119901) subcarrier 119899 should be allocatedto user 119894lowast This is expressed as follows
119894lowast= argmax
119894
119903119894119899
119877119894
(11)
In addition the allocated power of user 119894lowast with subcarrier 119899is given as
119901119894lowast119899= max119891lowast
119894minus
1
SNR119894lowast119899
0 (12)
where 119891119894
lowast is the water-filling level of user 119894lowast [17]
32 Heuristic Approach Given that the scheduling problemformulated in Section 31 is a NP-hard problem we adopt a
suboptimal heuristic approach to solve this problem prac-tically The details of the heuristic algorithm are shown inAlgorithm 1
In this heuristic algorithm 119909119887119894indicates the amount of
base layer trafficwithin the buffer and119909119890119894indicates the amount
of enhancement layer traffic within the buffer 119879119894is the
requested traffic size for user 119894 and 119889119887119894and 119889119890
119894indicate the
queuing delay of base layer packets and enhancement layerpackets respectively 119878
119886is the total number of assigned RErsquos
at one time instance and Δ is the time margin for priorityadjustments
A base layer packet and an enhancement layer packet fora 3D video traffic are delivered from a 3D media server tologically separatedMAC layer buffers in eNBWhen a packetis generated and stored in the buffer the index value of thebuffer increases by one They are transmitted to the MAClayer of the destination UE when resources are allocated bythe scheduler at every slot time The heuristic algorithm inAlgorithm 1 is involved in the scheduling process
In the heuristic algorithm a user whose average SNR ishigher has priority in resource allocation So the user whohas the highest average SNR gets resource firstThe schedulerfinds and allocates the best subcarrier which maximizes 119903
119894119899
According to the userrsquos requested traffic size and allocatedsubcarrierrsquos capacity RErsquos are allocated Then the heuristic
6 The Scientific World Journal
Table 1 MCS level and data rate
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
MCS level Min le SNR leMax Modulation scheme Coding rate Data rate (bitslot)M1 SNR lt 2 db BPSK 14 4375M2 2 db le SNR lt 5 db BPSK 12 875M3 2 db le SNR lt 5 db BPSK 34 13125M4 2 db le SNR lt 5 db QPSK 12 175M5 2 db le SNR lt 5 db QPSK 58 21875M6 2 db le SNR lt 5 db QPSK 34 2625M7 2 db le SNR lt 5 db QPSK 78 30625M8 2 db le SNR lt 5 db 16QAM 12 350M9 2 db le SNR lt 5 db 16QAM 916 39375
algorithm decides how to share the allocated RErsquos betweenbuffered traffic
The basic rule is that the base layer traffic is processedfirst after which the enhancement layer traffic is handled onlywhen there are remaining resources However we can adjustthe level of priority for the base layer traffic by adjustingthe value of Δ If the queuing delay of the base layer packetdoes not exceed the adjusted threshold delay bound and thequeuing delay of the enhancement layer packet exceeds theadjusted threshold delay bound resources are used for thetransmission of the enhancement layer packet If we use ahigher value of Δ higher priority to the base layer traffic isgiven Δ has a value between 1 and119863th
After allocating resources for the subcarrier and the REthe power is also assigned Assuming that all subcarriers haveequal power 119901
119894119899= 119875119879119873 for any 119894 and 119899 In this case the total
and substitute this into (12) we can obtain 119875119894119899[17]
4 Performance Evaluation
To evaluate the performance of the proposed method asimulation is carried out In this section we describe the sim-ulation environment first and then evaluate the simulationresults using three performance measures
41 Simulation Environment To evaluate the proposed strat-egy we perform a series of simulations In the simulations weconsider a single cell model which consists of one eNB and a
