Adaptive L/sub p/ approach to speech coding

Some experimental results using L/sub p/ (1<or=p<or=3) normed models in speech coding are described. In particular, the residual steepest descent (RSD) algorithm has been used to generate L/sub p/ models of speech for a variety of values of p; these models have in turn been used to generate statistics of the residual sequence. The RSD algorithm has been incorporated into the government standard LPC-10 coding algorithm, where the norm p used is selected on the basis of a statistical measure of the residual.<<ETX>>

[1]  J. Bee Bednar,et al.  L1 deconvolution and its application to seismic signal processing , 1986, IEEE Trans. Acoust. Speech Signal Process..

[2]  M. Paez,et al.  Minimum Mean-Squared-Error Quantization in Speech PCM and DPCM Systems , 1972, IEEE Trans. Commun..

[3]  R. Geary,et al.  Testing for Normality , 2003 .

[4]  Rui J. P. de Figueiredo,et al.  Maximum likelihood estimation of a class of non-Gaussian densities with application to deconvolution , 1987, ICASSP '87. IEEE International Conference on Acoustics, Speech, and Signal Processing.

[5]  Robert V. Hogg.  An introduction to robust procedures , 1977 .

[6]  J. Rice,et al.  Norms for Smoothing and Estimation , 1964 .

[7]  Etienne Denoel,et al.  Linear prediction of speech with a least absolute error criterion , 1985, IEEE Trans. Acoust. Speech Signal Process..

[8]  J. Bee Bednar,et al.  Fast algorithms for lpdeconvolution , 1985, IEEE Trans. Acoust. Speech Signal Process..