-
SPEECH PROCESSING FOR MAKHRAJ RECOGNITION
(DESIGN ADAPTIVE FILTER FOR NOISE REMOVAL)
SITI NURMAISARAH BT ABDUL AZIZ
This thesis is submitted as partial fulfillment of the
requirement
for the award of the
Bachelor of Electrical Engineering
(Electronics)
Faculty of Electrical & Electronics Engineering
Universiti Malaysia Pahang
NOVEMBER, 2010
-
ii
“I hereby acknowledge that the scope and quality of this thesis
is qualified for the award
of the Bachelor Degree of Electrical Engineering
(Electronics)”
Signature : ________________________________
Name : NURUL WAHIDAH BT ARSHAD
Date : 29 NOVEMBER 2010
-
iii
“All the trademark and copyrights use herein are property of
their respective owner.
References of information from other sources are quoted
accordingly; otherwise the
information presented in this report is solely work of the
author.”
Signature : ________________________________
Author : SITI NURMAISARAH BT ABDUL AZIZ
Date : 29 NOVEMBER 2010
-
v
ACKNOWLEDGEMENTS
First of all, I want to thanks to Allah for giving me this
opportunity, the strength
and the patience to complete my project successfully, after all
the challenges and
difficulties that I have face it.
Foremost, I would like to express my greatest gratitude to my
supervisor Madam
Nurul Wahidah Bt Arshad, who have guide and helped me a lot
throughout this final
year project. This appreciation is also dedicated to Mr. Mohd
Zamri Bin Ibrahim,
Madam Nurul Hazlina Bt Nordin, and Madam Rosyati Bt Hamid and
all the FKEE
staffs, those who are really generous and helpful.
I also would like to thanks to my parents, for supporting me
mentally and
physically not just during finishing this tasks but also during
my whole studies in order
to become a good Muslims.
Finally, I would like to take this opportunity to thank all my
friends and
colleagues who have given their support and help.
Hopefully, this final year project will not be the end of my
journey in seeking for
more knowledge to understand the meaning of life.
-
vi
ABSTRACT
Speech Processing for MAKHRAJ Recognition is a topic that very
useful in many
applications and environments in our daily day to improve
MAKHRAJ for Arabic
alphabets. In this project, it needs to design Adaptive Filter
for noise removal. There are
30 Arabic, أ until ي but for this project, only 7 Arabic will be
used as samples, أ until خ.
The speech processing will be used to obtain same waveform
output from two different
situations, road and cafeteria. Least Mean Square (LMS)
Algorithm based on Adaptive
Filter technique is used to remove noise. Filter Design Toolbox
provides many adaptive
filter design functions that use the LMS algorithms to search
for the optimal solution to
adaptive filter, including system identification and noise
cancellation. The filtered data
will be processed to match the standard pronunciations and it
will be integrated with
filter design process in MATLAB. As a result, the noise will be
removing and produce
same waveform signal.
-
vii
ABSTRAK
Pemprosesan Suara untuk Pengakuan Makhraj adalah satu topik yang
sangat
berguna dalam pelbagai aplikasi dan persekitaran dalam kehidupan
seharian kita untuk
meningkatkan Makhraj untuk huruf Arab. Dalam projek ini, ia
perlu untuk mereka
Penapis Adaptif untuk menyingkirkan bunyi bising. Ada 30 huruf
Arab, أ sampai ي tapi
untuk projek ini, hanya 7 huruf Arab akan digunakan sebagai
sampel, أ sampai خ.
Pemprosesan suara akan digunakan untuk mendapatkan keluaran
gelombang yang sama
dari dua situasi yang berbeza, jalan raya dan kafetaria. Least
Mean Square (LMS)
Algoritma berdasarkan teknik Penapis Adaptif digunakan untuk
menyingkirkan bunyi
bising. Filter Design Toolbox mempunyai banyak fungsi mereka
penapis adaptif yang
menggunakan algoritma LMS untuk mencari penyelesaian optimum
untuk menapis
adaptif, termasuk pengenalan sistem dan penyingkiran bunyi. Data
yang ditapis akan
diproses untuk menyesuaikan dengan sebutan sebenar dan akan
diintegrasikan dengan
proses penapis desain di MATLAB. Akibatnya, bunyi bising akan
disingkirkan dan
menghasilkan isyarat gelombang yang sama.
