矢量量化的等误差竞争学习算法  

Vector quantization based on competitive learning algorithm with equidistortion

作　　者：陈善学[1] 

机构地区：[1]重庆邮电学院,重庆400065

出　　处：《重庆邮电学院学报（自然科学版）》2004年第1期63-65,共3页Journal of Chongqing University of Posts and Telecommunications(Natural Sciences Edition)

基　　金：重庆邮电学院青年教师创新基金资助。

摘　　要：提出了一种使各区域子误差相等的矢量量化算法,该算法利用小波变换后各子带间的相关性,合理构造矢量,采用最优矢量量化器设计原则,通过调整学习过程中各子区域的误差,使之趋于相等,改善总的期望误差,获得更接近全局最优的码书。实验表明,这种算法获得的码本优于其它几种算法。A vector quantization algorithm with equidistortion on each region is introduced. The correlation of the wavelet coefficients in different frequency bands is effectively exploited through constructing the band-cross vector. Using the designing principle of the best vector quantizer, a nearly optimal codebook can be obtained by modulating the subdistortion of each region and can be made to equal to each other. Experimental results show that the algorithm is superior to other methods.

关 键 词：小波变换 矢量量化 等误差原则 

分 类 号：TN911.21[电子电信—通信与信息系统]