摘要
针对在LBG算法中存在初始码书的选择极易影响码书训练的收敛速度和最终码书性能的缺陷,提出了一种基于微粒群的矢量量化码书设计算法.首先产生具有一定全局性特点的初始码书,然后再应用LBG算法进行优化得到同时具有局部特性的码书.实验结果验证了该算法的合理性.
A codebook design algorithm for vector quantization based on particle swarm optimization （PSO） is proposed. To solve the shortcoming that the initial codebook selection affects the convergence speed of codebook training and the performance of the final codebook, the initial codebook with global characteristics is produced using PSO, and iteratively produce final codebook with local characteristics using Linde-Buzo-Gray（LBG）. Both of the theory and practical effect show the rationality of the proposed algorithm.
出处
《微电子学与计算机》
CSCD
北大核心
2008年第6期1-3,共3页
Microelectronics & Computer
基金
国家自然科学基金项目(60572074)
关键词
微粒群算法
LBG算法
码书设计
particle swarm optimization
LBG algorithm
codebook design