-
viii
TABLE OF CONTENT
CHAPTER TITLE PAGE
TITLE i
DECLARATION ii
DEDICATION iv
ACKNOWLEDGEMENTS v
ABSTRACT vi
ABSTRAK vii
TABLE OF CONTENTS viii
LIST OF TABLES xi
LIST OF FIGURES xii
LIST OF ABBREVIATION xiv
LIST OF APPENDICES xv
1 INTRODUCTION
1.1 Introduction 1
1.2 Objective 3
1.3 Scope of Project 3
1.4 Problem Statement 3
1.5 Thesis Outlines 4
2 LITERATURE REVIEW
2.1 Introduction 5
2.2 Speech Processing For MAKHRAJ Recognition 6
2.3 Adaptive Filter 7
-
ix
2.4 Least-Mean-Square (LMS) Based 10
2.4.1 Implementation of the LMS Algorithm 12
2.4.2 Convergence Properties 12
2.4.3 Wiener Filter Theory 14
2.5 Previous Research 15
3 METHODOLOGY
3.1 Introduction 19
3.2 Input Loading 20
3.3 Pre-Processing 21
3.4 Adaptive Filter 22
3.4.1 Create the Signals for Adaptation 23
3.4.2 Generate the Noise Signal 23
3.4.3 Corrupt the Desired Signal to Create a Noisy
Signal 24
3.4.4 Create a Reference Signal 24
3.5 Least-Mean-Square (LMS) Algorithm 25
3.5.1 System Identification Using Least Mean
Square (LMS) Algorithm 26
3.5.2 System Identification Using Least Mean
Square (LMS) Algorithm 27
3.5.3 Noise Cancellation using LMS Algorithm 28
4 RESULT AND DISCUSSION
4.1 Introduction 31
4.2 Input Loading 32
4.3 Adaptive Filter 33
4.4 Least Mean Square (LMS) Algorithms 36
-
x
5 CONCLUSION AND RECOMMENDATION
5.1 Conclusion 47
5.2 Recommendation 48
REFERENCES 49
APPENDICES
APPENDIX A 52
-
xi
LIST OF TABLES
TABLE NO. TITLE PAGE
2.1 LMS Algorithm Characteristics 13
3.1 Output Scaling Based On Typical Bit-
Widths for Native Formats 21
3.2 Output Scaling Based On Typical Bit-
Widths for Double Formats 21
3.3 Input Arguments for adaptfilt.nlms 27
3.4 Input Arguments for adaptfilt.ss 30
4.1 Table of Accuracy alif at Food Court 45
4.2 Table of Accuracy alif at Road 45
-
xii
LIST OF FIGURES
FIGURE NO. TITLE PAGE
2.1 Using an Filter to Remove
Noise from an Unknown System 8
2.2 Least-Mean-Square Implementation 10
2.3 Performance Surface Contours and
Weight Value Tracks for the LMS 13
2.4 The Wiener Filter Configuration 14
3.1 Flow Chart for Speech Recognition 20
3.2 Flow Chart for Adaptive Filter 22
3.3 Generate the Signals for Adaptation 23
3.4 Create a Noisy Signal 24
3.5 Create Reference Signal 24
3.6 Flow Chart for Least Mean Square
(LMS) algorithm 25
3.7 Syntax of adaptfilt.lms 26
3.8 Syntax of adaptfilt.nlms 27
3.9 Syntax of adaptfilt.ss 29
4.1 Waveform of Original Signal, y 32
4.2 Waveform of Desire Signal, signal 33
4.3 Waveform of Noise Signal v1 34
4.4 Waveform of Noisy Signal, A 35
4.5 Waveform of Reference Signal, v2 36
4.6 Waveform of System Identification by
adaptfilt.lms 37
-
xiii
4.7 Stem of System Identification by
adaptfilt.lms 38
4.8 Waveform of System Identification by
adaptfilt.nlms 39
4.9 Stem of System Identification by
adaptfilt.nlms 40
4.10 Noise Cancellation using LMS algorithms 41
4.11 Result of Filtering alphabet “alif”
at Food Road 42
4.12 Result of Filtering alphabet “alif”
at Road 44
4.13 Result of Filtering alphabet “alif”
-
xiv
LIST OF ABBREVIATIONS
LMS
FIR
MSE
SNR
NLMS
SSLMS
SDLMS
SELMS
RAM
Least Mean Square
Finite Impulse Response
Mean Square Error
Signal Noise Ratio
Normalized Least Mean Square
Sign-Sign Least Mean Square
Sign-Data Least Mean Square
Sign-Error Least Mean Square
Random Access Memory
-
xv
LIST OF APPENDICES
APPENDIX NO. TITLE PAGE
A Coding For Filtering Noise 54
-
CHAPTER 1
INTRODUCTION
1.1 INTRODUCTION
This project is about Speech Processing for MAKHRAJ Recognition
by using
Adaptive Filter. This filter is use to remove or filter the
noise and it is more efficient method.