number of UEsWe assume that the cell radius is 1 km and thecarrier frequency of eNB is 19 GHz We also assume that thetotal available transmit power of eNB is 10W and the noisepower is minus100 dB The path loss model given below is usedfor the channel between eNB and the UE [18]
119871 = 1281 + 371 log119889 (dB) (15)
Shadowing is lognormally distributed with a mean of 0 dBand a standard deviation of 8 dB The target BER is assumedto be 10minus4
Because we consider an OFDMA based wireless systemwe also set several OFDMA parameters The system band-width is assumed to be 10MHz and the slot time is set to05ms We adopt the channel structure of the 3GPP Ericssonmodel [19 20] To mitigate the long simulation time weassume that the maximum number of RErsquos during a slot 119878maxis 12 and the number of subcarriers is 6 The modulationmethod and coding rate are changed according to the SNRThe modulation and coding scheme (MCS) has nine levelsas shown in Table 1
UEs are uniformly distributed in a cell and each UE isserved only one type of traffic In the simulation we use twotraffic models 3D AV traffic models and web traffic modelsWe assume that 10 of the UEs are served WEB services and90 of UEs are served the 3D AV service
The traffic model of 3D AV used in this simulation isformulated based on the streaming video trafficmodel whichis used for IEEE 802 systems [21] Because the base layer trafficof 3D AV is identical to the characteristics of the streaming2D video traffic we use the streaming video traffic modelas the basis of the 3D traffic model We then append theenhancement layer traffic model which is changed accordingto the number of 3D viewpoints In this 3DAV trafficmodeleach frame consists of a constant number of packets Thepacket size and its arrival time within one frame are definedas a truncated Pareto distribution Figure 3 shows the 3DAVtraffic model and Table 2 shows its parameter values
In Figure 3 119879 represents the interarrival time betweenthe beginnings of each frame and the packet coding delayis determined by the interarrival time between packets in aframe as shown in Table 2The parameter 120573 of the packet sizeof the enhancement layer changes according to the numberof 3D viewpoints For 8-viewpoint 3D video traffic 120573 is 045
The Scientific World Journal 7
Table 2 Parameters of the 3D AV traffic model
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
Information ParameterInterarrival time between the beginnings of each frame (ms) (Deterministic) 100Number of packets in a frame (Deterministic) 8Packet size of the base layer (byte) Truncated Pareto (mean = 50 max = 125) 119870 = 20 120572 = 12Packet size of the enhancement layer (byte) (Packet size of the base layer)lowast120573 120573 = 045 09 18Interarrival time between packets in a frame (ms) Truncated Pareto (mean = 6 max = 125)119870 = 25 120572 = 12
Buffering window Packetcoding delay
Packet size0 T 2T (K minus 1)T KT
middot middot middot
middot middot middot
Figure 3 3D AV traffic model
[22] For 16-viewpoint and 32-viewpoint 3D video traffic 120573 is09 and 18 respectively
WEB traffic has a form similar to the ONOFF modelWEB traffic consists of the main object comprising the webpage and several embedded objects They are transmitted if aweb page is requested Table 3 shows the WEB traffic model
We also assume that the buffer of UE has infinite capacityThe maximum delay bound of 3D AV frame 119863th is set to250ms and the time margin for priority adjustment Δ is alsoset to 250ms
42 Numerical Results The performances of the proposedstrategy are evaluated through the packet drop rate theservice success rate and the QoS level Figures 4 5 and 6show the packet drop rate when 8-viewpoint 3D video 16-viewpoint 3D video and 32-viewpoint 3D video are servedrespectively A packet in a buffer drops when it is unable tobe transmitted within the maximum delay bound In thesefigures the proposed strategy is compared with an existingstrategy A similar scheduling algorithm is used in both theexisting strategy and the proposed strategy but they handle3D traffic in different ways The proposed strategy placeshigher priority on the base layer trafficOn the other hand theexisting strategy treats the base and the enhancement layertraffic in the same manner In the figures the label ldquoExistingrdquoindicates the packet drop rate of 3D traffic when the existingstrategy is employed The label ldquoProposed (B)rdquo indicates thepacket drop rate of the base layer traffic when the proposedstrategy is employed The label ldquoProposed (E)rdquo indicates the
