基于小波包变换的自适应门限的语音激活检测

Voice Activity Detection Based on Wavelet Packet Transform and Adaptive Threshold

语音激活检测是语音信号处理的一个重要环节。在低信噪比的情况下,传统的检测方法已不适用。为了提高语音激活检测的性能和鲁棒性,针对主要由白噪声组成的噪声背景,提出了一种基于小波包变换的自适应门限的语音激活检测方法(VAD),它将语音信号进行小波包变换,得到各个子带信号,各个子带信号通过Teager能量算子(TEO)将有声部分强化,同时衰减无声部分,最后进行自适应门限判决。实验结果表明在低信噪比的情况下,算法能够正确判别语音段和噪声段。

Voice activity detection is a crucial part in speech processing. Some traditional detection methods are ineffective in low SNR. To improve the performance and robustness, a new voice activity detection algorithm based on wavelet packet transform and adaptive threshold is proposed for the situation with white noise environment. The critical sub - band signals are obtained by wavelet packet transform, then smoothing the TEO of corresponding wavelet coefficients to enhance the discrimination capability of speech components, and lastly the adaptive threshold judgement is carried out. The experimental results prove that the new algorithm can efficiently distinguish the beginning and end points of voice segments in very low SNR environments.