基于矢量模型的语音线性预测研究

Speech Linear Prediction Based on Parallel Processing Algorithms

研究语音参数线性预测的并行处理问题。通过把语音源序列的相邻样本分组能够构成一个协方差平稳的语音向量自回归序列 ,在 Hilbert空间中运用正交投影原理可导出具有高度并行处理能力的向量预测编码策略 ,由此可推出语音参数线性预测的并行处理自适应算法。同传统格型算法相比 ,这种算法的计算复杂度及存贮量有明显改善。最后通过仿真运算检验了算法的性能。

The parallel processing algorithms of linear prediction for speech parameters are studied. The autoregressive speech vector series with stable convariance may be gained after a classification of adjacent sample of source speech series. A strategy of vector predictive coding with high parallel processing ability can be achieved with the application of the principle of orthogonal projection in Hilbert space. Adaptive parallel processing algorithms can be given by their properties. Comparised with the traditional lattice method of calculation complexity and storage, the algorithms have obvious advantages. Finally, it is confirmed that the algorithms are effective through computer simulation.