Discrete-Time Speech Signal Processing: Principles and Practice (Prentice Hall Signal Processing Series)

Essential principles, practical examples, current applications, and leading-edge research. In this book, Thomas F. Quatieri presents the field's most intensive, up-to-date tutorial and reference on discrete-time speech signal processing. Building on his MIT graduate course, he introduces key principles, essential applications, and state-of-the-art research, and he identifies limitations that point the way to new research opportunities. Quatieri provides an excellent balance of theory and application, beginning with a complete framework for understanding discrete-time speech signal processing. Along the way, he presents important advances never before covered in a speech signal processing text book, including sinusoidal speech processing, advanced time-frequency analysis, and nonlinear aeroacoustic speech production modeling. Coverage includes: Speech production and speech perception: a dual view Crucial distinctions between stochastic and deterministic problems Pole-zero speech models Homomorphic signal processing Short-time Fourier transform analysis/synthesis Filter-bank and wavelet analysis/synthesis Nonlinear measurement and modeling techniques The book's in-depth applications coverage includes speech coding, enhancement, and modification; speaker recognition; noise reduction; signal restoration; dynamic range compression, and more. Principles of Discrete-Time Speech Processing also contains an exceptionally complete series of examples and Matlab exercises, all carefully integrated into the book's coverage of theory and applications. Contents......Page 8 Foreword......Page 16 Preface......Page 18 1.1 Discrete-Time Speech Signal Processing......Page 22 1.2 The Speech Communication Pathway......Page 23 1.3 Analysis/Synthesis Based on Speech Production and Perception......Page 24 1.4 Applications......Page 26 1.5 Outline of Book......Page 28 Bibliography......Page 30 2.2 Discrete-Time Signals......Page 32 2.3 Discrete-Time Systems......Page 35 2.4 Discrete-Time Fourier Transform......Page 36 2.5 Uncertainty Principle......Page 41 2.6 z-Transform......Page 44 2.7 LTI Systems in the Frequency Domain......Page 49 2.8.1 Difference Equation Realization......Page 54 2.8.2 Magnitude-Phase Relationships......Page 55 2.8.4 IIR Filters......Page 58 2.9 Time-Varying Systems......Page 59 2.10 Discrete Fourier Transform......Page 62 2.11.1 Sampling Theorem......Page 64 2.11.2 Sampling a System Response......Page 66 2.12 Summary......Page 68 Exercises......Page 69 Bibliography......Page 75 3.1 Introduction......Page 76 3.2.1 Lungs......Page 78 3.2.2 Larynx......Page 79 3.2.3 Vocal Tract......Page 87 3.2.4 Categorization of Sound by Source......Page 92 3.3 Spectrographic Analysis of Speech......Page 93 3.4 Categorization of Speech Sounds......Page 98 3.4.1 Elements of a Language......Page 100 3.4.2 Vowels......Page 102 3.4.3 Nasals......Page 105 3.4.4 Fricatives......Page 106 3.4.5 Plosives......Page 109 3.4.6 Transitional Speech Sounds......Page 113 3.5 Prosody: The Melody of Speech......Page 116 3.6.1 Acoustic Cues......Page 120 3.6.2 Models of Speech Perception......Page 121 3.7 Summary......Page 122 Exercises......Page 123 Bibliography......Page 129 4.1 Introduction......Page 132 4.2.1 Basics......Page 133 4.2.2 The Wave Equation......Page 136 4.3.1 Lossless Case......Page 140 