Audio Processing and Speech Recognition (eBook)

Soumya Sen is an Assistant Professor at A. K. Choudhury School of Information Technology, University of Calcutta. He received his Ph.D. (Tech) degree from the Department of Computer Science and Engineering, at the same university, in 2016. Before joining A. K. Choudhury School of Information Technology, he worked at IBM India Pvt. Ltd and RS Software. His industrial expertise includes ERP and data warehousing. Currently his research interests are data warehousing and OLAP tools, data mining, big data, service engineering, distributed databases, and machine learning. He has published 1 book, 70 research papers in peer-reviewed journals and international conferences and registered 3 patents in USA, Japan and South Korea. Dr. Sen is a PC member and reviewer for numerous International conferences. Anjan Dutta was born in Kolkata, India, in 1986. He received his B.Tech degree in Information Technology from West Bengal University of Technology in 2008 and M.Tech in Information Technology in 2011 from Calcutta University.He served in IXIA Technologies LTD and TATA Consultancy Services Ltd. (TCSL) over 6 years of period. Initially he worked as a protocol developer in IXIA Technologies LTD and worked on 3gpp wireless protocols. Thereafter he worked as an IT Analyst in TATA Consultancy Services Ltd.(TCSL) Form July, 2011 to July, 2017. He is now employed as an Assistant Professor in Department of Information Technology, Techno India College of Technology, India. He is an active researcher in the field of Big Data, Data Mining, Audio processing and Audio classification etc. Nilanjan Dey was born in Kolkata, India, in 1984. He received his B.Tech. degree inInformation Technology from West Bengal University of Technology in 2005, M.Tech.in Information Technology in 2011 from the same University and Ph.D. in digital image processing in 2015 from Jadavpur University, India.In 2011, he was appointed as an Assistant Professor in the Department of Information Technology at JIS College of Engineering, Kalyani, India followed by Bengal College of Engineering College, Durgapur, India in 2014. He is now employed as an Assistant Professor in Department of Information Technology, Techno India College of Technology, India. He is a visiting fellow of the University of Reading, UK. His research topic is signal processing, machine learning and information security.Dr. Dey is an Associate Editor of IEEE ACCESS and is currently the Editor in-Chief of the International Journal of Ambient Computing and Intelligence. Series Co-editor of Advances in Ubiquitous Sensing Applications for Healthcare (AUSAH), Elsevier and Springer Tracts in Nature-Inspired Computing (STNIC).

Preface 7
Objective of the Book 7
Organization of the Book 7
Chapter 1: Audio Indexing 8
Chapter 2: Speech Processing and Recognition System 8
Chapter 3: Feature Extraction 9
Chapter 4: Audio Classification 10
Contents 11
About the Authors 13
1 Audio Indexing 15
1.1 Introduction 15
1.2 Audio Indexing and Classic Information Retrieval Problem 16
1.3 Large Vocabulary Continuous Speech Recognition (LVCSR) 16
1.3.1 Recognition Errors and Vocabulary Limitations 17
1.3.2 The Out-of-Vocabulary Problem 19
1.3.3 Pros and Cons of LVCSR Speech Analytics 19
1.4 Phonetic Search 20
1.4.1 Phases of Phonetic Search 21
1.4.2 Pros and Cons of Phonetic Search 22
1.5 Comparison Between LVCSR and Phonetic Search 22
1.6 Summary 23
References 24
2 Speech Processing and Recognition System 26
2.1 Introduction 26
2.2 Human Speech Production System 27
2.2.1 Speech Generation 27
2.2.2 Speech Perception 28
2.2.3 Voiced and Unvoiced Speech 28
2.2.4 Model of Human Speech 29
2.3 Automatic Speech Recognition System 30
2.3.1 History of ASR 31
2.3.2 Structure of an ASR System 32
2.3.3 Neural Network and Speech Recognition System 41
2.3.4 Pronunciation Model 46
2.3.5 Language Model 48
2.3.6 Central Decoder 49
2.4 Summary 52
References 54
3 Feature Extraction 57
3.1 Introduction 57
3.2 Basic Audio Features 58
3.2.1 Pitch 58
3.2.2 Timbral Features 59
3.2.3 Rhythmic Features 61
3.2.4 Inharmonicity 61
3.2.5 Autocorrelation 61
3.2.6 Other Features 62
3.2.7 MPEG-7 Features 63
3.3 Feature Extraction Techniques 64
3.3.1 Linear Prediction Coding (LPC) 64
3.3.2 Mel-Frequency Cepstral Coefficient (MFCC) 66
3.3.3 Perceptual Linear Prediction (PLP) 68
3.3.4 Discrete Wavelet Transform (DWT) 70
3.4 Summary 75
References 76
4 Audio Classification 79
4.1 Introduction 79
4.2 Classification Strategies 80
4.2.1 k-Nearest Neighbors (k-NN) 80
4.2.2 Naïve Bayes (NB) Classifier 83
4.2.3 Decision Tree and Speech Classification 86
4.2.4 Support Vector Machine (SVM) and Speech Classification 97
4.3 Neural Network in Speech Classification 99
4.4 Deep Neural Network in Speech Recognition and Classification 101
4.5 Summary 101
References 102
5 Conclusion 106