用弗马数论变换作语音信号自相关分析

Autocorrelation analysis of speech signal using Fast Fermat Number Transform

线性预测编码(LPC)是语音信号处理的一种强有力的分析技术.此时把信号看作非平稳随机过程,用自相关函数来计算短时预测参数.求解线性方程组的方法是简便的,但自相关系数的计算是十分繁重的,从而影响了处理的速度和实时性.为了提高计算速度,本文选用快速弗马数论变换(FFNT)计算自相关系数.计算机仿真表明,它快于和优于直接法.本文中的FNT具有与FFT相同的结构,因而很便于用硬件实现。

LPC is one of the most powerful technigue in speech signal processing. The speech signal can be considered as non-stationary random process and the autocorrelation function are used to calculate theshort-time predictive parameters. The method of solving linear equatimss feasible, but the computation ofautocorrelation coefficients is very havy, therefore the speed of processing and real-time performance would beof influence. In order to speed up the processing this paper chaises the Fast Fermat Number Transform (PENT).The computer simulation shous the FNT is faster and better than the direct method The structure of FNT pres-ented in this paper is similar to the FFTs and suitable for hardware implementation.