基于频谱相关的时域盲分离排序算法

由于水声信道的时变性,水声信号分离后存在排序模糊性,限制了盲分离算法的应用.针对盲分离后信号存在排序模糊性的问题,提出了一种基于频谱相关的盲分离排序算法.根据水声环境中各信号的频谱差异,对分离后信号进行重新排序,来改善盲分离算法的排序模糊性.仿真实验表明,该算法可以有效消除分离信号的顺序不确定性,提高盲分离算法的性能.

A Permutation Method in BHnd Source Separation Based on Correlation of Spectrum

Because the signal channel changes every minute, acoustic signals underwater have permutation inconsistency after separation, which limits the performance of blind source separation. To solve permutation problem, a method in blind source separation based on correlation of spectrum is presented. According to contrast of signals ＇ spectrum, signals are per- mutated in order to improve performance of blind source separation. The result shows that it is an effective method in eliminating the permutation inconsistency.