带约束的卷积语音信号频域盲分离方法

Blind source separation scheme with restrictive condition for convolutive audio signals in frequency-domain

在实际的室内环境中,多通道的语音混合是一个卷积混合信号,在频域利用ICA进行分离时,不同频点上分解出的源信号的次序不确定,需要用后处理方法确定源的对应关系。提出了一种利用波达方向(DOA)作为约束条件的频域盲源分离方法,可以在线地解决频域中的次序不确定性,并且不需要已知传感器及源信号位置等先验知识。仿真结果表明,该方法能够有效地分离卷积混合语音信号,比现有相关的方法更精确。

In real environment, multiple audio mixtures are a convolutive mixture. When separating signals at each frequency bin in frequency domain by independent component analysis, the order ambiguity becomes a serious problem which needs to do post-processing A novel scheme using DOA as a restriction is propased so that the permutation problem can be solved on line and it does not need the location information of receivers like classical DOA method. Compared with the existing algorithms, the proposed algorithm has better performance.