Speaker Classification I: Fundamentals, Features, and Methods (Lecture Notes in Computer Science, 4343)

✍ Scribed by Christian Müller (editor)

• Publisher: Springer
• Year: 2007
• Tongue: English
• Leaves: 363
• Category: Library

✦ Synopsis

As well as conveying a message in words and sounds, the speech signal carries information about the speaker's own anatomy, physiology, linguistic experience and mental state. These speaker characteristics are found in speech at all levels of description: from the spectral information in the sounds to the choice of words and utterances themselves.

This volume and its companion volume, LNAI 4441, constitute a state-of-the-art survey for the field of speaker classification. They approach the following questions: What characteristics of the speaker become manifest in his or her voice and speaking behavior? Which of them can be inferred from analyzing the acoustic realizations? What can this information be used for? Which methods are the most suitable for diversified problems in this area of research? How should the quality of the results be evaluated?

The 19 contributions to this volume comprise general and overview-like articles that are organized in topical sections on fundamentals, characteristics, applications, methods and features, as well as evaluation.

✦ Table of Contents

Title page
Preface
Table of Contents
How Is Individuality Expressed in Voice? An Introduction to Speech Production and Description for Speaker Classification
Introduction
Vocal Apparatus
Sub-laryngeal Vocal Tract
The Larynx
Supra-laryngeal Vocal Tract
Sound Generation
Phonetic Classification
Place and Manner of Articulation
The IPA Chart
Vowel Classification
Further Aspects of Vowel Classification
Multiple Articulation
Non-pulmonic Airstreams
Beyond the Segment
Expressions of Individuality
Individuality in Language and Language Use
Individuality in the Sound System
Individuality in Controlling the Speech Production Process
Anatomical Influences on Individuality
Other Influences on Individuality
From Individuality to Identity
Applications
References
Speaker Classification Concepts: Past, Present and Future
Introduction
Reasons for Wanting Speaker Classification
More Structured Approaches
Some Comments on Speaker Classification in the Context of Verification/Identification
Some Comments on the Foulkes & Barron [18] Experiment
Perceptual Studies
The Gnuspeech System
Possibilities and Current Limitations of the Experimental System
Face Recognition as an Analog of the Speaker Recognition Problem
Back to the Main Goal - Speaker Classification
Intonation and Rhythm
Lower Level Cues: Segmental Level and Below
Dynamics and Longer Term Effects
Recognizing Speakers Gender and Age, and Sexual Orientation
Conclusions
References
Speaker Characteristics
Introduction
Why? - Applications to Speaker Characteristics
Human-Computer Interaction Systems
Human-Centered Systems
Adaptation of System Components
Summary
What? A Taxonomy of Speaker Characteristics
Language-Dependent Speaker Characteristics
How? - Automatic Classification of Speaker Characteristics
Speaker Recognition
Language Identification
A Classification System for Speaker Characteristics
Multilingual Phone Sequences
Classification of Speaker Identity
Classification of Gender
Classification of Accent
Classification of Proficiency Level
Classification of Language
Classification of Attentional State
Language Dependencies
Conclusion
References
Foreign Accent
Introduction
Acoustic Correlates of Foreign Accent
Durational Features of Foreign Accent: Speech Rate
Durational Features of Foreign Accent: Reduction
Pitch Range and Pitch Movement in Foreign-Accented Speech
Perceptual Correlates of Foreign Accent
Summary and Conclusion
References
Acoustic Analysis of Adult Speaker Age
Introduction
Ageing of the Speech Production Mechanism
Respiratory System
Larynx
Supralaryngeal System
Neuromuscular Control
Female and Male Ageing
Perception and Automatic Recognition of Speaker Age
Human Perception of Speaker Age
Automatic Recognition of Speaker Age
Acoustic Correlates of Adult Speaker Age
General Variation
Speech Rate
Sound Pressure Level (SPL)
Fundamental Frequency $\mathrm{F}{0}$
Variation in $\mathrm{F}{0}$ and Amplitude
Other Voice Measures
Resonance Measures
Factors Which May Influence Acoustic Analysis of Speaker Age
Speaker-Related Factors
Speech-Material-Related Factors
Methodological Factors
References
Speech Under Stress: Analysis, Modeling and Recognition
Introduction
Domains of Speech Under Stress
Domain A: Production
Domain B: Perception
Domain C: Speech Systems
Analysis
Analysis of Fundamental Frequency
Analysis of Duration
Analysis of Intensity
Glottal Pulse Shaping
Vocal Tract Spectrum
Applications
Speech Recognition
Stress Detection
Detection-Theory-Based Framework for Stress Classification
A Distance Measure for Stress Classification
Neural Network Based Systems