20 30 40 6050 70 80 900
100
20
40
60
80
Existing
100
Proposed (B)Proposed (E)
Pack
et d
rop
rate
()
Number of UEs
Figure 4 Packet drop rate of 8-viewpoint 3DAV traffic (120573 = 045)
packet drop rate of the enhancement layer traffic when theproposed strategy is employed
As shown in Figures 4ndash6 the base layer traffic has alower packet drop rate compared to the enhancement layertraffic In addition the packet drop rate of the enhancementlayer traffic increases as the ratio of the enhancement layertraffic becomes higher On the other hand the packet drop
8 The Scientific World Journal
Table 3 WEB traffic parameters
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
Information ParameterMain object size (byte) Truncated Lognormal (mean = 10710 min = 100 max = 2000000)Embedded object size (byte) Truncated Lognormal (mean = 7758 min = 50 max = 2000000)Number of pages Truncated Pareto (mean = 564 max = 53)119870 = 2 120572 = 11119898 = 55Reading time (s) Exponential (mean = 30)Parsing time (s) Exponential (mean = 013)
3020 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 5 Packet drop rate of 16-viewpoint 3D AV traffic (120573 = 09)
rate of the base layer traffic shows little change because theproposed strategy prioritizes the base layer traffic In factthe packet drop rate of the existing strategy is similar to thesum of the packet drop rates of the base layer traffic and theenhancement layer traffic as the same level of resources isallocated regardless of whether it uses the proposed strategyor the existing strategy
Figures 7 8 and 9 show the service success rate of theproposed strategy We defined the service success rate as therate received by an end user successfully for a type of serviceConsequently the service success rate is closely related to theQoS In the figures the label ldquoExistingrdquo indicates the rate atwhich 3D service is successfully served when the existingstrategy is employed The label ldquoProposed (3D)rdquo indicatesthe rate at which 3D service is successfully served when theproposed strategy is employed indicating that both the baselayer traffic and the enhancement layer traffic are successfullytransmitted to the UE and the end user can watch 3D videoThe label ldquoProposed (2D)rdquo indicates the rate at which 2Dservice is successfully served when the proposed strategy isemployed Thus only the base layer traffic is successfullytransmitted to the UE and the end user can only watch 2Dvideo
Therefore it is considered as ldquoProposed (3D)rdquo if both thebase layer traffic and the enhancement layer traffic are welltransmitted However it is considered as ldquoProposed (2D)rdquo
3020 40 50 60 70 80 90 1000
20
40
60
80
100
ExistingProposed (B)Proposed (E)
Number of UEs
Pack
et d
rop
rate
()
Figure 6 Packet drop rate of 32-viewpoint 3D AV traffic (120573 = 18)
if the enhancement layer traffic cannot be delivered dueto a lack of resources and if only the base layer traffic istransmitted well It is regarded as a service failure if both thebase layer traffic and the enhancement layer traffic cannot bedelivered or if only the enhancement layer traffic is deliveredwell
When there are a small number of users nearly all userscan receive both the base layer traffic and the enhancementlayer traffic successfully As the number of users increasesthe number of users who can receive only the base layertraffic increases because the proposed strategy places priorityon the base layer traffic Therefore the service success rateof ldquoProposed (2D)rdquo increases and the service success rate ofldquoProposed (3D)rdquo decreases with an increase in the number ofusers
