一种混叠语音清浊音判断方法

A Voicing Decision for Mixed Speech

提出了一种混叠语音清浊音判断方法。该方法在分频带自相关函数的混叠语音基频分离提取算法中,引入了双周期度及频带组的分布度两个参数用于对混叠语音清浊判断,从而使基于分频带自相关函数的混叠语音基频分离提取方法不再局限于浊音,可推广到清音信号。实验结果证明算法有一定的清浊音判断能力。

A voicing decision for mixed speech is presented in this paper. Two characters: double periodic degree and frequencies groups distribution degree are introduced into multiband autocorrelation functions pitch extracting method for mixed speechs voicing decision. Therefore multiband autocorrelation functions pitch extracting method is no longer limited to vowels, but unvoiced signal as well. The experiment results show that the method has comparative voice/unvoiced classification ability.