The main purpose of this project is to remove the noise in
MAKHRAJ recording. It is because
the existing system cannot recognize the wanted alphabets
because of the noise. As an
example "ha", with the disturbance from the noise, the system
may recognize wrong alphabet
like "kho".
This project uses two inputs. The first input is the distorted
signal, the MAKHRAJ
recording without noise. The second input is the desired signal,
the unfiltered noise. The filter
works to eliminate the difference between the output signal and
the desired signal and outputs
the difference, which, in this case, is the clean MAKHRAJ
recording. When start the
simulation, we hear both noisy signal from environment and voice
from human. Over time,
the adaptive filter filters out the noise so we hear only the
voice from human.
For this project, the application that use is noise or
interference cancellation where the
filter adapts in real-time to remove noise by keeping the error
small. The term of filter is
often used to describe a device in the form of piece of physical
hardware or software that is
applied to a set of noisy data in order to extract information
about a prescribed quantity of
interest.
-
2
And the technique that applied in this project is
Least-Mean-Square (LMS) algorithm
to remove noise because it is easy and stable but the only
disadvantage is its weak
convergence. Besides that, it enjoys less computational
complexity because of the sign
present in the algorithm and good filtering capability because
of the normalized term. LMS
algorithm also represents the simplest and most easily applied
adaptive algorithms.
According to the MATLAB software, there is Adaptive Filter by
using Least Mean
Square (LMS) algorithms Toolbox that helps this project to train
the network.
-
3
1.2 OBJECTIVE
The objectives of this project are to:
i. Remove noise from unknown system.
ii. Design the system based on Least Mean Square (LMS) technique
on adaptive filter.
iii. Developed MAKHRAJ recognition software using Adaptive
Filter.
1.3 SCOPE OF PROJECT
There are three scopes of this project:
i. To remove noise of the speech recognition that able to
recognize in road environment
and cafeteria environment.
ii. To remove noise from human voice that produces filtered
speech MAKHRAJ
recognition by using Least Mean Square (LMS) algorithm.
iii. To develop software that can remove noise by using MATLAB
environment.
1.4 PROBLEM STATEMENT
In our daily life, speech recognition is very important in order
to improve the quality
of our speech but most of the people take it for granted
especially Muslim. They prefer
improve their English rather than MAKHRAJ.
For that reason, this project is proposed in order to create a
system that can be
improving their speech of MAKHRAJ. This system can easily
recognize the MAKHRAJ of
human voice in two different environments, cafeteria and
road.
-
4
1.5 THESIS OUTLINE
The Speech Processing for MAKHRAJ Recognition final thesis is a
combination of 5
chapters that contains and elaborates specific topics such as
Introduction, Literature Review,
Methodology, Result and Discussions and Conclusions and
Recommendation that applied in
this project.
Chapter 1 basically is an introduction of the project. In this
chapter, the main idea
about the background and objectives of the project will be
discussed. The basic concept of the
project will be focused in this chapter.
Chapter 2 is about literature review to review the critical
points of current knowledge
including substantive findings as well as theoretical and
methodological contributions to a
particular topic about this project.
Chapter 3 will be discussed more detail about the method that
used to achieve an
objective of this project. It wills shows and explain the flow
chart that been used to write the
coding, developing the process using the MATLAB.
Chapter 4 discusses all the results obtained and the limitation
of the project. All
discussions are concentrating on the result and performance of
the speech recognizer.
Chapter 5 will be explained about the problem and the
recommendation for this project.
-
CHAPTER 2
LITERATURE REVIEW
2.1 INTRODUCTION
MAKHRAJ is a set of range of organs in speech that will create
variety of letter
with its own character that is one of the vocalized forms of
human communication.
Each letter is created out of the phonetic combination of a
limited set of vowel and
consonant speech sound units that can be differentiate from
others.
