在分析LSP参数降维量化可行性的基础上,设计预处理机制实现了一种无码书LSP参数降维格型矢量量化算法,并提出了两个不同量化速率的VQ方案与MELP算法进行了对比。在24 bit/frame的实验方案中,组合4合成语音的PESQ均分超过了2.5,在无码书的条件下接近MELP有码书MSVQ的性能,可以满足低速率语音编码的音质要求并降低系统对空间复杂度和计算复杂度的需求。 On the basis of analyzing the feasibility of dimensionality reduction of LSP parameters, a dimensionless vector quantization algorithm based on code-free LSP parameters is designed by preprocessing mechanism, and two VQ schemes with different quantization rates are compared with MELP algorithm . In the 24 bit / frame experimental scheme, the PESQ of combined 4 synthesized speech is more than 2.5, which is close to that of MELP with codebook MSVQ without codebook, which can meet the requirement of low-rate speech coding and reduce the system’s Space complexity and computational complexity.