基于前馈复式的BSS去噪方法

在BSS中，含噪声信号的盲分离是困难的，因为存在太多的未知条件。针对噪声环境，提出了一种基于前馈复式BSS去噪新方法。由于BSS算法的目的是尽可能恢复源信号，因而其实际输出信号总是由某个较强的源信号迭加另一些较弱的源信号及噪声组成。该方法是把那些较弱的源信号及噪声又假定为一个新的混合源信号Snew，那么通过适当的选择和组合前一次盲分离的结果作为下一次盲分离的源，反复使用盲分离技术就可把这个假定的Snew分离出来，再将Snew与选定的某个恢复源信号取谱差，达到消除噪声，提高分离结果性能指标的目的。

A Feed Forward Multi-Layer BSS Denoising Algorithm

BSS in noisy signals is difficult because there are a great many unknown terms.A new BSS denoising method with additive noise is proposed.The BSS algorithm always does its best to retrieve source signals,so in fact the output is generally a stronger source added by some other weaker sources and noises.The algorithm proposed in this paper assumes the mixture of the weaker sources and noises as a new source Snew,and the outputs of last separation can be regarded as the sources of next separation after proper choice and combination.Then Snew can be separated by use of BSS technique repeatedly.The spectrum difference between Snew and a selected restored source helps to eliminate the influence of noises and improve the performance of this method.What is more,similitude phase graph is also proposed in this paper,which can show the performance of blind separation algorithm straightly.