Automatic Speech and Speaker Recognition.pdf

发布时间：2022-05-29 发布人：admin 分类：说明书资料大小：2.59M 资料格式：pdf 举报版权申诉

michaelshen7310-5071527-4744302543017757063.pdf-第1页.png

第1页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第2页.png

第2页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第3页.png

第3页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第4页.png

第4页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第5页.png

第5页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第6页.png

第6页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第7页.png

第7页 / 共257页

michaelshen7310-5071527-4744302543017757063.pdf-第8页.png

第8页 / 共257页

文本预览

Automatic Speech and Speaker Recognition Au tomatic Speech and Speaker R ecognition: Large Margin and Ker n el Methods Edited by Joseph Keshet and Samy Bengio © 2009 John Wiley & Sons, Ltd. ISBN: 978-0-470-69683-5

Automatic Speech and Speaker Recognition Large Margin and Kernel Methods Joseph Keshet IDIAP Research Institute, Martigny, Switzerland Samy Bengio Google Inc., Mountain View, CA, USA A John Wiley and Sons, Ltd, Publication

This edition ﬁrst published 2009 c 2009 John Wiley & Sons Ltd Registered ofﬁce John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom For details of our global editorial ofﬁces, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com. The right of the author to be identiﬁed as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Library of Congress Cataloging-in-Publication Data Automatic speech and speaker recognition : large margin and kernel methods / edited by Joseph Keshet, Samy Bengio. p. cm. Includes bibliographical references and index. ISBN 978-0-470-69683-5 (cloth) 1. Automatic speech recognition. I. Keshet, Joseph. II. Bengio, Samy. TK7895.S65A983 2009 006.4’54–dc22 A catalogue record for this book is available from the British Library. ISBN 9780470696835 (H/B) Set in 10/12pt Times by Sunrise Setting Ltd, Torquay, UK. Printed in Singapore by Markono Print Media Pte Ltd. 2008038551

Contents List of Contributors Preface I Foundations 1 Introduction Samy Bengio and Joseph Keshet 1.1 The Traditional Approach to Speech Processing . . . . . 1.2 Potential Problems of the Probabilistic Approach . . . . 1.3 Support Vector Machines for Binary Classiﬁcation . . . . . . . 1.4 Outline References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Theory and Practice of Support Vector Machines Optimization Shai Shalev-Shwartz and Nathan Srebo xi xv 1 3 3 5 7 8 9 11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction . . . . . . . . . 2.2.1 Binary Classiﬁcation and the Traditional SVM . . . . . 2.2.2 More General Loss Functions 2.2.3 Examples . . . . . . . . . . 2.2.4 Kernels . . . 2.2.5 . . . . . . . . . . . . . . Incorporating a Bias Term . . . . . . . . . . . . . . . . . . . . . . . . 2.1 2.2 SVM and L2-regularized Linear Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Optimization Accuracy From a Machine Learning Perspective . . . . 2.4 Stochastic Gradient Descent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 . . . . 2.4.2 Rate of Convergence and Stopping Criteria . . . . . . . . . . . . . . . . . . . 2.5.1 Duality . . . . . . . . . . . 2.6 Summary . . References . . . . . 2.5 Dual Decomposition Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sub-gradient Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 . . . . 12 . . . . 12 . . . . 13 . . . . 13 . . . . 14 . . . . 15 . . . . 16 . . . . 18 . . . . 20 . . . . 21 . . . . 22 . . . . 23 . . . . 25 . . . . 26

vi 3 From Binary Classiﬁcation to Categorial Prediction CONTENTS 27 . . . . . . . . . . . . . . . . . . . . . . . . . . Koby Crammer 3.1 Multi-category Problems . . . . . 3.2 Hypothesis Class 3.3 Loss Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Combinatorial Loss Functions . . . . . . . 3.4 Hinge Loss Functions . . . . . 3.5 A Generalized Perceptron Algorithm . . . . . . . . 3.6 A Generalized Passive–Aggressive Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 A Batch Formulation . 3.8 Concluding Remarks . . . . . 3.9 Appendix. Derivations of the Duals of the Passive–Aggressive Algorithm . . . . and the Batch Formulation . 3.9.1 Derivation of the Dual of the Passive–Aggressive Algorithm . . . . . . 3.9.2 Derivation of the Dual of the Batch Formulation . . . . . . . . . . . . 3.6.1 Dual Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . II Acoustic Modeling 4 A Large Margin Algorithm for Forced Alignment . . . 27 . . . 31 . . . 32 . . . 33 . . . 35 . . . 36 . . . 39 . . . 40 . . . 41 . . . 43 . . . 44 . . . 44 . . . 46 . . . 48 51 53 . . . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . Joseph Keshet, Shai Shalev-Shwartz, Yoram Singer and Dan Chazan . . . . 4.1 . . . . . . . . 4.2 Problem Setting . . . . . . . . 4.3 Cost and Risk . . . . . . 4.4 A Large Margin Approach for Forced Alignment 4.5 An Iterative Algorithm . . . . . . . 4.6 Efﬁcient Evaluation of the Alignment Function . . 4.7 Base Alignment Functions . . . . . . . . . . . . 4.8 Experimental Results . . . . . . . . . . . . 4.9 Discussion . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 . . . 54 . . . 55 . . . 56 . . . 57 . . . 62 . . . 64 . . . 66 . . . 67 . . . 67 5 A Kernel Wrapper for Phoneme Sequence Recognition 69 . . . . . . Introduction . . Joseph Keshet and Dan Chazan . . . . 5.1 . . . . . . . . . . . . 5.2 Problem Setting . . . . . . . . 5.3 Frame-based Phoneme Classiﬁer . . . . . . 5.4 Kernel-based Iterative Algorithm for Phoneme Recognition . . . . . . . 5.5 Nonlinear Feature Functions . . . . . . . . . . . . 5.5.1 Acoustic Modeling . . . . . 5.5.2 Duration Modeling . . . . . 5.5.3 Transition Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 . . . 70 . . . 71 . . . 71 . . . 75 . . . 75 . . . 77 . . . 78

