Speech Communications Human and Machine

by O'Shaughnessy, Douglas

• ISBN13: 9780780334496
• ISBN10: 0780334493
• Edition: 2nd
• Format: Hardcover
• Copyright: 1999-11-30
• Publisher: Wiley-IEEE Press

Summary

"Today the wireless communications industry is heavily dependent upon advanced speech coding techniques, while the integration of personal computers and voice technology is poised for growth. In this revised and updated second edition, a timely overview of the science of speech processing helps you keep pace with these rapidly developing advances. Students of electrical engineering, along with computer scientists, systems engineers, linguists, audiologists, and psychologists, will find in this one concise volume an interdisciplinary introduction to speech communication. This reference book addresses how humans generate and interpret speech and how machines simulate human speech performance and code speech for efficient transmission. With a skillful blending of the basic principles and technical detail underlying speech communication, this broad-based book offers you essential insights into the field. You will learn state-of-the-art techniques to analyze, code, recognize, and synthesize speech. In addition, you will gain a better understanding of the limits of today's technology and an informed view of future trends for speech research. SPEECH COMMUNICATIONS brings you an integrated approach to human and machine speech production and perception that is unmatched in the field. This book is complete with up-to-date references and Web addresses that will lead you to a wealth of resources for your own research into speech communication."

Author Biography

About the Author Douglas O’Shaughnessy is a professor at INRS-Telecommunication at the University of Quebec, Montreal, Canada. He is associate editor of the Journal of the Acoustical Society of America, and just completed a four-year term as an associate editor of the IEEE Transactions on Speech and Audio Processing. Dr. O’Shaughnessy’s research team is currently working on voice dialogs in English and French. He is a senior member of the Institute of Electrical and Electronics Engineers (IEEE), and a Fellow of the Acoustical Society of America.

