一种高效的线谱对参数分段量化方案  

Efficient segment quantization scheme for LSP parameters

作　　者：张雄伟[1] 闵刚[1,2] 杨吉斌[1] 闫佩君[2] 

机构地区：[1]解放军理工大学通信工程学院,江苏南京210007 [2]西安通信学院,陕西西安710106

出　　处：《解放军理工大学学报（自然科学版）》2009年第1期1-6,共6页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：江苏省自然科学基金资助项目(BK2006001)

摘　　要：为了适应300 b/s以下超低速声码器的需要,提出了一种基于离散余弦变换DCT高效的线谱对参数分段量化方案——DCTSQ(discrete cosine transform segment quantization)方案。采用分段量化的思想,针对语音分段后提取出的线谱对参数矩阵维数不固定问题,对维数转换进行优化分析并利用DCT实现。在此基础上,推导出了DCTSQ方案的码本设计方法。仿真结果表明,相对于目前的分段量化方案和矩阵量化方案,DCTSQ方案可有效提高LSP参数量化效率。In order to meet the need of very low bit rate vocoder below 300 b/ps,an efficient segment quantization scheme called DCT based segment quantization(DCTSQ) scheme for LSP parameters was proposed.The dimension of LSP matrix extracted from speech segment was unfixed,so the dimension conversion matrix was optimized and implemented by DCT.The codebook designing method for DCTSQ scheme was also deduced.Simulation results show that the DCTSQ scheme performs better than the segment quantization scheme and the matr...

关 键 词：线谱对 离散余弦变换 分段量化 语音编码 

分 类 号：TN912.3[电子电信—通信与信息系统]