The figures show that the 3D service success rate ofthe proposed strategy is similar to that of the existingstrategy However if we assume that 2D video watching isalso regarded as successful service the proposed strategyoutperforms the existing strategy The meaning of theseresults can be clarified when we consider them as follows let1205953D be the success rate of the 3D service and 120595
2D the successrate of the 2D service The QoS level 120603 is then defined as
120603 =
1205953D + 120596119902 sdot 1205952D
100 (16)
The Scientific World Journal 9
20 30 40 50 60 70 80 900
100
20
40
60
80
100
Serv
ice s
ucce
ss ra
te (
)
ExistingProposed (3D)Proposed (2D)
Number of UEs
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
Figure 7 Service success rate of 8-viewpoint 3D AV traffic (120573 =045)
20 30 40 50 60 8070 900
20
100
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)
Figure 8 Service success rate of 16-viewpoint 3D AV traffic (120573 =09)
where 120596119902is the userrsquos satisfaction weight for the 2D service
compared to that for the 3D service 120596119902has a value between
0 and 1 120596119902= 1 indicates that a 3D user is fully satisfied even
when watching 2D video instead of 3D video Additionally120596119902is zero when a 3D user is fully disappointed if he watches
2D video instead of 3D video The QoS level also has avalue between 0 and 1 A value of 1 indicates a state ofsatisfaction and a value of 0 indicates a state of dissatisfactionIn Figures 10 11 and 12 the pause-less and flawless executionof not only 3D video but also 2D video are regarded as asuccessful service From the figures the QoS level of the
20 30 40 50 60 70 80 100900
20
40
60
80
100
ExistingProposed (3D)Proposed (2D)
Number of UEs
Serv
ice s
ucce
ss ra
te (
)Figure 9 Service success rate of 32-viewpoint 3D AV traffic (120573 =18)
20 30 40 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 10 QoS level of 8-viewpoint 3D AV traffic (120573 = 045)
proposed strategy outperforms that of the existing strategyAs these strategies allocate similar amounts of resourcesto 3D traffic we can conclude that the proposed strategyutilizes the resources more efficientlymdashin terms of the QoSThis result stems the use of the base layer traffic whichcontains 2D video information Successful transmission ofthis information can improve the QoS level Of course ifthe enhancement layer traffic and the base layer traffic aresuccessfully transmitted together the user will be completelysatisfied However complete transmission of the base layerinformation alone can also improve the QoS level
From these numerical results it is clear that the proposedstrategy can guarantee better QoS compared to the existing
10 The Scientific World Journal
20 4030 50 60 70 80 90 10000
02
04
06
08
10
QoS
leve
l
Existing
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
Figure 11 QoS level of 16-viewpoint 3D AV traffic (120573 = 09)
20 30 40 50 60 70 80 9000
100
02
04
06
08
QoS
leve
l
Number of UEs
Proposed (120596q = 0)Proposed (120596q = 1)
Existing
10
Figure 12 QoS level of 32-viewpoint 3D AV traffic (120573 = 18)
strategy The advantage of the proposed strategy becomesmore evident as the rate of the enhancement layer trafficincreases or when the number of viewpoints of the 3D trafficgrows
5 Conclusions
In this paper we propose a novel resource allocation strategyfor 3D video over a wireless system to guarantee the QoSThe proposed strategy focuses on the relationship between2D video and the depth map and handles them with differentpriorities Performance evaluations show that our strategyis a good choice to guarantee better QoS In terms ofseveral performance measures such as the packet drop rate
the service success rate and the QoS level our strategy isshown to outperform an existing strategy Moreover theadvantage of the proposed strategy increases as the number ofviewpoints of 3D video is increasedTherefore we expect thatthe proposed strategy can be very useful for more realistic 3Dtraffic delivery
