Robust Automatic Speech Recognition
A Bridge to Practical Applications
- 1st Edition - October 12, 2015
- Imprint: Academic Press
- Authors: Jinyu Li, Li Deng, Reinhold Haeb-Umbach, Yifan Gong
- Language: English
- Hardback ISBN:9 7 8 - 0 - 1 2 - 8 0 2 3 9 8 - 3
- eBook ISBN:9 7 8 - 0 - 1 2 - 8 0 2 6 1 6 - 8
Robust Automatic Speech Recognition: A Bridge to Practical Applications establishes a solid foundation for automatic speech recognition that is robust against acoustic environme… Read more
Purchase options
Institutional subscription on ScienceDirect
Request a sales quoteRobust Automatic Speech Recognition: A Bridge to Practical Applications establishes a solid foundation for automatic speech recognition that is robust against acoustic environmental distortion. It provides a thorough overview of classical and modern noise-and reverberation robust techniques that have been developed over the past thirty years, with an emphasis on practical methods that have been proven to be successful and which are likely to be further developed for future applications.The strengths and weaknesses of robustness-enhancing speech recognition techniques are carefully analyzed. The book covers noise-robust techniques designed for acoustic models which are based on both Gaussian mixture models and deep neural networks. In addition, a guide to selecting the best methods for practical applications is provided.The reader will:
- Gain a unified, deep and systematic understanding of the state-of-the-art technologies for robust speech recognition
- Learn the links and relationship between alternative technologies for robust speech recognition
- Be able to use the technology analysis and categorization detailed in the book to guide future technology development
- Be able to develop new noise-robust methods in the current era of deep learning for acoustic modeling in speech recognition
- The first book that provides a comprehensive review on noise and reverberation robust speech recognition methods in the era of deep neural networks
- Connects robust speech recognition techniques to machine learning paradigms with rigorous mathematical treatment
- Provides elegant and structural ways to categorize and analyze noise-robust speech recognition techniques
- Written by leading researchers who have been actively working on the subject matter in both industrial and academic organizations for many years
Researchers and engineers in the area of speech processing, both in industry and academia; Undergraduate and graduate students in the area of signal and speech processing
- About the Authors
- List of Figures
- List of Tables
- Acronyms
- Notations
- Chapter 1: Introduction
- Abstract
- 1.1 Automatic Speech Recognition
- 1.2 Robustness to Noisy Environments
- 1.3 Existing Surveys in the Area
- 1.4 Book Structure Overview
- Chapter 2: Fundamentals of speech recognition
- Abstract
- 2.1 Introduction: Components of Speech Recognition
- 2.2 Gaussian Mixture Models
- 2.3 Hidden Markov Models and the Variants
- 2.4 Deep Learning and Deep Neural Networks
- 2.5 Summary
- Chapter 3: Background of robust speech recognition
- Abstract
- 3.1 Standard Evaluation Databases
- 3.2 Modeling Distortions of Speech in Acoustic Environments
- 3.3 Impact of Acoustic Distortion on Gaussian Modeling
- 3.4 Impact of Acoustic Distortion on DNN Modeling
- 3.5 A General Framework for Robust Speech Recognition
- 3.6 Categorizing Robust ASR Techniques: An Overview
- 3.7 Summary
- Chapter 4: Processing in the feature and model domains
- Abstract
- 4.1 Feature-Space Approaches
- 4.2 Model-Space Approaches
- 4.3 Summary
- Chapter 5: Compensation with prior knowledge
- Abstract
- 5.1 Learning from Stereo Data
- 5.2 Learning from Multi-Environment Data
- 5.3 Summary
- Chapter 6: Explicit distortion modeling
- Abstract
- 6.1 Parallel Model Combination
- 6.2 Vector Taylor Series
- 6.3 Sampling-Based Methods
- 6.4 Acoustic Factorization
- 6.5 Summary
- Chapter 7: Uncertainty processing
- Abstract
- 7.1 Model-Domain Uncertainty
- 7.2 Feature-Domain Uncertainty
- 7.3 Joint Uncertainty Decoding
- 7.4 Missing-Feature Approaches
- 7.5 Summary
- Chapter 8: Joint model training
- Abstract
- 8.1 Speaker Adaptive and Source Normalization Training
- 8.2 Model Space Noise Adaptive Training
- 8.3 Joint Training for DNN
- 8.4 Summary
- Chapter 9: Reverberant speech recognition
- Abstract
- 9.1 Introduction
- 9.2 Acoustic Impulse Response
- 9.3 A Model of Reverberated Speech in Different Domains
- 9.4 The Effect of Reverberation on ASR Performance
- 9.5 Linear Filtering Approaches
- 9.6 Magnitude or Power Spectrum Enhancement
- 9.7 Feature Domain Approaches
- 9.8 Acoustic Model Domain Approaches
- 9.9 The REVERB Challenge
