自适应滤波器原理:英文版

Contents
Background and Preview 1
1. The Filtering Problem 1
2. Linear Optimum Filters 4
3. Adaptive Filters 4
4. Linear Filter Structures 6
5. Approaches to the Development of Linear Adaptive Filters 12
6. Adaptive Beamforming 13
7. Four Classes of Applications 17
8. Historical Notes 20
Chapter 1 Stochastic Processes and Models 30
1.1 Partial Characterization of a Discrete-Time Stochastic Process 30
1.2 Mean Ergodic Theorem 32
1.3 Correlation Matrix 34
1.4 Correlation Matrix of Sine Wave Plus Noise 39
1.5 Stochastic Models 40
1.6 Wold Decomposition 46
1.7 Asymptotic Stationarity of an Autoregressive Process 49
1.8 Yule?CWalker Equations 51
1.9 Computer Experiment: Autoregressive Process of Order Two 52
1.10 Selecting the Model Order 60
1.11 Complex Gaussian Processes 63
1.12 Power Spectral Density 65
1.13 Properties of Power Spectral Density 67
1.14 Transmission of a Stationary Process Through a Linear Filter 69
1.15 Cramér Spectral Representation for a Stationary Process 72
1.16 Power Spectrum Estimation 74
1.17 Other Statistical Characteristics of a Stochastic Process 77
1.18 Polyspectra 78
1.19 Spectral-Correlation Density 81
1.20 Summary and Discussion 84
Problems 85
Chapter 2 Wiener Filters 90
2.1 Linear Optimum Filtering: Statement of the Problem 90
2.2 Principle of Orthogonality 92
2.3 Minimum Mean-Square Error 96
2.4 Wiener?CHopf Equations 98
2.5 Error-Performance Surface 100
2.6 Multiple Linear Regression Model 104
2.7 Example 106
2.8 Linearly Constrained Minimum-Variance Filter 111
2.9 Generalized Sidelobe Cancellers 116
2.10 Summary and Discussion 122
Problems 124
Chapter 3 Linear Prediction 132
3.1 Forward Linear Prediction 132
3.2 Backward Linear Prediction 139
3.3 Levinson?CDurbin Algorithm 144
3.4 Properties of Prediction-Error Filters 153
3.5 Schur?CCohn Test 162
3.6 Autoregressive Modeling of a Stationary Stochastic Process 164
3.7 Cholesky Factorization 167
3.8 Lattice Predictors 170
3.9 All-Pole, All-Pass Lattice Filter 175
3.10 Joint-Process Estimation 177
3.11 Predictive Modeling of Speech 181
3.12 Summary and Discussion 188
Problems 189
Chapter 4 Method of Steepest Descent 199
4.1 Basic Idea of the Steepest-Descent Algorithm 199
4.2 The Steepest-Descent Algorithm Applied to the Wiener Filter 200
4.3 Stability of the Steepest-Descent Algorithm 204
4.4 Example 209
4.5 The Steepest-Descent Algorithm Viewed as a Deterministic Search Method 221
4.6 Virtue and Limitation of the Steepest-Descent Algorithm 222
4.7 Summary and Discussion 223
Problems 224
Chapter 5 Method of Stochastic Gradient Descent 228
5.1 Principles of Stochastic Gradient Descent 228
5.2 Application 1: Least-Mean-Square (LMS) Algorithm 230
5.3 Application 2: Gradient-Adaptive Lattice Filtering Algorithm 237
5.4 Other Applications of Stochastic Gradient Descent 244
5.5 Summary and Discussion 245
Problems 246
Chapter 6 The Least-Mean-Square (LMS) Algorithm 248
6.1 Signal-Flow Graph 248
6.2 Optimality Considerations 250
6.3 Applications 252
6.4 Statistical Learning Theory 272
6.5 Transient Behavior and Convergence Considerations 283
6.6 Efficiency 286
6.7 Computer Experiment on Adaptive Prediction 288
6.8 Computer Experiment on Adaptive Equalization 293
6.9 Computer Experiment on a Minimum-Variance Distortionless-Response
Beamformer
302
6.10 Summary and Discussion 306
Problems 308
Chapter 7 Normalized Least-Mean-Square (LMS) Algorithm and Its
Generalization 315
7.1 Normalized LMS Algorithm: The Solution to a Constrained Optimization Problem 315
7.2 Stability of the Normalized LMS Algorithm 319
7.3 Step-Size Control for Acoustic Echo Cancellation 322
7.4 Geometric Considerations Pertaining to the Convergence Process for Real-Valued
Data 327
7.5 Affine Projection Adaptive Filters 330