Table of Contents

Preface	xvii
Acknowledgments	xxi
Acronyms in Speech Communications	xxiii
Important Developments in Speech Communications	xxv
Introduction	1	(8)
What Is Speech Communication?	1	(1)
Developments in Speech Communication	1	(1)
Outline of the Book	2	(5)
Production of Speech	3	(1)
Sound Perception	3	(1)
Speech Analysis	3	(1)
Speech Coding	4	(1)
Speech Enhancement	5	(1)
Speech Synthesis	6	(1)
Speech and Speaker Recognition	6	(1)
Other Topics	7	(2)
Review of Mathematics for Speech Processing	9	(26)
Mathematical Preliminaries	9	(3)
Number Representations	9	(1)
Matrix Arithmetic	10	(2)
Signals and Linear Systems	12	(4)
Simple Signals	13	(3)
Filtering and Convolution	16	(1)
Frequency Analysis	16	(3)
Fourier Transform	17	(1)
Spectra and Correlation	18	(1)
Laplace Transform: Poles and Zeros	18	(1)
Circuits	19	(1)
Discrete-Time Signals and Systems	20	(5)
Sampling	20	(2)
Frequency Transforms of Discrete-Time Signals	22	(1)
Decimation and Interpolation	23	(2)
Filters	25	(4)
Bandpass Filters	26	(1)
Digital Filters	26	(1)
Difference Equations and Filter Structures	27	(2)
Probability and Statistics	29	(4)
Probability Densities and Histograms	30	(1)
Averages and Variances	31	(1)
Gaussian Probability Density	31	(1)
Joint Probability	32	(1)
Noise	33	(1)
Summary	33	(2)
Speech Production and Acoustic Phonetics	35	(74)
Introduction	35	(2)
Anatomy and Physiology of the Speech Organs	37	(11)
The Lungs and the Thorax	38	(1)
Larynx and Vocal Folds (Cords)	39	(6)
Vocal Tract	45	(3)
Articulatory Phonetics	48	(8)
Manner of Articulation	50	(2)
Structure of the Syllable	52	(1)
Voicing	52	(1)
Place of Articulation	53	(2)
Phonemes in Other Languages	55	(1)
Articulatory Models	55	(1)
Acoustic Phonetics	56	(12)
Spectrograms	56	(1)
Vowels	57	(3)
Diphthongs	60	(2)
Glides and Liquids	62	(2)
Nasals	64	(1)
Fricatives	65	(1)
Stops (Plosives)	65	(2)
Variants of Normal Speech	67	(1)
Acoustic Theory of Speech Production	68	(20)
Acoustics of the Excitation Source	68	(2)
Acoustics of the Vocal Tract	70	(8)
Transmission Line Analog of the Vocal Tract	78	(8)
Effects of Losses in the Vocal Tract	86	(1)
Radiation at the Lips	87	(1)
Model of Glottal Excitation	87	(1)
Quantal Theory of Speech Production	88	(1)
Practical Vocal Tract Models for Speech Analysis and Synthesis	88	(7)
Articulatory Model	89	(4)
Terminal-Analog Model	93	(2)
Coarticulation	95	(6)
Where Does Coarticulation Occur?	96	(1)
Coarticulation Effects for Different Articulators	96	(2)
Invariant Features	98	(2)
Effects of Coarticulation on Duration	100	(1)
Models for Coarticulation	100	(1)
Prosody (Suprasegmentals)	101	(6)
Duration	102	(1)
Effects of Stress and Speaking Rate	103	(1)
Fundamental Frequency (F0)	104	(3)
Conclusion	107	(2)
Problems	107	(2)
Hearing	109	(32)
Introduction	109	(1)
Anatomy and Physiology of the Ear	109	(10)
Outer Ear	110	(1)
Middle Ear	111	(1)
Inner Ear	111	(2)
Basilar Membrane (BM) Behavior	113	(2)
Electrical Activity in the Auditory Neurons	115	(4)
Adaptation	119	(1)
Sound Perception	119	(12)
Auditory Psychophysics	120	(1)
Thresholds	120	(2)
Just-Noticeable Differences (JNDs)	122	(1)
Pitch Perception	123	(2)
Masking	125	(2)
Critical Bands	127	(1)
Nonsimultaneous or Temporal Masking	128	(2)
Origins of Masking	130	(1)
Release from Masking (‡)	130	(1)
Sound Localization (‡)	131	(1)
Response of the Ear to Complex Stimuli	131	(7)
Speech Stimuli (‡)	132	(1)
Masking Due to Complex Stimuli (‡)	132	(1)
Adaptation (‡)	133	(1)
Just-Noticeable Differences (JNDs) in Speech	134	(2)
Timing	136	(1)
Separating Sound Sources	137	(1)
Conclusion	138	(3)
Problems	138	(3)
Speech Perception	141	(32)
Introduction	141	(1)
Perceptually-Important Features of Speech Signals	142	(2)
Synthetic vs. Natural Speech	143	(1)
Redundancy in Speech	143	(1)
Models of Speech Perception	144	(6)
Categorical Perception	144	(1)
Distinctive Features	145	(1)
Active Models	146	(2)
Passive Models	148	(2)
Vowel Perception	150	(2)
Perceived Formant Location in Synthetic Vowels	150	(1)
Context Normalization	150	(1)
Coarticulation Effects	151	(1)
Consonant Perception	152	(8)
Perception of the Manner of Articulation Feature	152	(2)
Perception of Place of Articulation of Consonants	154	(4)
Perception of Voicing in Obstruents	158	(2)
Duration as a Phonemic Cue	160	(2)
