Speech Processing in Modern Communication

by Cohen, Israel; Benesty, Jacob; Gannot, Sharon

• ISBN13: 9783642111297
• ISBN10: 3642111297
• Edition: 1st
• Format: Hardcover
• Copyright: 2010-02-15
• Publisher: Springer Verlag

Summary

More and more devices for human-to-human and human-to-machine communications require some sophisticated algorithms. This is due to the fact that the acoustic environment in which we live in and communicate is extremely challenging. More than ever, the fundamental problems of acoustic echo cancellation, interference and noise suppression, and dereverberation need to be tackled rigorously.

Table of Contents

Linear System Identification in the Short-Time Fourier Transform Domain	p. 1
Introduction	p. 2
Problem Formulation	p. 4
System Identification Using Crossband Filters	p. 6
Crossband Filters Representation	p. 6
Batch Estimation of Crossband Filters	p. 8
Selecting the Optimal Number of Crossband Filters	p. 11
System Identification Using the MTF Approximation	p. 14
The MTF Approximation	p. 14
Optimal Window Length	p. 15
The Cross-MTF Approximation	p. 18
Adaptive Estimation of Cross-Terms	p. 19
Adaptive Control Algorithm	p. 20
Experimental Results	p. 22
Crossband Filters Estimation	p. 23
Comparison of the Crossband Filters and MTF Approaches	p. 23
CMTF Adaptation for Acoustic Echo Cancellation	p. 25
Conclusions	p. 28
Appendix	p. 28
References	p. 29
Identification of the Relative Transfer Function between Sensors in the Short-Time Fourier Transform Domain	p. 33
Introduction	p. 33
Identification of the RTF Using Multiplicative Transfer Function Approximation	p. 34
Problem Formulation and the Multiplicative Transfer Function Approximation	p. 35
RTF Identification Using Non-Stationarity	p. 36
RTF Identification Using Speech Signals	p. 37
Identification of the RTF Using Convolutive Transfer Function Approximation	p. 38
The Convolutive Transfer Function Approximation	p. 39
RTF Identification Using the Convolutive Transfer Function Approximation	p. 40
Relative Transfer Function Identification in Speech Enhancement Applications	p. 41
Blocking Matrix	p. 42
The Transfer Function Generalized Sidelobe Canceler	p. 44
Conclusions	p. 45
References	p. 46
Representation and Identification of Nonlinear Systems in the Short-Time Fourier Transform Domain	p. 49
Introduction	p. 49
Volterra System Identification	p. 51
Representation of Volterra Filters in the STFT Domain	p. 55
Second-Order Volterra Filters	p. 55
High-Order Volterra Filters	p. 59
A New STFT Model For Nonlinear Systems	p. 61
Quadratically Nonlinear Model	p. 61
High-Order Nonlinear Models	p. 65
Quadratically Nonlinear System Identification	p. 65
Batch Estimation Scheme	p. 67
Adaptive Estimation Scheme	p. 72
Experimental Results	p. 76
Performance Evaluation for White Gaussian Inputs	p. 77
Nonlinear Undermodeling in Adaptive System Identification	p. 79
Nonlinear Acoustic Echo Cancellation Application	p. 81
Conclusions	p. 83
Appendix	p. 83
References	p. 84
Variable Step-Size Adaptive Filters for Echo Cancellation	p. 89
Introduction	p. 90
Non-Parametric VSS-NLMS Algorithm	p. 92
VSS-NLMS Algorithms for Echo Cancellation	p. 96
VSS-APA for Echo Cancellation	p. 103
VFF-RLS for System Identification	p. 106
Simulations	p. 112
VSS-NLMS Algorithms for AEC	p. 112
VSS-APA for AEC	p. 117
VFF-RLS for System Identification	p. 121
Conclusions	p. 123
References	p. 124
Simultaneous Detection and Estimation Approach for Speech Enhancement and Interference Suppression	p. 127
Introduction	p. 127
Classical Speech Enhancement in Nonstationary Noise Environments	p. 129
Simultaneous Detection and Estimation for Speech Enhancement	p. 131
Quadratic Distortion Measure	p. 134
Quadratic Spectral Amplitude Distortion Measure	p. 137
Spectral Estimation Under a Transient Noise Indication	p. 140
A Priori SNR Estimation	p. 142
Experimental Results	p. 144
Simultaneous Detection and Estimation	p. 146
Spectral Estimation Under a Transient Noise Indication	p. 147
Conclusions	p. 148
References	p. 149
Speech Dereverberation and Denoising Based on Time Varying Speech Model and Autoregressive Reverberation Model	p. 151
Introduction	p. 151
Goal	p. 152
Technological Background	p. 153
Minimum Mean-Squared Error Signal Estimation and Model-Based Approach	p. 154
Dereverberation Method	p. 156