MAKHRAJ recognition is important to help in practicing the
pronunciation the
letters correctly. So, in this chapter, the basic knowledge and
fundamental concept in
creating the MAKHRAJ recognition will be discussed. This MAKHRAJ
recognition
project is using Adaptive Filter as a main processer.
http://en.wikipedia.org/wiki/Phonetichttp://en.wikipedia.org/wiki/Vowelhttp://en.wikipedia.org/wiki/Consonant
-
6
2.2 SPEECH PROCESSING FOR MAKHRAJ RECOGNITION
Speech is the way of choice for humans to communicate. There are
no special
equipment required, no physical contact required, no visibility
required, and can
communicate while doing something else. Speech processing
includes speech coding,
speech synthesis, speech recognition, identity verification and
enhancement.
Speech coding is to transmit or store a speech waveform using a
few bits as
possible while retaining high quality because to save bandwidth
in telecoms
applications and to reduce memory storage requirements.
Speech synthesis is to convert a text string onto speech
waveform because for
technology to communicate when a display would be inconvenient
because too big,
eyes busy, via phone, in the dark and moving around [1].
Speech recognition is the process of converting spoken input to
text or
sometimes referred to as speech-to-text. There are a few of the
basic terms and concepts
that are fundamental to speech recognition:
i. Utterance - When the user says something
The speech recognition engine is "listening" for speech input.
When the engine
detects audio input (a lack of silence) the beginning of an
utterance is signaled.
Utterances are sent to the speech engine to be processed. If the
user doesn’t say
anything, the engine returns what is known as a silence timeout
that indicated there
was no speech detected within the expected timeframe. An
utterance can be a
phrase or a sentence.
ii. Pronunciations
One piece of information that the speech recognition engine uses
to process a word
is its pronunciation, which represents what the speech engine
thinks a word should
sound like. Words can have multiple pronunciations associated
with them. For
example, the word “pa” has at least two pronunciations in the
transliterating foreign
words in Arabic: “pa” in the Jawi script for “ف”and in Persian,
Urdu, and Kurdish
for “ب”.
iii. Grammars
Grammars define the domain, or context, within which the
recognition engine
works. The engine compares the current utterance against the
words and phrases in
http://en.wikipedia.org/wiki/Persian_languagehttp://en.wikipedia.org/wiki/Urdu_languagehttp://en.wikipedia.org/wiki/Kurdish_language
-
7
the active grammars. If the user says something that is not in
the grammar, the
speech engine will not be able to decipher it correctly.
iv. Accuracy
The performance of a speech recognition system is measurable and
perhaps the most
widely used measurement is accuracy. Arguably the most important
measurement of
accuracy is whether the desired end result occurred. Measurement
of recognition
accuracy is whether the engine recognized the utterance exactly
as spoken. This
measure of recognition accuracy is expressed as a percentage and
represents the number
of utterances recognized correctly out of the total number of
utterances spoken. It is a
useful measurement when validating grammar design. For example,
if the engine
returned “aaaliif” when the user said “alif” this would be
considered a recognition error.
Based on the accuracy measurement, there must analyze the
grammar to determine if
there is anything that can do to improve accuracy. For instance,
it might need to add
“aaliif” as a valid word to grammar [2].
2.3 ADAPTIVE FILTER
Adaptive filtering involves the changing of filter parameters
(coefficients) over time, to
adapt to changing signal characteristics. There are four
application of Adaptive Filter
[3]:
i. System Identification - Using adaptive filters to identify
the response of an
unknown system such as a communications channel or a telephone
line.
ii. Inverse System Identification - Using adaptive filters to
develop a filter that has a
response that is the inverse of an unknown system.
iii. Noise or Interference Cancellation - performing active
noise cancellation where the
filter adapts in real-time to remove noise by keeping the error
small.
iv. Prediction - describes using adaptive filters to predict a
signal's future values.
In noise cancellation, adaptive filters will remove noise from a
signal in real
time. Here, the desired signal, the one to clean up, combines
noise and desired
information. To remove the noise, feed a signal n'(k) to the
adaptive filter that represents
noise that is correlated to the noise to remove from the desired
signal.
http://www.mathworks.com/access/helpdesk/help/toolbox/filterdesign/ug/f1-5512.html#f1-5547http://www.mathworks.com/access/helpdesk/help/toolbox/filterdesign/ug/f1-5512.html#f1-5557http://www.mathworks.com/access/helpdesk/help/toolbox/filterdesign/ug/f1-5512.html#f1-5567http://www.mathworks.com/access/helpdesk/help/toolbox/filterdesign/ug/f1-5512.html#f1-5576
-
8
Figure 2.1: Using an Adaptive Filter to Remove Noise from an
Unknown
System.
