摘要
将矩阵的投影引入到矢量量化中 ,达到了在搜索最佳匹配码字的过程中首先在降维空间中将一些侯选码字排除 ,其中利用了最佳匹配码字的和待编码的矢量的均值和方差与MSE距离之间的关系不等式 .[1] 实际结果表明约 98%的侯选码字被排除 ,只有 2
In this paper,we introduce the projection of a matrix into the vector quantization.In the process of searching the best matching codeword,some candidate codewords are first discarded in the dimensionaligy reduced space.The inequality between the MSE distance and the mean value or the variance of the candidate codeword is used in the elimination process.The experiment shows 98% of the candidate codewords is discarded,only about 2% of the candidate codewords has to compute the MSE distance.
