清音检测对于语音信号的端点检测等技术具有重要作用。清音检测容易出现的问题主要在两个方面,一是检测的清音容易出现断裂情况,二是容易出现漏检或误检现象。本文利用小波滤波中的Haar小波的块化效应以及不同小波滤波器对清音信息熵的敏感程度不同,对清音的信息熵进行滤波处理,仿真结果表明,这种方法能够在白噪声中检测出清音,即使清音被噪声完全淹没。 Unvoiced sound detection for speech signal endpoint detection technology has an important role. Unclear voice detection prone problems mainly in two aspects, one is the detection of voiceless easily broken situation, the second is prone to missed or false detection phenomenon. In this paper, the blocking effect of Haar wavelet in wavelet filtering and the sensitivity of different wavelet filters to the entropy of unvoicing are different, and the information entropy of unvoiced speech is filtered. The simulation results show that this method can detect unvoiced speech in white noise , Even if the voiceless sound is completely submerged.