利用离散Hopfield网络的码本设计方法

摘要

论文提出了一种利用Hopfield网络的码本设计方法,分析了LBG算法和离散Hopfield网络的特点,针对该特点构造聚类表格,并按离散Hopfield神经网络串行方式运行,从而得到最终码字集.通过实验表明,在码本大小相同的情况下,峰值信噪比提高了2.742～3.825 dB,生成的码本质量较传统的LBG算法更加有效.%This paper proposes codebook design method of using discrete hopfield neural networks.Based on analysis of the LBG algorithm and features of discrete hopfield neural networks,this method builts a clustering form;the clustering form works with the asynchronous mode of discrete hopfield neural network,and abtains the final codebook.Experimental results indicate that PSNR can be increased 2.742 to 3.825 dB.This method is more effective than the traditional LBG algorithm.