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
This research was supported by Basic Science ResearchProgram through the National Research Foundation ofKorea funded by the Ministry of Education Science andTechnology (2010-0005684) and by Business for CooperativeRampD between Industry Academy and Research Institutefunded Korea Small andMediumBusiness Administration in2014 (C0191516) and by the Research Grant of KwangwoonUniversity in 2012 The authors express their thanks to thereviewers who checked their paper
References
[1] A Kubota A SmolicMMagnorM Tanimoto T Chen andCZhang ldquoMultiview imaging and 3DTVrdquo IEEE Signal ProcessingMagazine vol 24 no 6 pp 10ndash21 2007
[2] Y Wang H Wang and C Wang ldquoGraph-based authentica-tion design for color-depth-based 3D video transmission overwireless networksrdquo IEEE Transactions on Network and ServiceManagement vol 10 pp 245ndash254 2013
[3] C Fehn K Hopf and Q Quante ldquoKey technologies for anadvanced 3D-TV systemrdquo inThree-Dimensional TV Video andDisplay III Proceedings of SPIE pp 66ndash80 Philadelphia PaUSA October 2004
[4] A M Tekalp E Kurutepe and M R Civanlar ldquo3DTV over IPrdquoIEEE Signal Processing Magazine vol 24 no 6 pp 77ndash87 2007
[5] J Choi D Min and K Sohn ldquoReliability-based multiviewdepth enhancement considering interview coherencerdquo IEEETransactions on Circuits and Systems for Video Technology vol24 pp 603ndash616 2014
[6] M Schmeing and X Jiang ldquoDepth image based rendering afaithful approach for the disocclusion problemrdquo in Proceedingsof the True VisionmdashCapture Transmission and Display of 3DVideo (3DTV-CON rsquo10) pp 1ndash4 Tampere Finland June 2010
[7] I Ahn and C Kim ldquoA novel depth-based virtual view synthesismethod for free viewpoint videordquo IEEE Transactions on Broad-casting vol 59 no 4 pp 614ndash626 2013
[8] A Redert M Op de Beeck C Fehn et al ldquoATTESTmdashadvancedthree-dimensional television system technologiesrdquo in Proceed-ings of the International Symposium on 3D Data Processing pp313ndash319 2002
[9] R Knopp and P A Humblet ldquoInformation capacity and powercontrol in single-cell multiuser communicationsrdquo in Proceed-ings of the IEEE International Conference on Communications(ICC rsquo95) pp 331ndash335 June 1995
The Scientific World Journal 11
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005
[10] CM Aras J F Kurose D S Reeves andH Schulzrinne ldquoReal-time communication in packet-switched networksrdquoProceedingsof the IEEE vol 82 no 1 pp 122ndash139 1994
[11] A Jalali R Padovani and R Pankaj ldquoData throughput ofCDMA-HDR a high efficiency-high data rate personal commu-nication wireless systemrdquo in Proceedings of the IEEE VehicularTechnology Conference (VTC rsquo00) 2000
[12] P Ameigeiras Packet scheduling and quality of service inHSDPA[PhD thesis] University of Aalborg Aalborg Denmark 2003
[13] K W Choi W S Jeon and D G Jeong ldquoResource allocationin OFDMA wireless communications systems supporting mul-timedia servicesrdquo IEEEACM Transactions on Networking vol17 no 3 pp 926ndash935 2009
[14] T Heikkinen andAHottinen ldquoDelay-differentiated schedulingin a fading channelrdquo IEEE Transactions on Wireless Communi-cations vol 7 no 3 pp 848ndash856 2008
[15] D P Bertsekas and R Gallager Data Networks Prentice-HallEnglewood Cliffs NJ USA 1992
[16] J Jang and K B Lee ldquoTransmit power adaptation for multiuserOFDM systemsrdquo IEEE Journal on Selected Areas in Communi-cations vol 21 no 2 pp 171ndash178 2003
[17] T Nguyen and Y Han ldquoA proportional fairness algorithmwith QoS provision in downlink OFDMA systemsrdquo IEEECommunications Letters vol 10 no 11 pp 760ndash762 2006
[18] S Ryu B Ryu H Seo M Shin and S Park ldquoWireless packetscheduling algorithm for OFDMA system based on time-utilityand channel staterdquo ETRI Journal vol 27 no 6 pp 777ndash7872005