So long as the input noise to the filter remains correlated to
the unwanted noise
accompanying the desired signal, the adaptive filter adjusts its
coefficients to reduce the
value of the difference between output signal, y(k) and desired
signal, d(k), removing
the noise and resulting in a clean signal in estimation error,
e(k). Notice that in this
application, the error signal actually converges to the input
data signal, rather than
converging to zero [4]. On this basis of this measure, the
adaptive filter will change its
coefficients in an attempt to reduce the error. The coefficient
update relation is a
function of the error signal squared and is given by
ℎ𝑛+1 𝑖 = ℎ𝑛 𝑖 +𝜇
2 −
𝛿
𝛿ℎ𝑛 𝑖 |𝑒| 2
The term inside the parentheses represents the gradient of the
squared-error with
respect to the ith
coefficient. The gradient is a vector pointing in the direction
of the
change in filter coefficients that will cause the greatest
increase in the error signal.
Because the goal is to minimize the error, however, Equation 1
updates the filter
coefficients in the direction opposite the gradient; that is why
the gradient term is
negated. The constant, μ is a step-size, which controls the
amount of gradient
information used to update each coefficient. After repeatedly
adjusting each coefficient
in the direction opposite to the gradient of the error, the
adaptive filter should converge;
-
9
that is, the difference between the unknown and adaptive systems
should get smaller
and smaller. To express the gradient decent coefficient update
equation in a more usable
manner, we can rewrite the derivative of the squared-error term
as
𝛿
𝛿ℎ 𝑖 𝑒 2 = 2
𝛿
𝛿ℎ 𝑖 𝑒 (2.1)
= 2𝛿
𝛿ℎ 𝑖 𝑑 − 𝑦 𝑒 (2.2)
= 2𝛿
𝛿ℎ 𝑖 𝑑 − h i x[n− i]
N−1
i=0 𝑒
𝛿
𝛿ℎ 𝑖 𝑒 2 = 2(−(𝑥[𝑛 − 𝑖]))𝑒 (2.3)
which in turn gives us the final LMS coefficient update,
ℎ𝑛+1 𝑖 = ℎ𝑛 𝑖 + 𝜇𝑒𝑥[𝑛 − 𝑖] (2.4)
The step-size, μ directly affects how quickly the adaptive
filter will converge
toward the unknown system. If μ is very small, then the
coefficients change only a
small amount at each update, and the filter converges slowly.
With a larger step-size,
more gradient information is included in each update, and the
filter converges more
quickly; however, when the step-size is too large, the
coefficients may change too
quickly and the filter will diverge. (It is possible in some
cases to determine
analytically the largest value of μ ensuring convergence.)
[5]
The objects use various algorithms to determine the weights for
the filter
coefficients of the adapting filter. While the algorithms differ
in their detail
implementations, the LMS and RLS share a common operational
approach to minimize
the error between the filter output and the desired signal
[6].
-
10
2.4 LEAST-MEAN-SQUARE (LMS) BASED
The least-mean-square (LMS) algorithm is a linear adaptive
filtering algorithm
that consists of two basic processes:
i. A filtering process, which involves computing the output of a
transversal filtering
produced by a set of tap inputs and generating an estimation
error by comparing
this output to a desired response.
ii. An adaptive filtering, which involves the automatic
adjustment of the tap weights
of the filter in accordance with the estimation error.
The combination of these two processes working together
constitutes a feedback
loop around the LMS algorithm. First, have a transversal filter
(adaptive filter) that
responsible for performing the filtering process. Second, have a
mechanism (unknown
system) for performing the adaptive control process on the tap
weights of the transversal
filter [7]. The filter calculates the filter weights, or
coefficients that produce the least
mean squares of the error between the output signal and the
desired signal (minimize
the error).