- 9.10 To Probe Further
- 9.11 Summary
- Chapter 10: Multi-channel processing
- Abstract
- 10.1 Introduction
- 10.2 The Acoustic Beamforming Problem
- 10.3 Fundamentals of Data-Dependent Beamforming
- 10.4 Multi-Channel Speech Recognition
- 10.5 To Probe Further
- 10.6 Summary
- Chapter 11: Summary and future directions
- Abstract
- 11.1 Robust Methods in the Era of GMM
- 11.2 Robust Methods in the Era of DNN
- 11.3 Multi-Channel Input and Robustness to Reverberation
- 11.4 Epilogue
- Index
- Edition: 1
- Published: October 12, 2015
- Imprint: Academic Press
- No. of pages: 306
- Language: English
- Hardback ISBN: 9780128023983
- eBook ISBN: 9780128026168
JL
Jinyu Li
Jinyu Li received a Ph.D. degree from Georgia Institute of Technology, U.S. From 2000 to 2003, he was a Researcher at Intel China Research Center and a Research Manager at iFlytek, China. Currently, he is a Principal Applied Scientist at Microsoft, working as a technical lead to design and improve speech modeling algorithms and technologies that ensure industry state-of-the-art speech recognition accuracy for Microsoft products. His major research interests cover several topics in speech recognition and machine learning, including noise robustness, deep learning, discriminative training, and feature extraction. He has authored over 60 papers and awarded over 10 patents.
Affiliations and expertise
Principal Science Lead, MicrosoftLD
Li Deng
Li Deng received a Ph.D. degree from the University of Wisconsin-Madison, US. He was a professor (1989-1999) at the University of Waterloo, Canada. In 1999, he joined Microsoft Research, where he currently leads R&D of application-focused deep learning as Partner Research Manager of its Deep Learning Technology Center. He is also an Affiliate Professor at University of Washington. He is a Fellow of the Acoustical Society of America, Fellow of the IEEE, and Fellow of the International Speech Communication Association. He served as Editors-in-Chief for the IEEE Signal Processing Magazine and for the IEEE/ACM Transactions on Audio, Speech and Language Processing (2009-2014). His technical work has been focused on deep learning for speech, language, image, and multimodal processing, and for other areas of machine intelligence involving big data. He received numerous awards including the IEEE SPS Best Paper Awards, IEEE Outstanding Engineer Award, and APSIPA Industrial Distinguished Leader Award.
Affiliations and expertise
Research Manager/Principal Researcher, MicrosoftRH
Reinhold Haeb-Umbach
Reinhold Haeb-Umbach is a professor with the University of Paderborn, Germany. His main research interests are in the fields of statistical signal processing and pattern recognition, with applications to speech enhancement, acoustic beamforming and source separation, as well as automatic speech recognition. After having worked in industrial research laboratories for more than 10 years he joined academia as a full professor of Communications Engineering in 2001. He has published more than 150 papers in peer reviewed journals and conferences.
Affiliations and expertise
University of Paderborn, GermanyYG
Yifan Gong
Yifan Gong served the National Scientific Research Center (CNRS) and INRIA, France, as Research Engineer and then joined CNRS as Senior Research Scientist. He was a Visiting Research Fellow at the Communications Research Center of Canada. As Senior Member of Technical Staff, he worked for Texas Instruments at the Speech Technologies Lab, where he developed speech modeling technologies robust against noisy environments, designed systems, algorithms, and software for speech and speaker recognition, and delivered memory- and CPU-efficient recognizers for mobile devices. Yifan joined Microsoft in 2004, and is currently a Principal Science Manager in the areas of speech modeling, computing infrastructure, and speech model development for speech products. His research interests include automatic speech recognition/interpretation, signal processing, algorithm development, and engineering process/infrastructure and management. He has authored over 130 publications and awarded over 30 patents. Specific contributions include stochastic trajectory modeling, source normalization HMM training, joint compensation of additive and convolutional noises, and variable parameter HMM. In these areas, he gave tutorials and presentations in international conferences. He has been serving as member of technical committee and session chair for many international conferences, and with IEEE Signal Processing Spoken Language Technical Committees from 1998 to 2002 and since 2013.
Affiliations and expertise
Principal Science Manager, MicrosoftRead Robust Automatic Speech Recognition on ScienceDirect