Requirements for the Transmission of Streaming Video in ...

1

Requirements for the Transmission of Streaming Video

in Mobile Wireless Networks

Vasos Vassiliou, Pavlos Antoniou, Iraklis Giannakou, and Andreas Pitsillides

Networks Research Lab (NetRL)

Computer Science DepartmentUniversity of Cyprus

2

Scope of the presentation

Present the characteristics of mobile wireless networks.

Relate these characteristics to the requirements for video transmission.

Establish guidelines for the transmission of video, based on the limits of mobile and wireless networks as well as on objective and subjective (perceptual) measurements.

3

Research Goals

Investigate the types of errors that can be observed, using objective video quality metrics such as PSNR by adopting the transmission of MPEG4-encoded video streams over wireless network environments.

Provide subjective video quality estimation based on the evaluation of decoded video streams by informed viewers.

Provide high quality transmission of video streaming based on objective QoS metrics measured in the network, subjective quality measures obtained by video viewers at end

hosts.

4

Characteristics of Mobile Wireless Networks Cellular Wireless Networks

2G – GSM: Very stable, widely accepted but does not support high data rate (only up to 9.6Kbps).

2.5G – GPRS: Designed to provide a high data rate packet-switched service in a GSM network (up to 384Kbps). Provides important benefits with respect to data and video.

3G – UMTS: Provides global mobility with wide range of services. Different QoS classes for 4 types traffic (up to 2Mbps).

5

Characteristics of Mobile Wireless Networks (ctd.) WLAN – IEEE 802.11 (WiFi - CSMA/CA)

a) Operates on the 2.4GHz band with maximum theoretical data rate of 54Mbps.

b) Based on common resource sharing MAC protocol which does not guarantee any QoS.

c) Data rate allocated to each user is inversely proportional to the number of users in the local network.

In ternet

PD AL aptop

W ork station

6

Characteristics of Mobile Wireless Networks (ctd.)

WiFi (WLAN) offers higher data rates (up to 54Mbps) within a single Access Point range. No handoff support.

UMTS (3G) and GPRS (2.5G) provide service continuity and seamless mobility due to handoff mechanism but they both operate in lower data rates (especially GPRS).

O ffic e o rR o o m

B u ild in g

In d o o rs O u td o o rs

S ta tio n a ry W a lk in g V e h ic le

G P R S (2 .5G )

Wir

ed T

erm

inal

U M TS (3G )

W L A N/ W iF i

Mbi

ts/s

ec

1 00 .0

10 .0

1 .0

0 .1

7

Video Quality Assessment Schemes

Objective QoS measures Classified in 3 categories based on the existence of the original stream:

The original video sequence is compared to the decoded video sequence (Full Reference) i.e. PSNR,

The original video sequence and the impaired one are fed to a computer algorithm that calculates the distortion between the two (Reduced Reference),

Observations are made only on decoded video and estimation of the quality is done using only that information (No Reference).

8

Video Quality Assessment Schemes (ctd.) Subjective QoS measures

Perceived measure of the quality of a video is done through the human "grading" of streams which helps collect and utilize the general user view (Mean Opinion Score - MOS).

Subjective measures are extracted from marks given by people, and can help to evaluate users' opinion.

9

Video Quality Assessment Schemes (ctd.) Objective QoS metric: PSNR (Full Reference Method) –

Measured in dB. Subjective QoS metric: SAMVIQ (Subjective

Assessment Method for Video Quality evaluation)a) Submitted as draft standard to ITU-R SG6 WP 6Q,

b) Uses a continuous scale graded from 0 to 100 annotated by 5 quality items linearly arranged

(Excellent, Good, Fair, Poor, Bad),

c) Based on random playout of the test files.

The individual viewer can start and stop the evaluation process as he wishes and is allowed to determine his own pace for performing grading, modifying grades, repeating playout when needed.

10

Evaluation Setup & Scenarios

We used NS2 simulation package and video streams of different content and characteristics so that the influence of errors in the wireless network is visible at the video stream reconstructed at the receiver.

Simulation test-bed

11

Evaluation Setup & Scenarios (ctd.)

Variable Network Test Parameters

Test Sequences

12

ResultsVariable Link Bandwidth and Propagation Delay

64Kbps encoding rate 256Kbps encoding rate

Higher encoding rate increases PSNR

13

Results (ctd.)Variable Video Encoding Rate and Propagation Delay

500Kbps link bandwidth 1Mbps link bandwidth

Stronger relationship between link BW and encoding rate than between the link BW and the propagation delay.

Propagation delay does not affect PSNR

14

Results (ctd.)Variable Link Bandwidth and Video Encoding Rate

10ms delay 400ms delay

PSNR LOW if Encoding rate > Link BW

15

Perceived QoS Evaluations Relationship between PSNR and MOS

PSNR does not have linear relationship with the MOS. PSNR is inappropriate for the evaluation of the actual

user perceived quality of service (poorly correlated to human vision).

PSNR vs MOS

0

5

10

15

20

25

30

Bad Poor Fair Good Excellent

MOS

PSNR

(dB)

16

Conclusions

Objective video quality - as this is reflected through PSNR values - depends on the percentage of lost frames.

End-to-end delay does not play an essential role in the objective video quality but it remains a critical factor for real-time services and may influence the subjective video quality.

Results show that the higher the video encoding bit rate the higher the QoS in terms of objective and subjective video quality measures.

Results show the inappropriateness of PSNR for the evaluation of the actual user-perceived QoS.

17

THE END

THANK YOU VERY MUCH