Abstract: This paper proposes a new method of speech endpoint detection in high noisy environment.In order to solve the problem of the less effective detection of surd and the poor anti-noise performance in spectrum entropy characteristic，this paper combines short-time ZCR characteristic which has a better detection of surd with Mel cepstral distance characteristic which has a good anti-noise performance to realize the endpoint detection based on multi-characteristics in high noise environment.The simulation shows that the new method can significantly improve the detection performance in high noise environment.

Key words: high noise, Mel cepstral distance, spectral entropy, short-time ZCR, endpoint detection

摘要： 提出了一种抗噪声的端点检测新方法。针对谱熵特征对清音的检测性能以及抗噪声性能较差的缺点，结合对清音检测性能较好的短时过零率特征，以及抗噪声性能良好的美尔倒谱距离特征，实现了基于多种特征相结合的抗噪声的语音端点检测。仿真实验表明，该方法能显著提高端点检测在高噪声环境下的检测性能。

关键词: 高噪声, 美尔倒谱距离, 谱熵, 短时过零率, 端点检测