采用进化算法优化向量量化

Optimal Quantizer Design with Evolutionary Algorithm

把进化策略引入到矢量量化设计中 ,提出了一种选择优良后代码书矢量的判据和方法 .在使用 LBG算法进行聚类的前提下 ,利用所选择出来的后代矢量调整各个区域的子误差 ,从而进一步改善期望误差 ,以获得全局最优码书 .实验结果证明 ,此方法能克服局部最优 ,改善码书矢量 .

This article introduced evolutionary strategy into the vector quantizer design and proposed a criterion and method of selecting the survived offspring code-vectors. In the precondition of clustering with LBG, subdistortion of each subregion is adjusted by using selected code-vectors to improve the expected distortion so as to attain the globally optimal codebook. Experimental results show that the algorithm can improve the code-vectors without falling into local optimality.