7.6 Summary and Discussion 334
Problems 335
Chapter 8 Block-Adaptive Filters 339
8.1 Block-Adaptive Filters: Basic Ideas 340
8.2 Fast Block LMS Algorithm 344
8.3 Unconstrained Frequency-Domain Adaptive Filters 350
8.4 Self-Orthogonalizing Adaptive Filters 351
8.5 Computer Experiment on Adaptive Equalization 361
8.6 Subband Adaptive Filters 367
8.7 Summary and Discussion 375
Problems 376
Chapter 9 Method of Least-Squares 380
9.1 Statement of the Linear Least-Squares Estimation Problem 380
9.2 Data Windowing 383
9.3 Principle of Orthogonality Revisited 384
9.4 Minimum Sum of Error Squares 387
9.5 Normal Equations and Linear Least-Squares Filters 388
9.6 Time-Average Correlation Matrix ≥ 391
9.7 Reformulation of the Normal Equations in Terms of Data Matrices 393
9.8 Properties of Least-Squares Estimates 397
9.9 Minimum-Variance Distortionless Response (MVDR) Spectrum Estimation 401
9.10 Regularized MVDR Beamforming 404
9.11 Singular-Value Decomposition 409
9.12 Pseudoinverse 416
9.13 Interpretation of Singular Values and Singular Vectors 418
9.14 Minimum-Norm Solution to the Linear Least-Squares Problem 419
9.15 Normalized LMS Algorithm Viewed as the Minimum-Norm Solution to an
Underdetermined Least-Squares Estimation Problem 422
9.16 Summary and Discussion 424
Problems 425
Chapter 10 The Recursive Least-Squares (RLS) Algorithm 431
10.1 Some Preliminaries 431
10.2 The Matrix Inversion Lemma 435
10.3 The Exponentially Weighted RLS Algorithm 436
10.4 Selection of the Regularization Parameter 439
10.5 Updated Recursion for the Sum of Weighted Error Squares 441
10.6 Example: Single-Weight Adaptive Noise Canceller 443
10.7 Statistical Learning Theory 444
10.8 Efficiency 449
10.9 Computer Experiment on Adaptive Equalization 450
10.10 Summary and Discussion 453
Problems 454
Chapter 11 Robustness 456
11.1 Robustness, Adaptation, and Disturbances 456
11.2 Robustness: Preliminary Considerations Rooted in H∞ Optimization 457
11.3 Robustness of the LMS Algorithm 460
11.4 Robustness of the RLS Algorithm 465
11.5 Comparative Evaluations of the LMS and RLS Algorithms from the Perspective of
Robustness
470
11.6 Risk-Sensitive Optimality 470
11.7 Trade-Offs Between Robustness and Efficiency 472
11.8 Summary and Discussion 474
Problems 474
Chapter 12 Finite-Precision Effects 479
12.1 Quantization Errors 480
12.2 Least-Mean-Square (LMS) Algorithm 482
12.3 Recursive Least-Squares (RLS) Algorithm 491
12.4 Summary and Discussion 497
Problems 498
Chapter 13 Adaptation in Nonstationary Environments 500
13.1 Causes and Consequences of Nonstationarity 500
13.2 The System Identification Problem 501
13.3 Degree of Nonstationarity 504
13.4 Criteria for Tracking Assessment 505
13.5 Tracking Performance of the LMS Algorithm 507
13.6 Tracking Performance of the RLS Algorithm 510
13.7 Comparison of the Tracking Performance of LMS and RLS Algorithms 514
13.8 Tuning of Adaptation Parameters 518
13.9 Incremental Delta-Bar-Delta (IDBD) Algorithm 520
13.10 Autostep Method 526
13.11 Computer Experiment: Mixture of Stationary and Nonstationary Environmental
Data 530
13.12 Summary and Discussion 534
Problems 535
Chapter 14 Kalman Filters 540
14.1 Recursive Minimum Mean-Square Estimation for Scalar Random Variables 541
14.2 Statement of the Kalman Filtering Problem 544
14.3 The Innovations Process 547
14.4 Estimation of the State Using the Innovations Process 549
14.5 Filtering 555
14.6 Initial Conditions 557
14.7 Summary of the Kalman Filter 558
14.8 Optimality Criteria for Kalman Filtering 559
14.9 Kalman Filter as the Unifying Basis for RLS Algorithms 561
14.10 Covariance Filtering Algorithm 566
14.11 Information Filtering Algorithm 568
14.12 Summary and Discussion 571
Problems 572
C