Spectral coding of speech based on generalized sorted codebook vector quantization

Abstract

Sorted codebook vector quantization (SCVQ) is shown to be a very efficient vector quantization method. Generalization of SCVQ is suggested and its application to the spectral coding of speech using the quantization of line spectral frequencies (LSF), which are the most popular parameters to represent the linear prediction model for spectrum quantization in speech coders, is described. Computer simulations are conducted to evaluate the performance of the new method. We demonstrate that the new method achieves superior quality and has low implementation costs.