CONTENTS 5.6 Preliminary Experimental Results . . . . 5.7 Discussion: Can we Hope for Better Results? . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . vii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 . . . . 79 . . . . 80 6 Augmented Statistical Models: Using Dynamic Kernels for Acoustic Models 83 . . . . . . . . . . Introduction . Mark J. F. Gales 6.1 . . . . 6.2 Temporal Correlation Modeling . 6.3 Dynamic Kernels . . . . . . 6.3.1 6.3.2 Generative Kernels . . . . . . . . 6.3.3 . . . . . . . . . . . . . . Static and Dynamic Kernels . . . . . . . . . . . . . . . 6.4.1 Generative Augmented Models . 6.4.2 Conditional Augmented Models . . . . . . . . . . . . . . . . . 6.5 Experimental Results . . . . . . 6.6 Conclusions . . . . . . . . Acknowledgements . . . . References . . . . . . . . 6.4 Augmented Statistical Models . . . . . . . . . . . . . . . . Simple Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 . . . . 84 . . . . 86 . . . . 87 . . . . 88 . . . . 90 . . . . 92 . . . . 92 . . . . 94 . . . . 95 . . . . 97 . . . . 97 . . . . 98 7 Large Margin Training of Continuous Density Hidden Markov Models 101 . . . . . . . . . . . . . . . . . . . . . . Fei Sha and Lawrence K. Saul 7.1 Introduction . 7.2 Background . . . . . . . . . . . . . . . . . . . 7.3 Large Margin Training . . . . . . . . . . 7.2.1 Maximum Likelihood Estimation . . . . 7.2.2 Conditional Maximum Likelihood . . . . 7.2.3 Minimum Classiﬁcation Error . . . . . . . . . . . . . . . . 7.3.1 Discriminant Function . . 7.3.2 Margin Constraints and Hamming Distances 7.3.3 Optimization . . . 7.3.4 Related Work . . . 7.4 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.4.1 Large Margin Training . . . . . . 7.4.2 Comparison with CML and MCE . . . . . . . . 7.4.3 Other Variants . . . . . . . . . . . . . . . . 7.5 Conclusion . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 . . . . 103 . . . . 103 . . . . 104 . . . . 104 . . . . 105 . . . . 105 . . . . 106 . . . . 106 . . . . 107 . . . . 107 . . . . 108 . . . . 109 . . . . 109 . . . . 112 . . . . 113 III Language Modeling 8 A Survey of Discriminative Language Modeling Approaches for Large 115 117 Vocabulary Continuous Speech Recognition Brian Roark 8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

viii . . . . . . . . . . . 8.3 Further Developments . 8.2 General Framework . . . . . . 8.2.1 Training Data and the GEN Function . . . . . . . 8.2.2 . . . . 8.2.3 . . . . . . . . . . . . . . . . . . . . . . . . Feature Mapping . . . . . . Parameter Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8.3.1 Novel Features . . . 8.3.2 Novel Objectives . . 8.3.3 Domain Adaptation . 8.4 Summary and Discussion . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 . . . 120 . . . 123 . . . 127 . . . 130 . . . 130 . . . 131 . . . 132 . . . 133 . . . 134 9 Large Margin Methods for Part-of-Speech Tagging . . . . . . . Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Sequence Boosting . . Yasemin Altun . . . . 9.1 9.2 Modeling Sequence Labeling . . . . Feature Representation . . . . . . . . . . 9.3.1 Objective Function . . . . . 9.3.2 Optimization Method . . . . . . . . . . . . . . . . 9.4 Hidden Markov Support Vector Machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.1 . . . . 9.2.2 Empirical Risk Minimization . . . . . . . . 9.2.3 Conditional Random Fields and Sequence Perceptron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.1 Data and Features for Part-of-Speech Tagging . . 9.5.2 Results of Sequence AdaBoost . . . . . . . . . . 9.5.3 Results of Hidden Markov Support Vector Machines . . . . . . . . . . 9.4.1 Objective Function . . . . . 9.4.2 Optimization Method . . . . . . . . 9.4.3 Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6 Discussion . . . . . . . References . . 9.5 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 . . . 139 . . . 140 . . . 141 . . . 142 . . . 143 . . . 144 . . . 145 . . . 145 . . . 149 . . . 149 . . . 151 . . . 151 . . . 153 . . . 153 . . . 154 . . . 155 . . . 156 . . . 156 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 A Proposal for a Kernel Based Algorithm for Large Vocabulary Continuous . . . . . . . . . . . . . . . Speech Recognition Joseph Keshet 10.1 Introduction . . . . . . 10.2 Segment Models and Hidden Markov Models . . . . . . . 10.3 Kernel Based Model . . . . . . . . . . . . . . . 10.4 Large Margin Training . . . . . . . . . . . 10.5 Implementation Details . . . . . . . 10.5.1 Iterative Algorithm . . . . . 10.5.2 Recognition Feature Functions . . . . . . . . . . . . . . . 10.5.3 The Decoder . 10.5.4 Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 . . . 161 . . . 163 . . . 164 . . . 166 . . . 166 . . . 167 . . . 169 . . . 169

分享到：

赞收藏

资料库

Automatic Speech and Speaker Recognition.pdf

相关推荐

开发技术

热门标签

最新资料