Foreword

vii

Keith M. Ponting

Insight vs. performance

viii

Dangers

Hot topics

Towards the future

Integrating knowledge sources

A unified theory?

References

xii

Speech Pattern Processing

(9)

Roger K. Moore

The State-of-the-Art in Speech

(1)

Speech Patterning

(1)

Speech Pattern Processing

(2)

Whither a Unified Theory?

(1)

Towards a Theory

(1)

Practical Issues

(1)

What We Know

(1)

Some Things We Don't Know

(1)

The Way Forward

(3)

References

(2)

Psycho-acoustics and Speech Perception

(8)

Louis C. W. Pols

Introduction

(1)

Psycho-acoustics

(2)

Speech Perception

(2)

Vowel Reduction and Schwa

(1)

Spectro-temporal Dynamics of Formant Transitions

(1)

Consonant Reduction

(1)

Discussion

(3)

References

(2)

Acoustic Modelling for Large Vocabulary Continuous Speech Recognition

(22)

Steve Young

Introduction

(1)

Overview of LVCSR Architecture

(3)

Front End Processing

(1)

Basic Phone Modelling

(8)

HMM Phone Models

(2)

HMM Parameter Estimation

(2)

Context-Dependent Phone Models

(4)

Adaptation for LVCSR

(3)

Maximum Likelihood Linear Regression

(1)

Estimating the MLLR Transforms

(2)

Progress in LVCSR

(1)

Discriminative Training for LVCSR

(3)

Conclusions

(3)

References

(2)

Tree-based Dependence Models for Speech Recognition

(14)

Mari Ostendorf

Ashvin Kannan

Orith Ronen

Introduction

(1)

Hidden Tree Framework

(2)

Hidden Dependence Trees

(4)

The Mathematical Framework

(1)

Application to Speech

(1)

Topology Design and Parameter Estimation

(2)

Experiments

(1)

Multiscale Tree Processes

(4)

The Mathematical Framework

(1)

Application to Speech

(1)

Topology Design and Parameter Estimation

(1)

Experiments

(1)

Discussion

(3)

References

(2)

Connectionist and Hybrid Models for Automatic Speech Recognition

(13)

Jean-Paul Haton

Introduction

(1)

A Brief Overview of Neural Networks

(2)

Basic Principles

(1)

Main Models for ASR

(1)

Signal Processing and Feature Extraction using ANNs

(1)

Neural Networks as Static Pattern Classifiers

(1)

Speech Pattern Classification with Perceptrons

(1)

Feature Maps

(1)

Dynamic Aspects

(2)

Position of the Problem

(1)

Time Delays

(1)

Dynamic Classifiers

(1)

Recurrent NNs

(1)

Hybrid Models

(2)

Position of the Problem

(1)

Proposed Solutions

(2)

Conclusion

(4)

References

(4)

Computational Models for Auditory Speech Processing

(11)

Li Deng

Introduction

(1)

A nonlinear computational model for basilar membrane wave motions

(1)

Frequency-domain and time-domain computational solutions to the BM model

(2)

Interval analysis of auditory model's outputs for temporal information extraction

(1)

IPIH representation of clean and noisy speech sounds

(2)

Speech recognition experiments

(2)

Summary and discussions

(3)

References

(2)

Speaker Adaptation of CDHMMs Using Bayesian Learning

(6)

Claudio Vair

Luciano Fissore

Introduction

(1)

Bayesian Estimation of CDHMMs

(3)

Prior Density Definition

(1)

Forgetting Mechanism

(1)

Prior Parameter Estimation and MAP Solution

(1)

Acoustic Normalization

(1)

Tasks, Corpus and System

(1)

Speaker Adaptation Experiments

(1)

Conclusions

(1)

References

(1)

Discriminative Improvement of the Representation Space for Continuous Speech Recognition

(6)

Angel de la Torre

Antonio M. Peinado

Antonio J. Rubio

Jose C. Segura

Introduction

(1)

Discriminative Feature Extraction

(1)

SGDFE Algorithm for CSR

(1)

Experimental Results

(3)

Conclusions

(1)

References

(1)