Manner Cues	160	(1)
Place Cues	161	(1)
Speaking Rate Effects	161	(1)
Intonation: Perception of Prosody	162	(6)
Stress: Lexical and Sentential	163	(1)
Acoustic Correlates of Stress	164	(1)
Perception of Syntactic Features	165	(3)
Perceptually Relevant Pitch Movements	168	(1)
Other Aspects of Speech Perception (‡)	168	(3)
Adaptation Studies	168	(1)
Dichotic Studies	169	(1)
Phase Effects	169	(1)
Word and Syllable Effects	170	(1)
Perception of Distorted Speech	170	(1)
Speech Perception by the Handicapped	171	(1)
Conclusion	171	(2)
Problems	172	(1)
Speech Analysis	173	(56)
Introduction	173	(1)
Short-Time Speech Analysis	174	(5)
Windowing	175	(1)
Spectra of Windows: Wide-and Narrow-Band Spectrograms	176	(3)
Time-Domain Parameters	179	(6)
Signal Analysis in the Time Domain	179	(1)
Short-Time Average Energy and Magnitude	180	(1)
Short-Time Average Zero-Crossing Rate (ZCR)	181	(2)
Short-Time Autocorrelation Function	183	(2)
Frequency-Domain (Spectral) Parameters	185	(7)
Filter-Bank Analysis	186	(1)
Short-Time Fourier Transform Analysis	186	(2)
Spectral Displays	188	(1)
Formant Estimation and Tracking	189	(2)
Other Spectral Methods (‡)	191	(1)
Energy Separation (‡)	191	(1)
Linear Predictive Coding (LPC) Analysis	192	(18)
Basic Principles of LPC	192	(1)
Least-Squares Autocorrelation Method	193	(3)
Least-Squares Covariance Method	196	(1)
Computational Considerations	197	(1)
Spectral Estimation via LPC	198	(4)
Updating the LPC Model Sample by Sample	202	(1)
Transversal Predictors	202	(1)
Lattice LPC Models	203	(3)
Window Considerations	206	(1)
Modifications to Standard LPC	207	(1)
Emphasizing Low Frequencies	208	(1)
Pole-Zero LPC Models	209	(1)
Cepstral Analysis	210	(5)
Mathematical Details of Cepstral Analysis	211	(1)
Applications for the Cepstrum	212	(2)
Mel-Scale Cepstrum	214	(1)
Other Spectral Estimation Methods (‡)	215	(2)
Karhunen-Loeve Transform (KLT)	215	(1)
Wavelets	216	(1)
Winger Distribution	216	(1)
Other Recent Techniques	217	(1)
F0 (``Pitch'') Estimation	217	(5)
Time-Domain F0 Techniques	218	(1)
Short-Time Spectral Techniques	219	(3)
Robust Analysis	222	(1)
Reducation of Information	222	(3)
Taking Advantage of Gradual Vocal Tract Motion	223	(1)
Smoothing: Linear and Nonlinear	224	(1)
Summary	225	(4)
Problems	226	(3)
Coding of Speech Signals	229	(94)
Introduction	229	(3)
Coding Noise	229	(1)
Applications	230	(1)
Quality	230	(1)
Classes of Coders	231	(1)
Chapter Overview	231	(1)
Quantization	232	(13)
Quantization Error or Noise	233	(2)
Bit Protection	235	(1)
Signal-to-Noise Ratio (SNR)	236	(1)
Nonuniform Quantization	237	(3)
Relationship of Bandwidth and Noise to Coding Rate	240	(1)
Vector Quantization (VQ)	241	(4)
Speech Redundancies (Characteristics to Exploit)	245	(1)
Measures to Evaluate Speech Quality	245	(1)
Time-Domain Waveform Coding	246	(12)
Basic Time-Adaptive Waveform Coding	247	(3)
Exploiting Properties of the Spectral Envelope	250	(3)
Exploiting the Periodicity of Voiced Speech	253	(3)
Exploiting Auditory Limitations (Noise Shaping)	256	(2)
Linear Predictive Coding (LPC)	258	(20)
Linear Delta Modulation (LDM)	260	(2)
Adaptive Delta Modulation (ADM)	262	(2)
Adaptive Differential Pulse-Code Modulation (ADPCM)	264	(2)
Linear Predictive Coders (LPC) vs. Linear Predictive Analysis-by-Synthesis (LPAS)	266	(1)
Equivalent Forms for LPC Coefficients	267	(1)
Line Spectrum Pairs/Frequencies	268	(1)
Parameter Updating and Transmission	269	(1)
Variable-Frame-Rate (VFR) Transmission	269	(1)
Transmission Details	270	(1)
Time-Varying LPC Coefficients (‡)	270	(1)
Different Excitation Models	271	(6)
Waveform Interpolation	277	(1)
Spectral (Frequency-Domain) Coders	278	(17)
Filter Bank Analysis	278	(2)
Sub-Band Coders (SBC)	280	(4)
Adaptive Transform Coders (ATC)	284	(5)
Harmonic Coding	289	(6)
Other Vocoders (Non-LP Source Voice Coders) (‡)	295	(7)
Phase Vocoder	296	(2)
Channel Vocoders	298	(1)
Excitation for Vocoders	299	(1)
Homomorphic (Cepstral) Vocoder	300	(1)
Other Vocoders	301	(1)
Vector Quantization (VQ) Coders	302	(11)
Split VQ Coders	303	(1)
Gain/Shape Vector Quantization	303	(1)
Other Types of Vector Quantization	304	(3)
Vector Quantization of LPC	307	(1)
Code-Excited Linear Prediction (CELP)	308	(4)
Very-Low-Rate LPC Vocoders	312	(1)
Network and Application Considerations	313	(5)
