ISBN: 3540419535-c
TITLE: Microphone Arrays
AUTHOR: Brandstein/Ward
TOC:

Part I. Speech Enhancement
1 Constant Directivity Beamforming
Darren B. Ward, Rodney A. Kennedy, Robert C. Williamson 3
1.1 Introduction 3
1.2 Problem Formulation 6
1.3 Theoretical Solution 7
1.3.1 Continuous sensor 7
1.3.2 Beam-shaping function 8
1.4 Practical Implementation 9
1.4.1 Dimension-reducing parameterization 9
1.4.2 Reference beam-shaping filter 11
1.4.3 Sensor placement 12
1.4.4 Summary of implementation 12
1.5 Examples 13
1.6 Conclusions 16
References 16
2 Superdirective Microphone Arrays
Joerg Bitter, K. Uwe Simmer 19
2.1 Introduction 19
2.2 Evaluation of Beamformers 20
2.2.1 Array-Gain 21
2.2.2 Beampattern 22
2.2.3 Directivity 23
2.2.4 Front-to-Back Ratio 24
2.2.5 White Noise Gain 24
2.3 Design of Superdirective Beamformers 24
2.3.1 Delay-and-Sum Beamformer 26
2.3.2 Design for spherical isotropic noise 26
2.3.3 Design for Cylindrical Isotropic Noise 30
2.3.4 Design for an Optimal Front-to-Back Ratio 30
2.3.5 Design for Measured Noise Fields 32
2.4 Extensions and Details 33
2.4.1 Alternative Form 33
2.4.2 Comparison with Gradient Microphones 35
2.5 Conclusion 36
References 37
3 Post-Filtering Techniques
K. Uwe Simmer, Joerg Bitzer, Claude Marro 39
3.1 Introduction 39
3.2 Multi-channel Wiener Filtering in Subbands 41
3.2.1 Derivation of the Optimum Solution 41
3.2.2 Factorization of the Wiener Solution 42
3.2.3 Interpretation 45
3.3 Algorithms for Post-Filter Estimation 46
3.3.1 Analysis of Post-Filter Algorithms 47
3.3.2 Properties of Post-Filter Algorithms 49
3.3.3 A New Post-Filter Algorithm 50
3.4 Performance Evaluation 51
3.4.1 Simulation System 52
3.4.2 Objective Measures 52
3.4.3 Simulation Results 54
3.5 Conclusion 57
4 Spatial Coherence Functions for Differential Microphones in Isotropic Noise Fields
Gary W. Elko 61
4.1 Introduction 61
4.2 Adaptive Noise Cancellation 61
4.3 Spherically Isotropic Coherence 65
4.4 Cylindrically Isotropic Fields 73
4.5 Conclusions 77
References 84
5 Robust Adaptive Beamforming
Osamu Hoshuyama, Akihiko Sugiyama 87
5.1 Introduction 87
5.2 Adaptive Beamformers 88
5.3 Robustness Problem in the GJBF 90
5.4 Robust Adaptive Microphone Arrays - Solutions to Steering-Vector Errors 92
5.4.1 LAF-LAF Structure 92
5.4.2 CCAF-LAF Structure 94
5.4.3 CCAF-NCAF Structure 95
5.4.4 CCAF-NCAF Structure with an AMC 97
5.5 Software Evaluation of a Robust Adaptive Microphone Array 99
5.5.1 Simulated Anechoic Environment 99
5.5.2 Reverberant Environment 101
5.6 Hardware Evaluation of a Robust Adaptive Microphone Array 104
5.6.1 Implementation 104
5.6.2 Evaluation in a Real Environment 104
5.7 Conclusion lO6
References 106
6 GSVD-Based Optimal Filtering for Multi-Microphone Speech Enhancement
Simon Doclo, Marc kfoonen 111
6.1 Introduction 111
6.2 GSVD-Based Optimal Filtering Technique 113
6.2.1 Optimal Filter Theory 114
6.2.2 General Class of Estimators 116
6.2.3 Symmetry Properties for Time-Series Filtering 117
6.3 Performance of GSVD-Based Optimal Filtering 118
6.3.1 Simulation Environment 118
6.3.2 Spatial Directivity Pattern 119
6.3.3 Noise Reduction Performance 121
6.3.4 Robustness Issues 121
6.4 Complexity Reduction 122
6.4.1 Linear Algebra Techniques for Computing GSVD 122
6.4.2 Recursive and Approximate GSVD-Updating Algorithms 123
6.4.3 Downsampling Techniques 125
6.4.4 Simulations 125
6.4.5 Computational Complexity 126
6.5 Combination with ANC Postprocessing Stage 127
6.5.1 Creation of Speech and Noise References 127
6.5.2 Noise Reduction Performance of ANC Postprocessing Stage 128
6.5.3 Comparison with Standard Beamforming Techniques 129
6.6 Conclusion 129
References 130
7 Explicit Speech Modeling for Microphone Array Speech Acquisition
Michael Brandstein, Scott Griebel 133
7.1 Introduction 133