Heuristic Derivation of Weighted Prediction Error Method	p. 156
Reverberation Model	p. 160
Clean Speech Model	p. 165
Clean Speech Signal Estimator and Parameter Optimization	p. 165
Combined Dereverberation and Denoising Method	p. 167
Room Acoustics Model	p. 168
Clean Speech Model	p. 171
Clean Speech Signal Estimator	p. 172
Parameter Optimization	p. 173
Experiments	p. 179
Conclusions	p. 180
References	p. 181
Codebook Approaches for Single Sensor Speech/Music Separation	p. 183
Introduction	p. 183
Single Sensor Source Separation	p. 185
Problem Formulation	p. 185
GSMM-Based Source Separation	p. 186
AR-Based Source Separation	p. 187
Bayesian Non-Negative Matrix Factorization	p. 188
Learning the Codebook	p. 190
Multi-Window Source Separation	p. 190
General Description of the Algorithm	p. 190
Choice of a Confidence Measure	p. 191
Practical Choice of the Thresholds	p. 192
Estimation of the Expansion Coefficients	p. 193
Median Filter	p. 193
Smoothing Prior	p. 194
GMM Modeling of the Amplitude Coefficients	p. 195
Experimental Study	p. 195
Evaluation Criteria	p. 195
Experimental Setup and Results	p. 195
Conclusions	p. 196
References	p. 197
Microphone Arrays: Fundamental Concepts	p. 199
Introduction	p. 199
Signal Model	p. 200
Array Model	p. 202
Signal-to-Noise Ratio	p. 203
Array Gain	p. 204
Noise Rejection and Desired Signal Cancellation	p. 206
Beampattern	p. 207
Anechoic Plane Wave Model	p. 208
Directivity	p. 210
Superdirective Beamforming	p. 210
White Noise Gain	p. 211
Spatial Aliasing	p. 212
Monochromatic Signal	p. 215
Broadband Signal	p. 217
Mean-Squared Error	p. 218
Wiener Filter	p. 220
Minimum Variance Distortionless Response	p. 221
Conclusions	p. 222
References	p. 222
The MVDR Beamformer for Speech Enhancement	p. 225
Introduction	p. 226
Problem Formulation	p. 227
From Speech Distortion Weighted Multichannel Wiener Filter to Minimum Variance Distortionless Response Filter	p. 230
Speech Distortion Weighted Multichannel Wiener Filter	p. 230
Minimum Variance Distortionless Response Filter	p. 232
Decomposition of the Speech Distortion Weighted Multichannel Wiener Filter	p. 234
Equivalence of MVDR and Maximum SNR Beamformer	p. 235
Performance Measures	p. 235
Performance Analysis	p. 237
On the Comparison of Different MVDR Beamformers	p. 237
Local Analyzes	p. 239
Global Analyzes	p. 241
Non-Coherent Noise Field	p. 242
Coherent plus Non-Coherent Noise Field	p. 243
Performance Evaluation	p. 244
Influence of the Number of Microphones	p. 245
Influence of the Reverberation Time	p. 245
Influence of the Noise Field	p. 247
Example Using Speech Signals	p. 249
Conclusions	p. 251
Appendix	p. 251
References	p. 253
Extraction of Desired Speech Signals in Multiple-Speaker Reverberant Noisy Environments	p. 255
Introduction	p. 256
Problem Formulation	p. 259
Proposed Method	p. 261
The LCMV and MVDR Beamformers	p. 261
The Constraints Set	p. 262
Equivalent Constraints Set	p. 263
Modified Constraints Set	p. 264
Estimation of the Constraints Matrix	p. 266
Interferences Subspace Estimation	p. 267
Desired Sources RTF Estimation	p. 269
Algorithm Summary	p. 270
Experimental Study	p. 271
The Test Scenario	p. 272
Simulated Environment	p. 274
Real Environment	p. 276
Conclusions	p. 277
References	p. 278
Spherical Microphone Array Beamforming	p. 281
Introduction	p. 281
Spherical Array Processing	p. 282
Regular Beam Pattern	p. 284
Delay-and-Sum Beam Pattern	p. 286
Dolph-Chebyshev Beam Pattern	p. 288
Optimal Beamforming	p. 290
Beam Pattern with Desired Multiple Nulls	p. 293
2D Beam Pattern and its Steering	p. 295
Near-Field Beamforming	p. 297
Direction-of-Arrival Estimation	p. 300
Conclusions	p. 303
References	p. 303
Steered Beamforming Approaches for Acoustic Source Localization	p. 307
Introduction	p. 307
Signal Model	p. 309
Spatial and Spatiotemporal Filtering	p. 310
Parameterized Spatial Correlation Matrix (PSCM)	p. 311
Source Localization Using Parameterized Spatial Correlation	p. 313
Steered Response Power	p. 313
Minimum Variance Distortionless Response	p. 315
Maximum Eigenvalue	p. 316
Broadband MUSIC	p. 318
Minimum Entropy	p. 320
Sparse Representation of the PSCM	p. 326
Linearly Constrained Minimum Variance	p. 329
Autoregressive Modeling	p. 331
Challenges	p. 333
Conclusions	p. 334
References	p. 335
Index	p. 339
Table of Contents provided by Ingram. All Rights Reserved.