基于邻近频点相关特性的水声信号盲源分离

Blind Source Separation of Underwater Acoustic Signals Based on the Correlation between Neighbor Frequency Bins

排序和幅度不一致性是信号频域盲源分离的主要困难。该文建立了邻近频点相关特性理论,并针对水声信号进行深入研究,结论表明单个水声信号邻近频点间相关特性良好,且性能非常稳定;而两个不同水声信号邻近频点相关性非常弱。提出基于邻近频点相关特性的盲源分离算法,用于消除卷积信号盲源分离过程中排序不确定性,实验表明该方法对卷积混合形式的水声信号能取得较好分离效果。

Indeterminacies in amplitude and permutation are the two main cumbersome aspects in frequency domain blind signal separation. A correlation theory between neighbour frequency bins is developed in this paper. It is found that the correlation between neighbor frequency bins of single underwater acoustic signal is strong and stable, while it is very weak in the case of two different signals. Based on these correlation characters, a novel blind source separation method is developed, which can get rid of the permutation indeterminacy. Simulation experiments prove that convolution mixed underwater acoustic signals can be well separated by this method.