| p. 1 |
| Background - Linear predictive processing of speech | p. 1 |
| The LP model of speech | p. 2 |
| The LP estimation algorithm | p. 4 |
| LP processing in practice | p. 5 |
| Linear predictive coders | p. 7 |
| MATLAB proof of concept: ASP_cell_phone.m | p. 11 |
| Examining a speech file | p. 12 |
| Linear prediction synthesis of 30 ms of voiced speech | p. 13 |
| Linear prediction synthesis of 30 ms of unvoiced speech | p. 18 |
| Linear prediction synthesis of a speech file, with fixed0 | p. 19 |
| Unvoiced linear prediction synthesis of a speech file | p. 21 |
| Linear prediction synthesis of speech, with original0 | p. 21 |
| CELP analysis-synthesis of a speech file | p. 23 |
| Going further | p. 29 |
| Conclusion | p. 29 |
| References | p. 30 |
| p. 33 |
| Background - Delta-sigma modulation | p. 34 |
| Uniform quantization: Bits vs. SNR | p. 34 |
| Conventional DACs | p. 36 |
| Oversampling DACs | p. 36 |
| Oversampling DACs - Noise shaping | p. 40 |
| Delta-sigma DACs | p. 42 |
| MATLAB proof of concept: ASP_audio_cd.m | p. 45 |
| Uniform quantization | p. 45 |
| Dithering | p. 47 |
| Conventional DAC | p. 49 |
| Oversampling DAC | p. 53 |
| Oversampling and noise-shaping DAC | p. 57 |
| Delta-sigma DAC | p. 59 |
| Going further | p. 62 |
| Conclusion | p. 62 |
| References | p. 63 |
| p. 65 |
| Background - Sub-band and transform coding | p. 65 |
| Perfect reconstruction filters | p. 67 |
| Filter banks and lapped transforms | p. 73 |
| Using the masking properties of the human ear | p. 76 |
| Audio coders | p. 77 |
| MATLAB proof of concept: ASP_mp3.m | p. 80 |
| Two-channel filter bank | p. 81 |
| Two-channel QMF filter bank | p. 84 |
| 32-channel pseudo-QMF filter bank | p. 86 |
| Filter banks and lapped transforms | p. 89 |
| Perceptual audio coding | p. 92 |
| Going further | p. 100 |
| Conclusion | p. 100 |
| References | p. 101 |
| p. 103 |
| Background - Statistical pattern recognition | p. 104 |
| The statistical formalism of ASR | p. 105 |
| Markov models | p. 108 |
| Hidden Markov models | p. 111 |
| Training HMMs | p. 115 |
| MATLAB proof of concept: ASP_dictation_machine.m | p. 118 |
| Gaussian modeling and Bayesian classification of vowels | p. 118 |
| Gaussian Mixture Models (GMM) | p. 123 |
| Hidden Markov models (HMM) | p. 134 |
| N-grams | p. 139 |
| Word-based continuous speech recognition | p. 144 |
| Going further | p. 147 |
| Conclusion | p. 147 |
| References | p. 147 |
| p. 149 |
| Background - The phase vocoder | p. 149 |
| DFT-based signal processing | p. 150 |
| STFT-based signal processing | p. 152 |
| Perfect reconstruction | p. 156 |
| Time scale modification with the phase vocoder | p. 157 |
| Pitch shifting with the phase vocoder | p. 163 |
| MATLAB proof of concept: ASP_audio_effects.m | p. 166 |
| STFT-based audio signal processing | p. 166 |
| Time-scale modification | p. 172 |
| Pitch modification | p. 179 |
| Going further | p. 182 |
| Conclusion | p. 183 |
| References | p. 184 |
| p. 187 |
| Background - Source localization | p. 188 |
| Sperm whale sounds | p. 188 |
| The Teager-Kaiser energy operator | p. 190 |
| TDOA estimation based on the generalized cross-correlation | p. 192 |
| Adaptive TDOA estimation | p. 195 |
| Multilateration | p. 198 |
| MATLAB proof of concept: ASP_audio_effects.m | p. 199 |
| Sperm whale sounds | p. 199 |
| Teager-Kaiser filtering | p. 203 |
| TDOA estimation using generalized cross-correlation | p. 210 |
| TDOA estimation using least-mean squares | p. 215 |
| Multilateration | p. 218 |
| Going further | p. 220 |
| Conclusion | p. 220 |
| References | p. 221 |
| p. 223 |
| Background - Audio watermarking seen as a digital communication problem | p. 225 |