Dealing with Loss of Synchronism in Multi-Band Continuous Speech Recognition Systems

(6)

Christophe Cerisara

Introduction

(1)

Forcing Synchronism Between the Bands

(1)

First Approach

(1)

Experiments

(1)

Modeling Loss of Synchronism

(2)

Theoretical Approach

(1)

Experimental Approach

(1)

Conclusion

(2)

References

(1)

K-Nearest Neighbours Estimator in a HMM-Based Recognition System

(6)

Fabrice Lefevre

Claude Montacie

Marie-Jose Caraty

Introduction

(1)

K-NN Assessment

(1)

K-NN estimator in HMM

(2)

Adaptation Principle

(1)

HMM Estimation Improvement

(1)

Evaluations

(2)

Recognition rates

(1)

SNALC Evaluation

100

(1)

Perspectives

101

(1)

References

101

(1)

Robust Speech Recognition

102

(10)

Sadaoki Furui

Mismatches between Training and Testing

102

(2)

Speech Variation

102

(2)

Inter-Speaker Variation

104

(1)

Reducing Mismatches to Improve Speech Recognition

104

(5)

Principles of Adaptive Speech Recognition

104

(2)

Three Principal Adaptation Methods for Reducing Mismatches

106

(1)

Important Practical Issues

107

(1)

N-Best-Based Unsupervised Adaptation

108

(1)

Conclusion

109

(3)

References

109

(3)

Channel Adaptation

112

(10)

Keith M. Ponting

Introduction

112

(1)

Matched condition training

112

(1)

Robust features

112

(1)

Model adaptation

113

(1)

Channel adaptation

113

(1)

Speech enhancement

113

(1)

Models of distortion

113

(2)

Minimum mean square error

114

(1)

Additive noise estimation

114

(1)

Methods for channel adaptation

115

(4)

Global transformations

115

(1)

Class-specific corrections

116

(1)

Empirical methods based on stereo data

117

(1)

Model-based compensation

118

(1)

Conclusion

119

(3)

References

120

(2)

Speaker Characterization, Speaker Adaptation and Voice Conversion

122

(10)

Sadaoki Furui

Introduction

122

(1)

Speaker-Characterization

122

(1)

Speaker Recognition

123

(1)

Speaker-Adaptation Techniques for Speech Recognition

124

(4)

Classification of Speaker-Adaptation/Normalization Methods

124

(1)

Speaker Cluster Selection Methods

124

(1)

Interpolated Re-Estimation Algorithm

125

(1)

Spectral Mapping Algorithm

125

(3)

Individuality Problems in Speech Synthesis and Coding

128

(1)

Conclusion

129

(3)

References

130

(2)

Speaker Recognition

132

(11)

Sadaoki Furui

Principles of Speaker Recognition

132

(1)

Text-Independent Speaker Recognition Methods

133

(4)

Long-Term-Statistics-Based Methods

133

(2)

VQ-Based Methods

135

(1)

Ergodic-HMM-Based Methods

135

(1)

Speech-Recognition-Based Methods

136

(1)

Text-prompted Speaker Recognition

137

(1)

Normalization and Adaptation Techniques

137

(3)

Parameter-Domain Normalization

138

(1)

Likelihood Normalization

138

(1)

HMM Adaptation for Noisy Conditions

139

(1)

Updating Models and A Priori Threshold for Speaker Verification

139

(1)

Open Questions and Concluding Remarks

140

(3)

References

140

(3)

Application of Acoustic Discriminative Training in an Ergodic HMM for Speaker Identification

143

(6)

Leandro Rodriguez Linares

Carmen Garcia Mateo

Introduction

143

(1)

Experimental Conditions

144

(1)

System Architecture

145

(1)

Acoustic Segmentation

145

(1)

The PTE-HMM Model

145

(1)

Experimental Results

145

(2)

Conclusions

147

(2)

References

148

(1)

Comparison of Several Compensation Techniques for Robust Speaker Verification

149

(8)

Laura Docio-Fernandez

Carmen Garcia-Mateo

Introduction

149

(2)

The HMM recognition system

151

(1)

Mismatch Compensation Techniques

151

(1)

CMS

151

(1)