Stress Classification Using Nonlinear Speech Features
Synthesis and Conversion of Speech Under Stress
Speech Coding System
Discussions and Future Directions
References
Speaker Characteristics and Emotion Classification
Introduction
Setting the Scene
Concepts: Emotion and Speaker Characteristics
Personalization and Data Acquisition: A Problem
A Tentative Relevance Hierarchy for Speaker-Independent Emotion Recognition in Spontaneous Speech
An Example: Laryngealizations
Another Example: Pitch
Implications from Applications
Voice Application Setup
System Architecture
Concluding Remarks
References
Emotions in Speech: Juristic Implications
Introduction
Effects in Legal Contexts
Assessment of Emotion in Others
Emotions and Memory
Emotions and Culture
Emotions in Legal Scholarship
Emotions in Speech
Perceptual Studies
Acoustic Studies
Implications for Training and Assessment of Legal Actors
Interaction Skills Training
Situated Assessment
Admissibility of Machine-Detected Emotion as Evidence
Summary
References
Application of Speaker Classification in Human Machine Dialog Systems
Contents and Motivation
A Taxonomy for Speaker Classification Applications
Application Scenarios
Conclusion
References
Speaker Classification in Forensic Phonetics and Acoustics
Introduction
Speaker Classification Characteristics in Current Forensic $Practice^2$
Gender
Age
Dialect
Foreign Accent
Sociolect
Medical Conditions
Current Issues in Forensic Speaker Classification: Auditory vs. Acoustic-Phonetic Analysis
Introduction
New Information on Established Speaker Classification Characteristics: Gender
New Speaker Classification Characteristics: Speaker Height
Conclusion
General Conclusion
References
Forensic Automatic Speaker Classification in the “Coming Paradigm Shift”
Introduction
The Coming Paradigm Shift and Forensic Speaker Recognition
The Bayesian Methodology
Proficiency Testing in Automatic Forensic Speaker Classification
Calibration in Bayesian Forensic Speaker Classification
Assessing Calibration in Forensic Speaker Classification
Calibration Example
Experiments
Results
Conclusions
References
The Many Roles of Speaker Classification in Speaker Verification and Identification
Introduction
Variability
Anti-speaker Modeling
Disguised Voices
Stress and Lie Detection
Speaker Segmentation and Clustering
Conclusion
References
Frame Based Features
Introduction
Linear Prediction Coding
A Simple Model of Speech Tract
Yule-Walker-Equations
Mel-Frequency Cepstrum Coefficients
Derivation
Dynamic Features
Wavelet Based Features
Comparison of MFCCs and Wavelets for Speaker Recognition and Speech Recognition
Data Sets
Feature Derivation
The Speech Recognition System
The Speaker Recognition System
Results
Discussion
References
Higher-Level Features in Speaker Recognition
Introduction
Overview and Classification of Approaches
Cepstral and Cepstral-Derived Features
Acoustic Tokenization (“Phonetic”) Features
Prosodic Features
Lexical Features
Performance in a Recent System
Task and Data
ASR system
Session Variability Compensation and TNORM
Systems
Results
Conclusions and Implications for Speaker Classification
References
Enhancing Speaker Discrimination at the Feature Level
Introduction
Data
Baseline Speaker Identification System
Feature Enhancement
Size of the Speaker Basis
Improving Speaker Basis Selection
Feature Enhancement for Channel Noise
Feature Enhancement for Added Noise
Matched Noise Conditions
Mismatched Noise Conditions
Multi-condition Training
Conclusions and Discussion
References
Classification Methods for Speaker Recognition
Introduction
Feature Extraction
Models and Classifiers
Gaussian Mixture Modeling (GMM)
Sequence Kernels for Speaker Recognition—Specific Examples
Support Vector Machine (SVM)
Hidden Markov Modeling (HMM)
Artificial Neural Networks
Normalization Techniques
Classifier Choice
Conclusions
References
Multi-stream Fusion for Speaker Classification
Introduction
Feature Space
Score Space
Model Space
Factorial HMMs
Coupled HMMs
Kernel Space
Affect Recognition
Empirical Results
Discussion
References
Evaluations of Automatic Speaker Classification Systems
The Challenge
The NIST Evaluations
Evaluation Parameters
Channel Variability
Speaker Variability
Measuring Progress
Multi-speaker
Other Evaluations
Future of Speaker Evaluation
References
An Introduction to Application-Independent Evaluation of Speaker Recognition Systems
Introduction
Recognition, Verification, Detection, Identification
The Traditional Approach of the Evaluation of Speaker Recognition Systems
The Errors in Detection
The DET-Plot: A Measure of Discrimination
The Detection Cost Function: Simultaneous Measure of Discrimination and Calibration
The Log-Likelihood-Ratio
Log-Likelihood-Ratio Cost Function
Discrimination/Calibration Decomposition: The PAV Algorithm
The APE-Curve: Graph of the $C_{llr}$ Integral
Information-Theoretic Interpretation of $C_{llr}$
Comparison of Systems: DETs and APEs
Conclusion
References
Author Index