基于有理多项式非线性函数的FastICA算法

FastICA Algorithm Based on Rational Polynomial Nonlinear Function

源信号数目较多的混合信号盲源分离中,快速独立分量分析FastICA算法存在迭代次数过高、收敛速度慢、分离性能恶化等问题,为此,提出了一种基于有理多项式非线性函数的FastICA算法N-FastICA,并通过实验分析了有理多项式函数的阶数和系数选取。仿真结果表明,NFastICA算法的PI值小于FastICA算法的PI值,且都接近于0,分离性能优于FastICA算法。随着信号数目的增多,由相似系数矩阵判断N-FastICA算法能成功分离源信号,而FastICA算法已不能成功分离出源信号,前者迭代次数远少于后者。

Fast independent component analysis(FastICA)algorithm has many problems such as high number of iterations,slow convergence rate and poor separation performance in mixed-signal blind source separation with a large number of source signals.To improve the performance of FastICA,one FastICA algorithm based on rational polynomial nonlinear function(N-FastICA)is proposed.The order and coefficients selection of the rational polynomial function is analyzed experimentally.The results of computer simulation show that the PI values of N-FastICA algorithm is smaller than FastICA algorithm and closing to 0,which means the separation performance of N-FastICA algorithm superior to FastICA algorithm.With the increase of the number of signals,the similarity coefficient matrix shows the N-FastICA algorithm can successfully separate the source signal,and the FastICA algorithm can not successfully separate the source signal.The former iterations number is far less than the latter.