7.2 Model-Based Strategies 136
7.2.1 Example 1: A Frequency-Domain Model-Based Algorithm 137
7.2.2 Example 2: A Time-Domain Model-Based Algorithm 140
7.3 Conclusion 148
References 151
Part II. Source Localization
8 Robust Localization in Reverberant Rooms
Joseph H. DiBiase, Harvey F. Silvemnan, Michael S. Brandstein 157
8.1 Introduction 157
8.2 Source Localization Strategies 158
8.2.1 Steered-Beamformer-Based Locators 159
8.2.2 High-Resolution Spectral-Estimation-Based Locators 160
8.2.3 TDOA-Based Locators 161
8.3 A Robust Localization Algorithm 164
8.3.1 The Impulse Response Model 164
8.3.2 The GCC and PHAT Weighting Function 166
8.3.3 ML TDOA-Based Source Localization 167
8.3.4 SRP-Based Source Localization 169
8.3.5 The SRP-PHAT Algorithm 170
8.4 Experimental Comparison 172
References 178
9 Multi-Source Localization Strategies
Elio D. Di Claudia, Raffaele Parisi 181
9.1 Introduction 181
9.2 Background 184
9.2.1 Array Signal Model 184
9.2.2 Incoherent Approach 185
9.2.3 Coherent Signal Subspace Method (CSSM) 185
9.2.4 Wideband Weighted Subspace Fitting (WB-WSF) 186
9.3 The Issue of Coherent Multipath in Array Processing 187
9.4 Implementation Issues 188
9.5 Linear Prediction-ROOT-MUSIC TDOA Estimation 189
9.5.1 Signal Pre-Whitening 189
9.5.2 An Approximate Model for Multiple Sources in Reverberant Environments 191
9.5.3 Robust TDOA Estimation via ROOT-MUSIC 192
9.5.4 Estimation of the Number of Relevant Reflections 194
9.5.5 Source Clustering. 195
9.5.6 Experimental Results 196
References 198
10 Joint Audio-Video Signal Processing for Object Localization and Tracking
Norbert Strobel, Sascha Spars, Rudolf Rabenstein 203
10.1 Introduction 203
10.2 Recursive State Estimation 205
10.2.1 Linear Kalman Filter 206
10.2.2 Extended Kalman Filter due to a Measurement Nonlinearity 210
10.2.3 Decentralized Kalman Filter 212
10.3 Implementation 218
10.3.1 System description 218
10.3.2 Results 219
10.4 Discussion and Conclusions 221
References 222
Part III. Applications
11 Microphone-Array Hearing Aids
Julie E. Greenberg, Patrick M. Zurek 229
ll.l Introduction 229
11.2 Implications for Design and Evaluation 230
11.2.1 Assumptions Regarding Sound Sources 230
11.2.2 Implementation Issues 231
11.2.3 Assessing Performance 232
11.3 Hearing Aids with Directional Microphones 233
11.4 Fixed-Beamforming Hearing Aids 234
11.5 Adaptive-Beamforming Hearing Aids 235
11.5.1 Generalized Sidelobe Canceler with Modifications 236
11.5.2 Scaled Projection Algorithm 242
11.5.3 Direction of Arrival Estimation 243
11.5.4 Other Adaptive Approaches and Devices 243
11.6 Physiologically-Motivated Algorithms 244
11.7 Beamformers with Binaural Outputs 245
11.8 Discussion 246
References 249
12 Small Microphone Arrays with Postfilters for Noise and Acoustic Echo Reduction
Rainer Martin 255
12.1 Introduction 255
12.2 Coherence of Speech and Noise 257
12.2.1 The Magnitude Squared Coherence 257
12.2.2 The Reverberation Distance 258
12.2.3 Coherence of Noise and Speech in Reverberant Enclosures 259
12.3 Analysis of the Wiener Filter with Symmetric Input Signals 263
12.3.1 No Near End Speech 265
12.3.2 High Signal to Noise Ratio 265
12.4 A Noise Reduction Application 266
12.4.1 An Implementation Based on the NLMS Algorithm 266
12.4.2 Processing in the 809 - 3600 Hz Band. 268
12.4.3 Processing in the 240 - 800 Hz Band 269
12.4.4 Evaluation 269
12.4.5 Alternative Implementations of the Coherence Based Postfilter
12.5 Combined Noise and Acoustic Echo Reduction 271
12.5.1 Experimental Results 274
12.6 Conclusions 275
References 276
13 Acoustic Echo Cancellation for Beamforming Microphone Arrays
Walter L. Kellermann 281
13.1 Introduction 281
13.2 Acoustic Echo Cancellation 282
13.2.1 Adaptation algorithms 284
13.2.2 AEC for multi-channel sound reproduction 287
