Speech coding based upon vector quantization

TLDR

With rare exception, all presently available narrow‑band speech coding systems implement scalar quantization of the transmission parameters such as reflection coefficients or transformed reflection coefficients in LPC systems. The paper introduces vector quantization as a new approach to speech coding, presenting its theory, practical results, and outlining future research directions. Vector quantization is a mathematically and computationally tractable method built on linear prediction analysis, providing a new approach to speech coding with theoretical foundations and practical results. For very low data rates, realistic experiments show that vector quantization achieves the same average distortion with 15 to 20 fewer bits per frame than optimized scalar quantization.

Abstract

With rare exception, all presently available narrow-band speech coding systems implement scalar quantization (independent quantization) of the transmission parameters (such as reflection coefficients or transformed reflection coefficients in LPC systems). This paper presents a new approach called vector quantization. For very low data rates, realistic experiments have shown that vector quantization can achieve a given level of average distortion with 15 to 20 fewer bits/frame than that required for the optimized scalar quantizing approaches presently in use. The vector quantizing approach is shown to be a mathematically and computationally tractable method which builds upon knowledge obtained in linear prediction analysis studies. This paper introduces the theory in a nonrigorous form, along with practical results to date and an extensive list of research topics for this new area of speech coding.