关键词: 语音活动检测（端点检测）, 自适应, 噪声评估, 特征融合

Abstract: In view of the question that the method of voice endpoint detection based on short-time energy and zero-crossing-
rate can’t adapt to different environments especially the low-SNR condition, this paper presents a new algorithm by evaluating the environment noise with the minimum short-time energy. Then by fusing the two optimized characteristic parameters together compensate each other’s drawback, the final parameter is able to narrow the spread among voice and enlarge the difference between voice and noise. Finally, a dynamic threshold is employed for environment adaptation.

Key words: Voice Activity Detection（VAD）, self-adaptive, noise evaluation, feature fusion