| Spread spectrum signals | p. 226 |
| Communication channel design | p. 228 |
| Informed watermarking | p. 233 |
| MATLAB proof of concept: ASP_watermarking.m | p. 238 |
| Audio watermarking seen as a digital communication problem | p. 239 |
| Informed watermarking with error-free detection | p. 244 |
| Informed watermarking made inaudible | p. 247 |
| Informed watermarking robust to MPEG compression | p. 259 |
| Going further | p. 261 |
| Conclusion | p. 262 |
| References | p. 262 |
| p. 265 |
| Background-JPEG | p. 266 |
| Color transform | p. 268 |
| Frequency transform: The discrete cosine transform | p. 269 |
| Entropy coding | p. 279 |
| A few specificities of the JPEG standard | p. 282 |
| Quality measures | p. 284 |
| MATLAB proof of concept | p. 285 |
| Block image transformation | p. 286 |
| Complete image block coding | p. 293 |
| DCT quantization | p. 295 |
| Spatial decorrelation between blocks | p. 298 |
| Entropy coding | p. 302 |
| Still image coding | p. 306 |
| Going further | p. 308 |
| Conclusions | p. 309 |
| References | p. 309 |
| p. 311 |
| Background - Motion estimation | p. 312 |
| Motion estimation: The block matching algorithm | p. 316 |
| A few specificities of video coding standards | p. 321 |
| MATLAB proof of concept | p. 325 |
| Macroblock processing | p. 325 |
| Block matching motion estimation | p. 326 |
| Motion compensation | p. 339 |
| Selection of search area | p. 341 |
| Selection of reference image | p. 343 |
| Backward motion estimation | p. 345 |
| Coding of the compensation error | p. 349 |
| Entropy coding | p. 350 |
| Video coding | p. 352 |
| Going further | p. 358 |
| Conclusion | p. 359 |
| References | p. 359 |
| p. 361 |
| Background - Introduction to wavelet and multi-resolution transforms | p. 365 |
| Think globally, act locally | p. 366 |
| Approximate... but details matter | p. 367 |
| Wavelet transform: Definition and computation | p. 370 |
| WT and discrete signals: DWT | p. 374 |
| WT and DWT for Images: 1+1 = 2 | p. 375 |
| Background - Context-based modeling of wavelet coefficients bit planes | p. 376 |
| Spatial and bit-depth scalability | p. 376 |
| Efficient entropy coding | p. 377 |
| Background - Rate-distortion optimal bit allocation across wavelet codeblocks | p. 379 |
| Problem definition | p. 380 |
| Lagrangian formulation and approximated solution | p. 381 |
| Lagrangian optimization: A non-image based example | p. 384 |
| MATLAB proof of concept | p. 386 |
| Experiments with the wavelet transform | p. 387 |
| A simplified JPEG2000 scheme | p. 393 |
| Going further: From concepts to compliant JPEG2000 codestreams | p. 406 |
| Conclusion | p. 408 |
| References | p. 408 |
| p. 411 |
| Background - Statistical pattern recognition for image classification | p. 414 |
| Statistical framework | p. 415 |
| Gaussian mixture models (GMM) | p. 417 |
| The Expectation-Maximization algorithm (EM) | p. 419 |
| Markov random fields (MRF) | p. 420 |
| Hidden Markov random fields (HMRF) | p. 423 |
| Gaussian hidden Markov random field model | p. 424 |
| MATLAB proof of concept | p. 426 |
| 3D data visualization | p. 426 |
| Image histogram | p. 429 |
| Gaussian mixture model (GMM) | p. 431 |
| Hidden Gaussian mixture model | p. 437 |
| Influence of the spatial parameter | p. 439 |
| Localization and quantification of brain degeneration | p. 441 |
| Going further | p. 445 |
| Nature and domain of the transformation | p. 447 |
| Features and cost function | p. 447 |
| Optimization | p. 447 |
| Conclusions | p. 448 |
| Acknowledgments | p. 448 |
| References | p. 449 |
| Index | p. 451 |
| Table of Contents provided by Ingram. All Rights Reserved. |