13.2.3 AEC for multi-channel acquisition 287
13.3 Beamforming 288
13.3.1 General structure 288
13.3.2 Time-invariant beamforming 290
13.3.3 Time-varying beamforming 291
13.3.4 Computational complexity 292
13.4 Generic structures for combining AEC with beamforming 292
13.4.1 Motivation 292
13.4.2 Basic options 293
13.4.3 'AEC first' 293
13.4.4 'Beamforming first' 296
13.5 Integration of AEC into time-varying beamforming 297
13.5.1 Cascading time-invariant and time-varying beamforming 297
13.5.2 AEC with GSC-type beamforming structures 301
13.6 Combined AEC and beamforming for multi-channel recording and multi-channel reproduction 302
13.7 Conclusions 303
References 303
14 Optimal and Adaptive Microphone Arrays for Speech In-put in Automobiles
Suen Nordholm, Ingvar Claesson, Nedelko Grbic 307
14.1 Introduction: Hands-Free Telephony in Cars 307
14.2 Optimum and Adaptive Beamforming 309
14.2.1 Common Signal Modeling 309
14.2.2 Constrained Minimum Variance Beamforming and the Generalized Sidelobe Canceler 310
14.2.3 In Situ Calibrated Microphone Array (ICMA) 312
14.2.4 Time-Domain Minimum-Mean-Square-Error Solution 313
14.2.5 Frequency-Domain Minimum-Mean-Square-Error Solution 314
14.2.6 Optimal Near-Field Signal-to-Noise plus Interference Beam-former 316
14.3 Subband Implementation of the Microphone Array 317
14.3.1 Description of LS-Subband Beamforming 318
14.4 Multi-Resolution Time-Frequency Adaptive Beamforming 319
14.4.1 Memory Saving and Improvements 319
14.5 Evaluation and Examples 320
14.5.Car Environment 320
14.5.2 Microphone Configurations 321
14.5.3 Performance Measures 321
14.5.4 Spectral Performance Measures 322
14.5.5 Evaluation on car data 323
14.5.6 Evaluation Results 323
14.6 Summary and Conclusions 324
References 326
15 Speech Recognition with Microphone Arrays
Maurizio Omologo, Marco Matassoni, Piergiorgio Svaizer 331
15.1 Introduction. 331
15.2 State of the Art 332
15.2.1 Automatic Speech Recognition 332
15.2.2 Robustness in ASR 336
15.2.3 Microphone Arrays and Related Processing for ASR 337
15.2.4 Distant-Talker Speech Recognition 339
15.3 A Microphone Array-Based ASR System 342
15.3.1 System Description 342
15.3.2 Speech Corpora and Task 345
15.3.3 Experiments and Results 346
15.4 Discussion and Future Trends 348
References 349
16 Blind Separation of Acoustic Signals
Scott C. Douglas 355
L6.l Introduction 355
16.1.1 The Cocktail Party Effect 355
16.1.2 Chapter Overview 356
16.2 Blind Signal Separation of Convolutive Mixtures 357
16.2.1 Problem Structure 357
16.2.2 Goal of Convolutive BSS 359
16.2.3 Relationship to Other Problems 360
16.3 Criteria for Blind Signal Separation 362
16.3.1 Overview of BSS Criteria 362
16.3.2 Density Modeling Criteria 362
16.3.3 Contrast Functions 364
16.3.4 Correlation-Based Criteria 366
16.4 Structures and Algorithms for Blind Signal Separation 367
16.4.1 Filter Structures 367
16.4.2 Density Matching BSS Using Natural Gradient Adaptation 368
16.4.3 Contrast-Based BSS Under Prewhitening Constraints 370
16.4.4 Temporal Decorrelation BSS for Nonstationary Sources 372
16.5 Numerical Evaluations 373
16.6 Conclusions and Open Issues 375
References 378
Part IV. Open Problems and Future Directions
17 Future Directions for Microphone Arrays
Gary W. Elko 383
17.1 Introduction 383
17.2 Hands-Free Communication 383
17.3 The "Future" of Microphone Array Processing 385


17.4 Conclusions 387
1 8 Future Directions in Microphone Array Processing
Dirk Van Cornpernolle 389
18.1 Lessons From the Past 389
18.2 A Future Focused on Applications 391
18.2.1 Automotive 391
18.2.2 Desktop 392
18.2.3 Hearing Aids 393
18.2.4 Teleconferencing 393
18.2.5 Very Large Arrays 393
18.2.6 The Signal Subspace Approach - An Alternative to Spatial Filtering? 393
18.3 Final Remarks 394
Index 395
END
