对含噪语音进行基频检测的方法

Pitch detection for the noisy speech

提出一种对含噪语音进行基频检测的新方法。先对含噪语音进行小波去噪,然后再经过预处理后,采用归一化的AMDF算法对语音进行基频提取,后期对基频信号采用搜索试探方法进行平滑处理,通过实验表明,该方法比传统方法有更好的鲁棒性,尤其在低信噪比的情况下。

This paper proposes a pitch detection method for the noisy speech.Firstly,the wavelet transform is used to de-noising the noisy speech,then after the pre-processing,the normalized AMDF is adopted to extract the speech＇s fundamental frequencies. During the post-treating,the searching method is used to smooth the fundamental frequencies.Experimental results show the new method has better robust than the traditional method,especially in the low snr.