4.3.2 Effect of Energy Loss......Page 148 4.3.3 Boundary Effects......Page 151 4.3.4 A Complete Model......Page 155 4.4 A Discrete-Time Model Based on Tube Concatenation......Page 157 4.4.1 Sound Propagation in the Concatenated Tube Model......Page 158 4.4.2 A Discrete-Time Realization......Page 164 4.4.3 Complete Discrete-Time Model......Page 169 4.5 Vocal Fold/Vocal Tract Interaction......Page 174 4.5.1 A Model for Source/Tract Interaction......Page 175 4.5.2 Formant Frequency and Bandwidth Modulation......Page 179 Exercises......Page 183 Bibliography......Page 194 5.1 Introduction......Page 196 5.2 Time-Dependent Processing......Page 197 5.3.1 Formulation......Page 198 5.3.2 Error Minimization......Page 202 5.3.3 Autocorrelation Method......Page 206 5.3.4 The Levinson Recursion and Its Associated Properties......Page 215 5.3.5 Lattice Filter Formulation of the Inverse Filter......Page 221 5.3.6 Frequency-Domain Interpretation......Page 226 5.4.1 Formulation......Page 228 5.4.2 Error Minimization......Page 230 5.5.1 Time Domain......Page 231 5.5.2 Frequency Domain......Page 233 5.6 Synthesis Based on All-Pole Modeling......Page 237 5.7 Pole-Zero Estimation......Page 241 5.7.1 Linearization......Page 242 5.7.2 Application to Speech......Page 243 5.7.3 High-Pitched Speakers: Using Two Analysis Windows......Page 248 5.8.1 Model......Page 249 5.8.2 Estimation......Page 251 5.9 Summary......Page 253 Random Processes......Page 254 Ensemble Averages......Page 256 Time Averages......Page 257 Power Density Spectrum......Page 258 Appendix 5.B: Derivation of the Lattice Filter in Linear Prediction Analysis......Page 259 Exercises......Page 261 Bibliography......Page 272 6.1 Introduction......Page 274 6.2 Concept......Page 275 6.3 Homomorphic Systems for Convolution......Page 278 6.4.1 Sequences with Rational z-Transforms......Page 282 6.4.2 Impulse Trains Convolved with Rational z-Transform Sequences......Page 286 6.4.3 Homomorphic Filtering......Page 287 6.4.4 Discrete Complex Cepstrum......Page 290 6.5 Spectral Root Homomorphic Filtering......Page 293 6.6 Short-Time Homomorphic Analysis of Periodic Sequences......Page 297 6.6.1 Quefrency-Domain Perspective......Page 298 6.6.2 Frequency-Domain Perspective......Page 300 6.7.1 Complex Cepstrum of Voiced Speech......Page 302 6.7.2 Complex Cepstrum of Unvoiced Speech......Page 307 6.8 Analysis/Synthesis Structures......Page 308 6.8.1 Zero- and Minimum-Phase Synthesis......Page 309 6.8.2 Mixed-Phase Synthesis......Page 311 6.8.3 Spectral Root Deconvolution......Page 313 6.9.2 Homomorphic Prediction......Page 314 6.10 Summary......Page 317 Exercises......Page 318 Bibliography......Page 327 7.1 Introduction......Page 330 7.2.1 Fourier Transform View......Page 331 7.2.2 Filtering View......Page 334 7.2.3 Time-Frequency Resolution Tradeoffs......Page 339 7.3.1 Formulation......Page 341 7.3.2 Filter Bank Summation (FBS) Method......Page 342 7.3.3 Overlap-Add (OLA) Method......Page 346 7.3.4 Time-Frequency Sampling......Page 349 7.4 Short-Time Fourier Transform Magnitude......Page 351 7.4.1 Signal Representation......Page 352 7.4.2 Reconstruction from Time-Frequency Samples......Page 355 7.5 Signal Estimation from the Modified STFT or STFTM......Page 356 7.5.1 Heuristic Application of STFT Synthesis Methods......Page 358 7.5.2 Least-Squared-Error Signal Estimation from the Modified STFT......Page 361 7.5.3 LSE Signal Estimation from Modified