Packet Transmission	314	(1)
Time Assignment Speech Interpolation (TASI)	315	(1)
Embedded Coding	316	(1)
Tandeming of Coders (‡)	317	(1)
Hardware Implementation: Integrated Circuits	318	(1)
Summary	319	(4)
Problems	320	(3)
Speech Enhancement	323	(14)
Introduction	323	(1)
Background	324	(1)
Nature of Interfering Sounds	325	(2)
Speech Enhancement (SE) Techniques	327	(1)
Spectral Subtraction and Filtering	327	(1)
Harmonic Filtering	327	(1)
Parametric Resynthesis	327	(1)
Spectral Subtraction (SS)	328	(1)
Filtering and Adaptive Noise Cancellation	329	(4)
Filtering	329	(1)
Multi-Microphone Adaptive Noise Cancellation (ANC)	330	(3)
Methods Involving Fundamental Frequency Tracking	333	(2)
Enhancement by Resynthesis	335	(1)
Summary	336	(1)
Speech Synthesis	337	(30)
Introduction	337	(1)
Principles of Speech Synthesis	338	(8)
Types of Stored Speech Units to Concatenate	339	(2)
Memory Size	341	(2)
Synthesis Method	343	(1)
Limited-Text (Voice-Response) Systems	343	(1)
Unrestricted-Text (TTS) Systems	344	(2)
Synthesizer Methods	346	(14)
Articulatory Synthesis (‡)	346	(3)
Formant Synthesis	349	(3)
Linear Predictive Coding (LPC) Synthesis	352	(1)
Specifying Parameter Trajectories (‡)	353	(2)
Intraframe Parameter Updating (‡)	355	(1)
Excitation Modeling (‡)	356	(3)
Waveform Concatenation	359	(1)
Synthesis of Intonation	360	(3)
Speech Synthesis for Different Speakers	363	(1)
Speech Synthesis in Other Languages	364	(1)
Evaluation of TTS Systems	364	(1)
Practical Speech Synthesis	365	(1)
Conclusion	365	(2)
Problems	366	(1)
Automatic Speech Recognition	367	(70)
Introduction	367	(5)
ASR Search: Vast and Expensive	367	(2)
Variability in Speech Signals	369	(1)
Segmenting Speech into Smaller Units	369	(1)
Performance Evaluation	370	(1)
Databases for Speech Recognition	371	(1)
Basic Pattern Recognition Approach	372	(3)
Pattern Recognition Methods	372	(1)
Different Viewpoints Toward ASR	373	(2)
Preprocessing	375	(1)
Parametric Representation	375	(3)
Parameters Used in Recognition	377	(1)
Feature Extraction	377	(1)
Evaluating the Similarity of Speech Patterns	378	(8)
Frame-Based Distance Measures	379	(5)
Making ASR Decisions	384	(2)
Accommodating Both Spectral and Temporal Variability	386	(16)
Segmenting Speech into Smaller Units	387	(2)
Dynamic Time Warping	389	(10)
Applying Vector Quantization to ASR	399	(3)
Networks for Speech Recognition	402	(10)
Hidden Markov Models (HMMs)	402	(10)
Adapting to Variability in Speech	412	(5)
Intraspeaker Variability (Speaker Freedom)	412	(1)
Interspeaker Variability (Everybody's Different)	412	(3)
Environmental Variability (Noise Robustness)	415	(2)
Language Models (LMs)	417	(3)
Grammars in ASR	417	(1)
Integrating Language Models into ASR	418	(2)
Search Design	420	(4)
Efficient Searches	420	(3)
Allowing Vocabulary Freedom	423	(1)
Out-of-Vocabulary Words	424	(1)
Artificial Neural Networks	424	(3)
Training ANNs	425	(1)
Accommodating Timing in ANNs	426	(1)
Expert-System Approach to ASR	427	(5)
Segmenting Speech into Syllables	427	(1)
Segmentation of Continuous Speech into Phones	428	(2)
Labeling Phones	430	(1)
Phonological Rules	431	(1)
Using Prosodics to Aid Recognition	432	(1)
Commercial Systems	432	(1)
Summary of Current ASR Design	433	(1)
Conclusion	433	(4)
Problems	434	(3)
Speaker Recognition	437	(24)
Introduction	437	(1)
Verification vs. Recognition	438	(1)
Recognition Techniques	439	(11)
Model Evaluation	440	(1)
Text Dependence	440	(1)
Statistical vs. Dynamic Features	441	(2)
Stochastic Models	443	(1)
Vector Quantization	443	(2)
Similarity and Distance Measures	445	(2)
Cepstral Analysis	447	(1)
Orthogonal LPC Parameters (‡)	447	(2)
Neural Network Approaches	449	(1)
Features that Distinguish Speakers	450	(3)
Measures of the Effectiveness of Features for Recognition	450	(1)
Techniques to Choose Features	451	(1)
Spectral Features	452	(1)
Prosodic Features	452	(1)
System Design	453	(3)
Data Collection	453	(1)
Sequential Decision Strategy	454	(1)
Multiple-Stage Recognition	454	(1)
Effects of Different Communication Channels (‡)	455	(1)
Language and Accent Identification	456	(1)
Speaker Recognition by Humans	457	(1)
Conclusion	458	(3)
Problems	458	(3)
Appendix: Computer Sites for Help on Speech Communication	461	(8)
References	469	(68)
Index	537	(10)
About the Author	547