本文提出了一种采用均方误差 (MSE)测度的矢量量化码书搜索的快速算法 .该算法在码书设计的每次迭代前预先计算各码字的和值 (一个矢量各分量的和 )并保存在码书中 .在迭代过程中 ,利用输入矢量的和值、各码字的和值以及均方误差三者之间的各种特性排除大部分候选码字而免去许多均方误差计算 .测试结果表明 ,相对于穷尽搜索方法 ,计算量得到明显的降低 ,计算时间减少约 90 % ,同时只需要很少的预先计算量和额外存储量 . In this paper, we propose a fast algorithm for vector quantization codebook search using mean square error (MSE) measure, which pre-calculates the sum of each codeword (the sum of the components of a vector) before each iteration of the codebook design And saved in the codebook.In the iterative process, most of the candidate codewords are eliminated by using the sum of the input vectors, the sums of the codewords and the mean square error, without many mean square errors The results of the tests show that the computational cost is significantly reduced relative to the exhaustive search method, the computational time is reduced by about 90%, and only a small amount of precalculation and additional storage is required.