STFTM......Page 363 7.6.1 Time-Scale Modification......Page 364 7.6.2 Noise Reduction......Page 370 7.7 Summary......Page 371 Appendix 7.A: FBS Method with Multiplicative Modification......Page 372 Exercises......Page 373 Bibliography......Page 382 8.1 Introduction......Page 384 8.2 Revisiting the FBS Method......Page 385 8.3.1 Analysis/Synthesis of Quasi-Periodic Signals......Page 388 8.3.2 Applications......Page 396 8.3.3 Motivation for a Sinewave Analysis/Synthesis......Page 401 8.4.1 Preservation of Temporal Envelope......Page 402 8.4.2 Phase Coherence of Quasi-Periodic Signals......Page 406 8.5 Constant-Q Analysis/Synthesis......Page 407 8.5.1 Motivation......Page 408 8.5.2 Wavelet Transform......Page 409 8.5.3 Discrete Wavelet Transform......Page 413 8.5.4 Applications......Page 418 8.6 Auditory Modeling......Page 422 8.6.1 AM-FM Model of Auditory Processing......Page 424 8.6.2 Auditory Spectral Model......Page 427 8.6.3 Phasic/Tonic View of Auditory Neural Processing......Page 429 Exercises......Page 433 Bibliography......Page 443 9.1 Introduction......Page 448 9.2 Sinusoidal Speech Model......Page 450 9.3 Estimation of Sinewave Parameters......Page 453 9.3.1 Voiced Speech......Page 456 9.3.2 Unvoiced Speech......Page 460 9.3.3 Analysis System......Page 461 9.3.4 Frame-to-Frame Peak Matching......Page 463 9.4 Synthesis......Page 466 9.4.1 Cubic Phase Interpolation......Page 467 9.4.2 Overlap-Add Interpolation......Page 471 9.4.3 Examples......Page 473 9.4.4 Applications......Page 477 9.4.5 Time-Frequency Resolution......Page 478 9.5.1 Signal Model......Page 481 9.5.2 Applications......Page 482 9.6.1 Signal Model......Page 495 9.6.2 Analysis/Synthesis......Page 496 9.6.3 Application to Signal Modification......Page 498 9.7 Summary......Page 499 Appendix 9.A: Derivation of the Sinewave Model......Page 500 Appendix 9.B: Derivation of Optimal Cubic Phase Parameters......Page 503 Exercises......Page 505 Bibliography......Page 520 10.1 Introduction......Page 524 10.2 A Correlation-Based Pitch Estimator......Page 525 10.3 Pitch Estimation Based on a "Comb Filter"......Page 526 10.4 Pitch Estimation Based on a Harmonic Sinewave Model......Page 530 10.4.1 Parameter Estimation for the Harmonic Sinewave Model......Page 531 10.4.2 Parameter Estimation for the Harmonic Sinewave Model with a priori Amplitude......Page 532 10.4.3 Voicing Detection......Page 537 10.4.4 Time-Frequency Resolution Perspective......Page 540 10.4.5 Evaluation by Harmonic Sinewave Reconstruction......Page 543 10.5.1 A Phase Model Based on Onset Time......Page 544 10.5.2 Onset Estimation......Page 546 10.5.3 Sinewave Amplitude Envelope Estimation......Page 548 10.5.4 Minimum-Phase Sinewave Reconstruction......Page 551 10.6.1 Harmonic Sinewave Model......Page 552 10.6.2 Multi-Band Voicing......Page 554 10.7 Summary......Page 555 Exercises......Page 556 Bibliography......Page 561 11.1 Introduction......Page 562 11.2 The STFT and Wavelet Transform Revisited......Page 563 11.2.2 Minimum Uncertainty......Page 564 11.2.3 Tracking Instantaneous Frequency......Page 567 11.3.1 Properties of a Proper Time-Frequency Distribution......Page 570 11.3.2 Spectrogram as a Time-Frequency Distribution......Page 573 11.3.3 Wigner Distribution......Page 574 11.3.5 Application to Speech Analysis......Page 579 11.4 Aeroacoustic Flow in the Vocal Tract......Page 583 11.4.1 Preliminaries......Page 584 11.4.2 Early Measurements and