Figure 2.2 : Least-Mean-Square Implementation
SPEECH PROCESSING FOR MAKHRAJ RECOGNITION
(DESIGN ADAPTIVE FILTER FOR NOISE REMOVAL)
SITI NURMAISARAH BT ABDUL AZIZ
This thesis is submitted as partial fulfillment of the
requirement
for the award of the
Bachelor of Electrical Engineering
(Electronics)
Faculty of Electrical & Electronics Engineering
Universiti Malaysia Pahang
NOVEMBER, 2010
“I hereby acknowledge that the scope and quality of this thesis
is qualified for the award of the Bachelor Degree of Electrical
Engineering (Electronics)”
Signature: ________________________________
Name: NURUL WAHIDAH BT ARSHAD
Date: 29 NOVEMBER 2010
“All the trademark and copyrights use herein are property of
their respective owner. References of information from other
sources are quoted accordingly; otherwise the information presented
in this report is solely work of the author.”
Signature: ________________________________
Author: SITI NURMAISARAH BT ABDUL AZIZ
Date: 29 NOVEMBER 2010
Dedicated to my beloved parents,
ABDUL AZIZ BIN MOHD ZAIN & ROS FARIZAN BT MAT ZAIN,
Sibling,
ANGAH, UDA, ALANG, ACHIK & ADIK
Supervisor,
PN NURUL WAHIDAH BT ARSHAD
and all of you for giving a constant source of support and
encouragement
ACKNOWLEDGEMENTS
First of all, I want to thanks to Allah for giving me this
opportunity, the strength and the patience to complete my project
successfully, after all the challenges and difficulties that I have
face it.
Foremost, I would like to express my greatest gratitude to my
supervisor Madam Nurul Wahidah Bt Arshad, who have guide and helped
me a lot throughout this final year project. This appreciation is
also dedicated to Mr. Mohd Zamri Bin Ibrahim, Madam Nurul Hazlina
Bt Nordin, and Madam Rosyati Bt Hamid and all the FKEE staffs,
those who are really generous and helpful.
I also would like to thanks to my parents, for supporting me
mentally and physically not just during finishing this tasks but
also during my whole studies in order to become a good Muslims.
Finally, I would like to take this opportunity to thank all my
friends and colleagues who have given their support and help.
Hopefully, this final year project will not be the end of my
journey in seeking for more knowledge to understand the meaning of
life.
ABSTRACT
Speech Processing for MAKHRAJ Recognition is a topic that very
useful in many applications and environments in our daily day to
improve MAKHRAJ for Arabic alphabets. In this project, it needs to
design Adaptive Filter for noise removal. There are 30 Arabic, أ
until ي but for this project, only 7 Arabic will be used as
samples, أ until خ. The speech processing will be used to obtain
same waveform output from two different situations, road and
cafeteria. Least Mean Square (LMS) Algorithm based on Adaptive
Filter technique is used to remove noise. Filter Design Toolbox
provides many adaptive filter design functions that use the LMS
algorithms to search for the optimal solution to adaptive filter,
including system identification and noise cancellation. The
filtered data will be processed to match the standard
pronunciations and it will be integrated with filter design process
in MATLAB. As a result, the noise will be removing and produce same
waveform signal.
ABSTRAK
Pemprosesan Suara untuk Pengakuan Makhraj adalah satu topik yang
sangat berguna dalam pelbagai aplikasi dan persekitaran dalam
kehidupan seharian kita untuk meningkatkan Makhraj untuk huruf
Arab. Dalam projek ini, ia perlu untuk mereka Penapis Adaptif untuk
menyingkirkan bunyi bising. Ada 30 huruf Arab, أ sampai ي tapi
untuk projek ini, hanya 7 huruf Arab akan digunakan sebagai sampel,
أ sampai خ. Pemprosesan suara akan digunakan untuk mendapatkan
keluaran gelombang yang sama dari dua situasi yang berbeza, jalan
raya dan kafetaria. Least Mean Square (LMS) Algoritma berdasarkan
teknik Penapis Adaptif digunakan untuk menyingkirkan bunyi bising.
Filter Design Toolbox mempunyai banyak fungsi mereka penapis
adaptif yang menggunakan algoritma LMS untuk mencari penyelesaian
optimum untuk menapis adaptif, termasuk pengenalan sistem dan
penyingkiran bunyi. Data yang ditapis akan diproses untuk
menyesuaikan dengan sebutan sebenar dan akan diintegrasikan dengan
proses penapis desain di MATLAB. Akibatnya, bunyi bising akan
disingkirkan dan menghasilkan isyarat gelombang yang sama.
