分段声码器中的LSP参数量化算法研究

为解决语音分段后提取出的LSP参数矩阵维数不固定问题,采用多项式拟合取代段长弯折以固定参数矩阵的维数。方案以降低维数转换引起的谱失真为出发点,对不同范围的段长进行不同阶次的多项式拟合,再对拟合后的矩阵进行量化编码。仿真表明,提出的方案具有相对良好的量化性能。

Research on quantization of LSP parameters in segment vocoder

In order to solve the problem of unfixed dimension of LSP parameter matrix extracted after speech segmentation, polynomial fitting is used to replace the duration bend to fix the dimension of the parameters matrix. The scheme takes reducing spectral distortion induced by the conversion of the dimension as the goal, and conducts different degree polynomial fitting with different duration range, and then the fitted matrix is quantized coded. The simulation results show that, the proposed solution has good quality of quantization.