Hypotheses of Aeroacoustic Flow in the Vocal Tract......Page 585 11.4.3 Aeroacoustic Mechanical Model......Page 588 11.4.4 Aeroacoustic Computational Model......Page 591 11.5.1 Motivation......Page 592 11.5.2 Energy Measurement......Page 593 11.5.3 Energy Separation......Page 598 11.6 Summary......Page 603 Exercises......Page 604 Bibliography......Page 613 12.1 Introduction......Page 616 12.3 Scalar Quantization......Page 619 12.3.1 Fundamentals......Page 620 12.3.2 Quantization Noise......Page 623 12.3.3 Derivation of the Max Quantizer......Page 627 12.3.4 Companding......Page 630 12.3.5 Adaptive Quantization......Page 631 12.3.6 Differential and Residual Quantization......Page 634 12.4.1 Approach......Page 637 12.4.2 VQ Distortion Measure......Page 639 12.4.3 Use of VQ in Speech Transmission......Page 641 12.5.1 Subband Coding......Page 642 12.5.2 Sinusoidal Coding......Page 646 12.6.1 Basic Linear Prediction Coder (LPC)......Page 656 12.6.2 A VQ LPC Coder......Page 658 12.6.3 Mixed Excitation LPC (MELP)......Page 659 12.7 LPC Residual Coding......Page 661 12.7.1 Multi-Pulse Linear Prediction......Page 662 12.7.2 Multi-Pulse Modeling with Long-Term Prediction......Page 666 12.7.3 Code-Excited Linear Prediction (CELP)......Page 670 12.8 Summary......Page 673 Exercises......Page 674 Bibliography......Page 681 13.1 Introduction......Page 686 13.2.1 Problem Formulation......Page 687 13.2.2 Spectral Subtraction......Page 689 13.2.3 Cepstral Mean Subtraction......Page 692 13.3 Wiener Filtering......Page 693 13.3.1 Basic Approaches to Estimating the Object Spectrum......Page 694 13.3.2 Adaptive Smoothing Based on Spectral Change......Page 696 13.3.3 Application to Speech......Page 699 13.3.4 Optimal Spectral Magnitude Estimation......Page 701 13.4 Model-Based Processing......Page 703 13.5 Enhancement Based on Auditory Masking......Page 705 13.5.1 Frequency-Domain Masking Principles......Page 706 13.5.2 Calculation of the Masking Threshold......Page 708 13.5.3 Exploiting Frequency Masking in Noise Reduction......Page 709 13.6.1 Formulation......Page 711 13.6.2 Temporal Filtering......Page 712 13.6.3 Nonlinear Transformations of Time-Trajectories......Page 715 13.7 Summary......Page 719 Appendix 13.A: Stochastic-Theoretic Parameter Estimation......Page 720 Exercises......Page 721 Bibliography......Page 726 14.1 Introduction......Page 730 14.2.1 Formulation......Page 732 14.2.2 Mel-Cepstrum......Page 733 14.2.3 Sub-Cepstrum......Page 736 14.3.1 Minimum-Distance Classifier......Page 738 14.3.2 Vector Quantization......Page 739 14.3.3 Gaussian Mixture Model (GMM)......Page 740 14.4.1 Glottal Flow Derivative......Page 746 14.4.3 Relative Influence of Source, Spectrum, and Prosody......Page 750 14.5 Signal Enhancement for the Mismatched Condition......Page 754 14.5.1 Linear Channel Distortion......Page 755 14.5.2 Nonlinear Channel Distortion......Page 758 14.5.3 Other Approaches......Page 767 14.6.1 Synthesized Coded Speech......Page 769 14.6.2 Experiments with Coder Parameters......Page 770 14.7 Summary......Page 772 Appendix 14.A: Expectation-Maximization (EM) Estimation......Page 773 Exercises......Page 775 Bibliography......Page 783 Speech Signal Processing......Page 788 Databases......Page 789 A......Page 790 D......Page 791 G......Page 792 L......Page 793 O......Page 794 P......Page 795 S......Page 796 T......Page 798 V......Page 799 Z......Page 800 About the Author......Page 802