TABLE OF CONTENT
CHAPTER TITLE PAGE
TITLE i
DECLARATION iiDEDICATION iv
ACKNOWLEDGEMENTS v
ABSTRACT viABSTRAK vii
TABLE OF CONTENTS viii
LIST OF TABLES xi
LIST OF FIGURES xii
LIST OF ABBREVIATION xivLIST OF APPENDICES xv
1 INTRODUCTION
1.1 Introduction 1
1.2 Objective 3
1.3 Scope of Project 3
1.4 Problem Statement 3
1.5 Thesis Outlines 4
2 LITERATURE REVIEW
2.1 Introduction 5
2.2 Speech Processing For MAKHRAJ Recognition 6
2.3 Adaptive Filter 7
iii
2.4
2.5 Least-Mean-Square (LMS) Based 10
2.4.1Implementation of the LMS Algorithm 12
2.4.2Convergence Properties 12
2.4.3Wiener Filter Theory 14
2.6 Previous Research 15
3METHODOLOGY
3.1Introduction 19
3.2Input Loading 20
3.3Pre-Processing 21
3.4Adaptive Filter 22
3.4.1Create the Signals for Adaptation 23
3.4.2Generate the Noise Signal 23
3.4.3Corrupt the Desired Signal to Create a Noisy
Signal 24
3.4.4Create a Reference Signal 24
3.5Least-Mean-Square (LMS) Algorithm 25
3.5.1System Identification Using Least Mean
Square (LMS) Algorithm 26
3.5.2System Identification Using Least Mean
Square (LMS) Algorithm 27
3.5.3Noise Cancellation using LMS Algorithm 28
4RESULT AND DISCUSSION4.1Introduction 31
4.2Input Loading 32
4.3Adaptive Filter 33
4.4 Least Mean Square (LMS) Algorithms 36
5 CONCLUSION AND RECOMMENDATION
5.1 Conclusion 47
5.2 Recommendation 48
REFERENCES 49
APPENDICES
APPENDIX A 52
LIST OF TABLES
TABLE NO.TITLE PAGE
2.1LMS Algorithm Characteristics 13
3.1Output Scaling Based On Typical Bit-
Widths for Native Formats 21
3.2Output Scaling Based On Typical Bit-
Widths for Double Formats 21
3.3Input Arguments for adaptfilt.nlms 27
3.4Input Arguments for adaptfilt.ss 30
4.1Table of Accuracy alif at Food Court 45
4.2Table of Accuracy alif at Road 45
LIST OF FIGURES
FIGURE NO. TITLE PAGE
2.1 Using an Filter to Remove
Noise from an Unknown System 8
2.2Least-Mean-Square Implementation 10
2.3Performance Surface Contours and
Weight Value Tracks for the LMS 13
2.4The Wiener Filter Configuration 14
3.1Flow Chart for Speech Recognition 20
3.2Flow Chart for Adaptive Filter 22
3.3Generate the Signals for Adaptation 23
3.4 Create a Noisy Signal 24
3.5 Create Reference Signal 24
3.6 Flow Chart for Least Mean Square
(LMS) algorithm 25
3.7 Syntax of adaptfilt.lms 26
3.8 Syntax of adaptfilt.nlms 27
3.9 Syntax of adaptfilt.ss 29
4.1 Waveform of Original Signal, y 32
4.2Waveform of Desire Signal, signal 33
4.3 Waveform of Noise Signal v1 34
4.4 Waveform of Noisy Signal, A 35
4.5 Waveform of Reference Signal, v2 36
4.6 Waveform of System Identification by
adaptfilt.lms 37
4.7 Stem of System Identification by
adaptfilt.lms 38
4.8 Waveform of System Identification by
adaptfilt.nlms 39
4.9 Stem of System Identification by
adaptfilt.nlms 40
4.10 Noise Cancellation using LMS algorithms 41
4.11 Result of Filtering alphabet “alif”
at Food Road 42
4.12 Result of Filtering alphabet “alif”
at Road 44
4.13 Result of Filtering alphabet “alif”
LIST OF ABBREVIATIONS
LMS
FIR
MSE
SNR
NLMS
SSLMS
SDLMS
SELMS
RAM
Least Mean Square
Finite Impulse Response
Mean Square Error
Signal Noise Ratio
Normalized Least Mean Square
Sign-Sign Least Mean Square
Sign-Data Least Mean Square
Sign-Error Least Mean Square
Random Access Memory
LIST OF APPENDICES
APPENDIX NO.
TITLE
PAGE
A
Coding For Filtering Noise
54