基于前置滤波和小波变换的带噪语音基音周期检测方法

根据语音信号的基音周期范围有限和在声门闭合时刻语音信号出现锐变的特点，提出一种基于前置滤波和小波变换的基音周期检测方法。带噪语音信号经过3阶椭圆低通滤波器滤波后，采用以二次样条小波作为小波函数，进行一级小波变换检测语音信号的锐变点，再计算基音周期。实验表明，本文提出的基音周期检测方法，与平均幅度差函数(AMDF)和自相关函数(ACF)方法相比，提高了提取基音周期的准确率；与多尺度小波变换的基音周期检测方法相比，减小了计算量，削弱了噪声信号和语音的共振峰对基音周期检测的影响。

Pitch Detection Method for Noisy Speech Signals Based on Pre-Filter and Wavelet Transform

According to the features of limited range of the pitch frequency of speech signals and the sharp change of speech signals at the glottal closure instant, a pitch detection method for noisy speech signals based on pre-filter and the discrete wavelet transform is proposed. The noisy speech signals are filtered by a 3-order lowpass elliptic filter. The wavelet function is the quadratic spline wavelet. The sharp change points are obtained by the one level wavelet transform and then the pitch is calculated. Experiments show that the algorithm can increase the pitch detection precision compared with average magnitude difference function (AMDF) and autocorrelation function (ACF) algorithms. The algorithm decreases computational expense and complexity compared with the multi-scale wavelet transform algorithm. The effects on the noisy signals and speech formants for pitch detection are decreased.