本文提出了基于快速胞腔划分的改进分裂法语音信号矢量量化码书设计新算法。新算法应用特征变量对码书进行合理的构造，使得码书在结构上具有秩序性。利用这种秩序性进行快速划分胞腔，从而压缩分裂法码书训练中庞大的运算量。计算机实验结果表明，新算法明显减少了码本训练的复杂度，提高了矢量量化码书的训练速度。实验还分析了码书扩展参数r的不同取值对码书性能的影响． In this paper, we propose a new algorithm for vector quantization codebook design of speech signal based on fast cell partition. The new algorithm uses feature variables to reasonably construct the codebook so that the codebook is structurally ordered. Utilize this kind of order to carry on the rapid division cell, thus compressing and splitting law yard book training huge computation quantity. Computer experimental results show that the new algorithm significantly reduces the complexity of codebook training and improves the training speed of vector quantization codebook. The experiment also analyzes the influence of different values ​​of codebook extended parameter r on the codebook performance.