盲源分离是信号处理领域中一门新的信号处理技术,由于不需要知道信号的先验信息而使其在信号处理中得到广泛的应用,语音分离是其重要的应用领域之一。然而由于语音的短时平稳性,在帧长小于100ms时,语音信号间存在着显著的相关性,因此,批处理模式的语音盲分离算法不能处理实时语音环境中的动态特性。 Blind source separation is a new signal processing technology in the field of signal processing. Because it does not need to know the prior information of signal, it is widely used in signal processing. Speech separation is one of the most important applications. However, due to the short-time stability of speech, there is a significant correlation between speech signals when the frame length is less than 100 ms. Therefore, the WPA algorithm in batch mode can not deal with the dynamic characteristics in real-time speech environment.