Spectral entropy-based wideband speech coding

Wideband speech coding is concerned with speech in the 50 Hz to 7 kHz band and is important in video-conferencing applications. For a new standard at 24 and 32 kbits/s, G.722.1, standardization efforts are underway to achieve good quality wideband speech at 14-16 kbits/s. We present a new approach to wideband speech compression based upon spectral entropy. This approach is sample function adaptive and falls within the class of nonlinear approximation methods in that it codes the best n basis functions rather than allocate bits to the first n. We develop spectral entropy-based wideband speech coders that operate at both 24 and 16 kbits/s.

[1]  Seymour Shlien,et al.  The modulated lapped transform, its time-varying forms, and its applications to audio coding standards , 1997, IEEE Trans. Speech Audio Process..

[2]  Wenye Yang,et al.  The coefficient rate and equivalent bandwidth in source coding , 1998, 1998 Information Theory Workshop (Cat. No.98EX131).

[3]  Thomas M. Cover,et al.  Elements of Information Theory , 2005 .

[4]  L. Lorne Campbell,et al.  Minimum Coefficient Rate for Stationary Random Processes , 1960, Inf. Control..

[5]  Jerry D. Gibson,et al.  Spectral entropy, equivalent bandwidth and minimum coefficient rate , 1997, Proceedings of IEEE International Symposium on Information Theory.

[6]  W. H. Holmes,et al.  Use of an auditory model to improve speech coders , 1993, 1993 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[7]  J. Gibson,et al.  Coefficient rate and adaptive coding of side information , 1998 .