Algorithms and Software for Predictive and Perceptual Modeling of Speech

This book delves into advances in speech coding, focusing on predictive and perceptual modeling techniques to enhance speech quality at low bit rates.

The field of speech coding has evolved significantly, moving from early pulse code modulation-based coders to modern multi-rate wideband speech coding standards. The primary goal remains achieving high-quality speech while minimizing the bit rate. In Algorithms and Software for Predictive and Perceptual Modeling of Speech, the author explores recent advancements in linear prediction (LP)-based speech analysis, particularly those that utilize perceptual models for both narrow- and wide-band speech coding. The LP analysis-synthesis framework has proven effective for speech coding due to its alignment with the source-system paradigm for speech synthesis.

Throughout the book, various limitations of conventional LP are examined, with numerous extensions proposed to enhance LP-based analysis-synthesis. These include discrete all-pole modeling, perceptual LP, warped LP, and more. Such algorithms aim to either improve the fitting performance of the LP spectral envelope or incorporate perceptual models into the LP framework. The initial sections delve into recent developments in predictive modeling, supported by practical Matlab™ simulation examples, emphasizing the benefits of integrating perceptual models into low bit rate speech coding.

The latter half of the book shifts focus to perceptual modeling of speech, discussing concepts such as masking thresholds and psychoacoustic models, again supplemented by Matlab™ simulations. Readers can access supplementary materials, including Matlab™ programs and simulation examples. This comprehensive overview makes Algorithms and Software for Predictive and Perceptual Modeling of Speech a valuable resource for those interested in the advancements in speech coding technologies.