• Sonuç bulunamadı

CONTENTS ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i ÖZET

N/A
N/A
Protected

Academic year: 2021

Share "CONTENTS ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i ÖZET"

Copied!
3
0
0

Yükleniyor.... (view fulltext now)

Tam metin

(1)

CONTENTS

ABSTRACT . . . i

ÖZET . . . iii

ACKNOWLEDGEMENTS . . . v

CONTENTS . . . vi

LIST OF TABELES . . . . . . ix

LIST OF FIGURES . . . x

1. INTRODUCTION . . . 1

2. SPEECH RECOGNITION . . . 3

2.1. Overview . . . 3

2.2. Applications of speech recognition system . . . 3

2.3. Types of speech . . . 6

2.4. Speech signal and its basic properties . . . 6

2.5. Speech production . . . 7

2.6. Review of speech recognition . . . .. . . 8

3. STRUCTURE OF SPEECH RECOGNITION SYSTEM . . . 12

3.1. Overview . . . 12

3.2. The basic structure . . . 12

3.2.1. Speech . . . 12

3.2.2. Signal preprocessing . . . 13

3.2.3. Features extraction . . . 14

3.2.4. Speech classification . . . 15

3.2.5. Output . . . 16

3.3. Features extraction techniques . . . 16

vi

(2)

3.3.1. Linear Predictive Coding (LPC) . . . 16

3.3.1.1. LPC estimation . . . 17

3.3.1.2. LPC analysis . . . 18

3.3.1.3. LPC synthesis . . . 18

3.3.1.4. LPC applications . . . 19

3.3.2. Mel Frequency Cepstral Coefficients (MFCC) . . . 20

3.3.3. Spectrogram . . . . . . 22

3.3.3.1. Wideband spectrogram . . . 23

3.3.3.2. Narrowband spectrogram . . . 23

3.3.3.3. How to get spectrogram of a speech signal . . . 24

4. NEURAL NETWORKS FOR SPEECH CLASSIFICATION . . . 26

4.1. Overview . . . 26

4.2. Fundamentals of neural networks . . . 26

4.3. Models of neuron . . . 30

4.4. Feed-Forward neural networks . . . . . . 32

4.5. Backpropagation algorithm . . . 34

5. DESIGN OF SPEECH RECOGNITION SYSTEM . . . 36

5.1. Overview . . . 36

5.2. General structure of the system . . . 36

5.3. Flowcharts of features extraction methods . . . 37

5.4. Implementation of features extraction methods . . . 40

5.4.1. Linear Predictive Coding (LPC) . . . 43

5.4.2. Mel Frequency Cepstral Coefficients (MFCC) . . . 48

5.4.3. Spectrogram . . . 50

vii

(3)

5.5. The design of speech recognition program . . . 52

5.5.1. The selection of the method for feature extraction . . . 55

5.5.2. Training the system . . . 55

5.5.3. Speech recognition using LPC . . . 58

5.5.4. Speech recognition using MFCC . . . 60

5.5.5. Speech recognition using Spectrogram . . . 61

5.5.6. Neural network training . . . 61

5.5.7. Testing the system . . . 64

5.5.8. Quit button . . . . . . 71

6. RESULTS AND CONCLUSIONS . . . 72

6.1. Results . . . . . . 72

6.2. Conclusions . . . 74

7. REFERENCES . . . 75

Appendix A: Main GUI functions for Speech Recognition system . . . 78

Appendix B: Functions that used in Speech Recognition system . . . 80

viii

Referanslar

Benzer Belgeler

Ex- perimental studies performed over the Berlin Emotional Speech Database and the FAU Aibo Emotion Corpus demonstrate that decision fusion configurations with LSF features bring

In this study, we propose a semi-supervised speech-music sep- aration method which uses the speech, music and speech-music segments in a given segmented audio signal to separate

Aşağıda daha ayrıntılı bir şekjlde yeniden gözden geçirilecek olan ve farklı düşünce okulları tarafından getirilen bütün bu eleştirilere rağmen, YBK' nın

bağımsız değişkenler ise, kişi başına gelir, işsizlik, okullaşma oranı, şehirleşme oranı, güvenlik görevlisi sayısı, göç ve yoksulluk gibi değişkenler yer

Esas ürüne ait olmak üzere ürün kalitesi bir diğer ifadeyle performansı endüstriyel ürünler için vazgeçilmez ve en önemli ürün bileşenlerinden birisi olarak kabul

Chapter 4 describes the speaker identification systems and the methods used to identify speech signals.. Both text dependent and text independent speech identification methods

Two of the challenging issues in recognition of Turkish text are determined as delayed strokes which introduce an extra source of variation in the sequence order of the

Deney sonuçlarına bakıldığında, beklenildiği gibi sınırlı dağarcıklı deneydeki hata yüzdesi, geniş dağarcıktakilere göre çok daha düşüktür. Çünkü sınırlı