非平稳语音序列的循环格型算法  

Adaptive circular lattice algorithms for unstable speech sequences

作　　者：钱正祥[1] 徐华[1] 杨溢[1] 

机构地区：[1]解放军炮兵学院,合肥230031

出　　处：《仪器仪表学报》2006年第z2期1419-1420,共2页Chinese Journal of Scientific Instrument

摘　　要：在语音信号序列的向量预测编码中,利用向量量化器把编码结构中产生的预测残差向量量化,可得到比传统标量量化更好的性能。向量的最优线性预测将转化为标量的最小二乘估计,通过标量运算就能实现并行处理。由此出发,运用Hilbert空间通过的正交投影原理,可导出语音信号序列参数线性预测的循环格型自适应算法。In the processing of vector predictive coding of speech signal sequences, the predictive residual error vector produced in coding structure, can be gauged by vector quantizer, whjch can obtain a better performance than that of the traditional scalar quantizer. The optimal linear prediction of vector will be translated into scalar estimation of least squares, then through scalar computation parallel processing can be realized. Based on these, the adaptive circular lattice algorithms of speech signal sequences can be achieved by using the principle of orthogonal projection in Hilbert space.

关 键 词：语音信号序列 向量预测 并行处理 格型滤波 自适应算法 

分 类 号：TH7-55[机械工程—仪器科学与技术]