SM1

152

(1)

SM2

152

(1)

Experiments and Results

152

(4)

Discussion and Conclusion

156

(1)

References

156

(1)

Segmental Acoustic Modeling for Speech Recognition

157

(16)

Mari Ostendorf

Introduction

157

(1)

Segmental and Hidden Markov Models

158

(7)

General Modeling Framework

159

(2)

Models of Feature Dynamics

161

(4)

Recognition and Training

165

(3)

Recognition Algorithms

165

(1)

Parameter Estimation Algorithms

166

(2)

Segmental Features

168

(1)

Summary

169

(4)

References

170

(3)

Trajectory Representations and Acoustic Descriptions for a Segment-Modelling Approach to Automatic Speech Recognition

173

(8)

Wendy J. Holmes

Introduction

173

(1)

Modelling Trajectories in Speech

174

(1)

Representing an Unobserved Trajectory with Segmental HMMs

175

(2)

Calculating segment probabilities

175

(1)

Recognition experiment

176

(1)

HMM Recognition with Formant Features

177

(1)

Modelling trajectories of cepstrum and formant features

178

(1)

Conclusions

178

(3)

References

179

(2)

Suprasegmental Modelling

181

(18)

E. Noth

A. Batliner

A. Kiebling

R. Kompe

H. Niemann

Introduction

181

(2)

The Verbmobil System

183

(1)

Computation of Prosodic Information

183

(7)

Extraction of Prosodic Features

185

(1)

Prosodic Classes

185

(1)

New Boundary Labels: The Syntactic-prosodic M-labels

186

(1)

Classification of Prosodic Events

187

(1)

Improving the Classification Results with Stochastic Language Models

187

(2)

Prosodic scoring of WHGs

189

(1)

The Use of Prosodic Information

190

(6)

Prosody and Syntax --- Interaction with the TUG-Grammar

190

(4)

Prosody and the Other Linguistic Modules

194

(2)

Concluding Remarks

196

(3)

References

196

(3)

Computational Models for Speech Production

199

(15)

Li Deng

Introduction

199

(1)

Speech production models in science/technology literatures

200

(2)

Derivation of discrete-time version of statistical task-dynamic model

202

(2)

Algorithms for learning task-dynamic model parameters and for likelihood computation

204

(6)

Model with deterministic, time-invariant parameters

205

(2)

Model with random, time-invariant parameters

207

(1)

Model with random, smoothly time-varying parameters

208

(2)

Discriminative learning of production models parameters

210

(1)

Other types of computational models of speech production

210

(2)

Summary and discussions

212

(2)

References

212

(2)

Articulatory Features and Associated Production Models in Statistical Speech Recognition

214

(11)

Li Deng

Introduction

214

(1)

Functional description of human speech communication as an encoding-decoding process

214

(1)

Overview of theories of speech perception

215

(1)

A general framework of statistical speech recognition

216

(1)

Brief analysis of weaknesses of current speech recognition technology

217

(1)

Phonological model: Overlapping articulatory features and related HMMs

218

(1)

Task-dynamic model of speech production

219

(1)

Interfacing overlapping features to task-dynamic model and a general architecture for speech recognition

220

(1)

Discussions: Machine speech recognition

220

(5)

References

223

(2)

Talker Normalization with Articulatory Analysis-by-Synthesis

225

(8)

Richard S. McGowan

Introduction

225

(1)

Normalization Procedure

226

(3)

Experiments

229

(2)

Conclusion

231

(2)

References

231

(2)

The Psycholinguistics of Spoken Word Recognition

233

(19)

Cynthia M. Connine

Thomas Deelman

Introduction

233

(1)

Overview: Models of spoken word recognition

233

(2)

Currency of mapping: units and the nature of lexical representations

235

(2)

Temporal nature of speech: early vs delayed commitment

237

(2)

Delayed commitment

238

(1)

Multiple lexical hypotheses, lexical competition and graded activation

239

(3)

Language architecture: Lexical and segmental levels

242

(2)

Language architecture: Lexical and sentential

244

(1)

Contribution of attention

245

(7)

References

247

(5)

Issues in Using Models for Self Evaluation and Correction of Speech

252

(7)

