Codebook Design Method Based on Genetic Algorithm for Chinese Continuous Digital Speech Recognition

Vector quantization is one of model training and pattern matching technologies with good performance which is applied to speech recognition at present. The traditional Linde-Buzo-Gray (LBG) algorithm is easy to get the local optimal result in the process of codebook design. According to capability of getting global optimal, genetic algorithm (GA) is used to improve the quality of codebook. The new algorithm is applied to mandarin continuous digit speech recognition and implementation procedure is given in detail. The experiments show it is more effective than traditional LBG algorithm, especially in condition of lower SNR.