A New 12.8 kbps LD-CELP Speech Codec Based on a Neural Search Codebook Module

Low Delay-Code Excited Linear Prediction (LD-CELP) is considered as an attractive algorithm in speech coding domain. This algorithm was adopted by the International Telephone and Telegraph Consultative Committee (CCITT) for the coding of speech at 16 kbps with toll quality. However, operation at transmission rates lower than 16 kbps is desirable so that traffic can be accommodated during system overload conditions. In this paper, a 12.8 kbps LD-CELP algorithm is proposed in which codebook search module employs a network of Self-Organizing Maps (SOMs) to determine the optimum index value of shape codebook. Based on the occurrence frequency characteristics of codevectors, the 6 bits for shape codebook and 2 bits for gain codebook are used in this work to produce a 12.8 kbps coder. The performance comparison of the proposed neural codebook search module in the structure of 12.8 kbps LD-CELP with a conventional implementation of LD-CELP coder shows that execution time of the algorithm is reduced significantly. However, the degradation of voice quality in terms of Perceived Evaluation of Speech Quality (PESQ) and segmental Signal to Noise Ratio (SNRseg) is not significant.