Genetic Vector Quantization of Line Spectrum Frequencies in Speech Coding

In this paper, we study the codebook generation of Line Spectrum Frequencies (LSF) vector quantization and we consider vector quantization of LSF in code excited linear predictive (CELP) speech coder. A partition-based genetic algorithm (GA) is applied to codebook design, which is referred to as genetic vector quantization (GVQ) method in vector quantization. The performance of ACELP using the codebook obtained by the proposed algorithm is compared with that of the conjugate structured CS-ACELP-based ITU-T G.729 coder.