信号盲分离的非线性主分量分析算法

Principal component analysis of blind signal separation with nonlinear approaches

主分量分析是统计信号处理中常用的算法 ,将非线性引入主分量分析算法 ,可以完成对输入信号独立分量的分离 .分析认为现有的非线性主分量分析算法只能实现实数信号的分离 ,对复数信号无效 .通过对非线性函数进行修改 ,提出了一种非线性主分量分析复数算法 ,成功地实现了复数信号的盲分离 .文中还借助于计算机仿真 ,对实数和复数算法分离亚高斯和超高斯信号混合的特性进行了分析评价 .

The independent source signals can be separated from their linear mixture by incorporating nonlinear function into a standard principal component analysis (PCA) which is a widely used technique in statistical signal processing. Some recent nonlinear PCA approaches to blind signal separation are reviewed in present paper. It is found that present nonlinear PCA is only valid to real valued signals but invalid to complex valued signal.With modifying the form of activation function, a class of complex nonlinear PCA algorithms which is successful to blind complex signals separation is proposed. Furthermore, the property for separating the mixture of sub-and super-Gaussian signals is analyzed and evaluated by computer simulation.