Marie-Christine Haton

Introduction

252

(1)

Using models

253

(1)

Norm building

254

(1)

Matching between the subject's world and the technical world

255

(1)

Settlement of the speech education program

256

(1)

Management of the education program

257

(1)

Conclusion

257

(2)

References

257

(2)

The Use of the Maximum Likelihood Criterion in Language Modelling

259

(21)

Hermann Ney

Introduction

259

(1)

Perplexity and Maximum Likelihood

260

(3)

Smoothing and Discounting for Sparse Data

263

(4)

Modelfree Discounting and Turing-Good Estimates

263

(3)

Absolute Discounting

266

(1)

Partitioning-Based Models

267

(5)

Equivalence Classes of Histories and Decision Trees

267

(3)

Two-Sided Partitionings and Word Classes

270

(2)

Word Trigger Pairs

272

(3)

Maximum Entropy Approach

275

(2)

Conclusions

277

(3)

References

277

(3)

Language Model Adaptation

280

(24)

Renato DeMori

Marcello Federico

Introduction

280

(1)

Background on Language Models

281

(2)

Adaptation paradigms

283

(2)

LM adaptation in dialogue systems

284

(1)

Basis statistical methods

285

(10)

Maximum a-posteriori estimation

285

(1)

Linear interpolation

286

(2)

Sublanguages mixture adaptation

288

(1)

Backing-off

288

(2)

Maximum Entropy

290

(1)

Minimum Discrimination Information

291

(1)

Generalized iterative scaling

292

(1)

Cache model and word triggers

293

(2)

Practical applications of adaptation paradigms

295

(6)

The 1993 ARPA evaluation method

295

(1)

Mixture based adaptation

296

(2)

Adaptation with a cache model

298

(1)

ME and MDI adaptation

299

(1)

LM adaptation in interactive systems

299

(2)

Conclusion

301

(3)

References

301

(3)

Using Natural-Language Knowledge Sources in Speech Recognition

304

(24)

Robert C. Moore

Introduction

304

(1)

Issues in Language Modeling for Speech Recognition

305

(2)

Formal Models for Natural Language

307

(5)

Finite-State Grammars

307

(1)

Context-Free Grammars

308

(1)

Augmented Context-Free Grammars

309

(1)

Expressive Power of Grammar Formalisms and the Requirements of Natural Language

310

(2)

Search Architectures for Natural-Language-Based Language Models

312

(2)

Word Lattice Parsing

312

(1)

N-best Filtering or Rescoring

312

(1)

Dynamic Generation of Partial Grammar Networks

313

(1)

Compiling Unification Grammars into Context-Free Grammars

314

(4)

Instantiating Unification Grammars

314

(2)

Removing Left Recursion from Context-Free Grammars

316

(2)

Robust Natural-Language-Based Language Models

318

(7)

Combining Linguistics and Statistics in a Language Model

318

(2)

Fully Statistical Natural-Language Grammars

320

(5)

Summary

325

(3)

References

326

(2)

How May I Help You?

328

(22)

A.L. Gorin

G. Riccardi

J.H. Wright

Introduction

328

(1)

A Spoken Dialog System

329

(2)

Database

331

(2)

Algorithms

333

(10)

Salient Fragment Acquisition

335

(4)

Recognizing Fragments in Speech

339

(2)

Call Classification

341

(2)

Experiment Results

343

(4)

Conclusions

347

(3)

References

348

(2)

Introduction of Rules into a Stochastic Approach for Language Modelling

350

(6)

Thierry Spriet

Marc El-Beze

Introduction

350

(1)

Stack Decoding Strategy

351

(2)

The Algorithm

351

(1)

The Evaluation Function

351

(1)

Peculiar Advantages of the Algorithm

352

(1)

Rules

353

(1)

Correction of Biases

353

(1)

Under-represented Structures and Long Span Dependencies

353

(1)

Multi Level Interactions

354

(1)

Linguistic and Syntactic

354

(1)

Phonology

354

(1)

Conclusion

355

(1)

References

355

(1)

History Integration into Semantic Classification

356

(6)

Mauro Cettolo

Anna Corazza

Introduction

356

(1)

