摘要
首先对语音信号进行小波变换,引入反映信号能量分布特性的小波熵,进行端点检测。并根据浊音的特点,改进了自然加权因子。仿真表明,在低信噪比条件下其检测效果要好于自适应子带谱熵法。
First,the wavelet transformation is used on vocal signals.Then the wavelet entropy reflecting energy distributing characteristics of the signal is introduced for detecting the point.Finally,the natural weighting factor is improved based on the characteristic of sonant sound.Simulation results indicate that the new algorithm outperforms the adaptive band-partitioning spectral entropy-voice activity detection(BSE-VAD) in detection result in condition of low signal-to-noise ratio.
出处
《计算机应用与软件》
CSCD
2011年第2期227-228,259,共3页
Computer Applications and Software
关键词
端点检测
能量熵
小波能量
小波熵
Voice endpoint detection(VED) Energy entropy Wavelet energy Wavelet entropy