Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


Adaptive Spectral Weighting for Vector Predictive Coding of the LPC-Spectra

D. J. Zarkadis

Department of Electronic and Electrical Engineering, University of Surrey, Guildford, Surrey, UK

We propose Adaptive Spectral Weighting (ASW) for improving the low-rate coding of the LPC spectral information which is represented in the form of spectral envelope vectors. This method employs intraframe weighted-error criteria applied to split-vector quantization of the residual LPC spectral components which in effect minimize a weighted distance between the input and the reconstructed LPC spectra. Spectral weighting is introduced by means of a weighting operator corresponding to an equalized version of the spectral envelope vector which is obtained by hierarchical processing. The weighting operator is adaptive and it is characterized by gain and shape. The ASW scheme utilizes residual codebooks which have been populated by embedding weighted-distance criteria to the clustering and centroid estimation rules of the design algorithm.

Full Paper

Bibliographic reference.  Zarkadis, D. J. (1991): "Adaptive spectral weighting for vector predictive coding of the LPC-spectra", In EUROSPEECH-1991, 901-904.