Classifier

357

(1)

Data

357

(1)

Dialogue History Integration

358

(2)

Discussion

360

(2)

References

361

(1)

Multilingual Speech Recognition

362

(13)

E. Noth

S. Harbeck

H. Niemann

Introduction

362

(1)

Architecture of the National SQEL Demonstrators

363

(1)

Language Identification with Different Amounts of Knowledge about the Training Data

364

(6)

A System with Explicit Language Identification

365

(2)

A System with Implicit Language Identification

367

(2)

Language Identification Based on Cepstral Feature Vectors

369

(1)

Results

370

(3)

Conclusions and Future Work

373

(2)

References

373

(2)

Toward ALISP: A proposal for Automatic Language Independent Speech Processing.

375

(14)

Gerard Chollet

Jan Cernocky

Andrei Constantinescu

Sabine Deligne

Frederic Bimbot

Introduction

375

(1)

Practical benefit of ALISP

376

(1)

Issues specific to ALISP

377

(2)

Selecting features

377

(1)

Modeling speech units

377

(1)

Defining a derivation criterion

378

(1)

Building a lexicon

378

(1)

Some tools for ALISP

379

(2)

Temporal Decomposition

379

(1)

The multigram model

380

(1)

Experiments

381

(5)

Cross-Language Recognition

381

(1)

Very low bit rate speech coding

382

(2)

Mono-Speaker Continuous Speech Recognition

384

(2)

Conclusions

386

(3)

References

387

(2)

Interactive Translation of Conversational Speech

389

(15)

Alex Waibel

Introduction

389

(1)

Background

390

(2)

The Problem of Spoken Language Translation

390

(1)

Research Efforts on Speech Translation

391

(1)

JANUS-II--A Conversational Speech Translator

392

(8)

Task Domains and Data Collection

392

(2)

System Description

394

(4)

Performance Evaluation

398

(2)

Applications and Forms of Deployment

400

(4)

Interactive Dialog Translation

401

(1)

Portable Speech Translation Device

402

(1)

Passive Simultaneous Dialog Translation

402

(1)

References

403

(1)

Multimodal Speech Systems

404

(27)

Francoise D. Neel

Wolfgang M. Minker

Introduction

404

(1)

System Architecture: Knowledge Sources and Controllers

405

(9)

Environment Model

406

(1)

System Model

406

(2)

User Model

408

(2)

Task Model

410

(1)

Dialogue Model

411

(2)

Models Interdependency

413

(1)

Role of Speech in Multimodal Applications

413

(1)

Information Speech Systems

414

(13)

Spontaneous Language Characteristics

414

(2)

Case Grammar Formalism used for Task Modelling

416

(1)

Different Parsing Methods

417

(9)

Task and Dialogue Model Integration

426

(1)

Conclusion

427

(4)

References

428

(3)

Multimodal Interfaces for Multimedia Information Agents

431

(9)

Alex Waibel

Bernhard Suhm

Minh Tue Vo

Ji Yang

Introduction

431

(1)

Interpretation of Multimodal Input

432

(1)

Multimodal Components

432

(1)

Joint Interpretation

432

(1)

Multimodal Error Correction

433

(1)

Multimodal Interactive Error Repair

433

(1)

Error Repair for Multimedia Information Agents

433

(1)

Evaluating Interactive Error Repair

434

(1)

Multimodal Information Agents

434

(3)

Information Access

434

(1)

Information Creation

435

(1)

Information Manipulation

435

(1)

Information Dissemination

436

(1)

Controlling the Interface

436

(1)

The QuickDoc Application

437

(1)

Conclusions

437

(3)

References

438

(2)

Index

440

Amazon no longer offers textbook rentals. We do!

Amazon no longer offers textbook rentals. We do!

We're the #1 textbook rental company. Let us show you why.

Computational Models of Speech Pattern Processing : Proceedings of the NATO Advanced Study Institute on Computational Models of Speech Pattern Processing, Held in St. Helier, Jersey, U. K., July 7-18, 1997

9783540654780

354065478X

Supplemental Materials

Summary

Table of Contents

Supplemental Materials

Rewards Program