Vector Quantization Codebook Design Method for Speech Recognition Based on Genetic Algorithm

In the process of codebook design of vector quantization, traditional LBG algorithm owns the advantage of fast convergence, but it is prone to local optimum and is influenced greatly by initial codebook. Given that the Genetic Algorithm has the capability to produce global optimal results,this paper proposes a new clustering algorithm GA-L based on GA and LBG to improve the quality of codebook. This paper applies GA-L algorithm to Mandarin Continuous Digit Speech Recognition,the experiments results show GA-L algorithm is more effective than traditional LBG algorithm.