盲信号分离输出与源信号的一致性判断

Judgment on Consistency of Blind Source Separation Outputs with Source Signals

针对盲信号分离需要判断分离结果与源信号是否一致的问题,基于统计独立变量函数仍然保持“统计独立”的性质,提出了独立分量分析(ICA)的输出分量与源信号的一致性判断方法.该方法通过计算混合信号及其差分值混合矩阵的相关矩阵,根据ICA各分量对应的最大相关系数来判断ICA各分量与源信号的一致性.模拟计算和实验结果表明:若差分前后混合矩阵的最大相关系数趋近于1,则ICA输出分量与对应的源信号一致;为保证分离的准确,差分前后混合矩阵的最大相关系数不应小于0.8.

In the blind source separation （BSS）, it is necessary to judge whether the BSS outputs are consistent with the source signals. In this paper, based on the statistical independence of statistical independence variable functions, a method to judge the consistency of the outputs in independent component analysis （ICA） with the source signals is presented. In this method, the correlation matrix of two mixing matrices which respectively belong to the source signals and the source differences is calculated, and the consistencies of the independent components with the source signals are then judged according to the maximum correlation coeflqcients corresponding to each independent component. Simulated and experimental results indicate that the independent components are consistent with the corresponding source signals when the maximum correlation coefficients converge and that the maximum correlation coeflqcient should not be less than 0. 8 to ensure the separation accuracy.