基于数据压缩的LSP参数高效量化算法  

High Quality Quantization of LSP Parameters based on Data Compression

作　　者：商晓燕[1] 张雄伟[2] 林龙[3] 于坚[3] 

机构地区：[1]解放军理工大学通信工程学院研究生1队 [2]解放军理工大学通信工程学院电子信息工程系 [3]中国人民解放军73683部队

出　　处：《军事通信技术》2008年第4期25-28,共4页Journal of Military Communications Technology

摘　　要：文中为解决超长帧编码系统中联合帧数过多导致LSP参数量化性能下降这个问题,将多项式拟合引入超长帧编码系统中,利用多项式拟合压缩LSP参数矩阵,以降低量化时实际联合的帧数。再用分裂矩阵量化器进行量化。仿真结果表明,在120 bps的参数量化速率下,此算法的量化性能优于矩阵量化算法(MQ),平均谱失真比MQ算法降低了0.0719 dB,平均分段信噪比提高了0.1971 dB。The polynomial approximation was introduced in order to solve the problem that too many frames reduce the quality of quantization of the LSP parameters in the super-frames coder system.The LSP parameters matrix was compressed by the polynomial approximation,reducing the number of the frames when quantized.Then the compressed matrix was quantized by split matrix quantizations.The results of the emulation show that,when the quantization rate is 120 bps,the quality of the arithmetic above is better than the MQ with ASDM reducing 0.0919 dB and SegSNR reducing 0.1971 dB.

关 键 词：多项式拟合 超长帧联合 分裂矩阵量化 

分 类 号：TN91[电子电信—通信与信息系统]