Source Coding and Speech Compression

Abstract

An overview of recent applications of source coding theory and techniques to Linear Predictive Coded Coded speech compression systems is presented. Several distortion measures proposed for use in speech compression systems are described and compared. These distortion measures are then combined with an algorithm for computing "optimum" (minimum distortion) vector quantizers to obtain optimum quantizers for reflection coefficient vectors in Linear Predictive Coded speech systems. The quality of the system is evaluated via the speech distortion measures and listening to demonstration tapes. Some implications for speech compression theory and practice are discussed.