Representing Acoustics of Speech for Speech Processing - 2009

Explore the intricacies of acoustic speech signal representation in this comprehensive lecture by renowned speech processing expert Bishnu Atal. Delve into the crucial role of proper acoustic representation in various speech processing applications, examining the advantages and disadvantages of using short and long time windows in speech analysis. Investigate the use of short-time Fourier transform, discussing unresolved issues such as window size and shape. Compare narrow-band and wideband analysis techniques, and learn about the importance of both short and long interval prediction in digital speech coding applications. Consider Hermansky's argument for using longer speech windows in automatic speech recognition, and evaluate the potential benefits of long-window Fourier transforms. Examine the debate surrounding the relevance of phase information in speech representations, and gain insights into this important aspect of speech processing. Benefit from Atal's extensive experience and contributions to the field, including his pioneering work in linear predictive coding of speech, as you explore cutting-edge concepts in speech analysis, synthesis, and coding.