基于奇异性的语音端点检测方法

Speech endpoint detection algorithm based on singularity

噪声信号对于语音信号是相对奇异的。小波变换是分析信号奇异性的有利工具。在利用小波对含噪语音进行分析研究的基础上,提出了一种新的端点检测方法。该算法利用了基于信号奇异性的统计特征和高低频能量比特征。实验结果表明,在低信噪比的情况下,该算法依然能有效地进行语音分割。

White noise signal is singularer than speech signal, Singularity theory ofdetecting signal based on wavelet transtorm is appned to the detection of speech endpoint, on which, a new speech endpoint detection algorithm is proposed. The algorithm is based on the statistic characteristics of singularity and the ratio of high/low spectral energy. The simulation results show that this method is efficient to segment speech even at a low signal-to-noise ratio （SNR）.