一种EFICA的稳定性改进算法  

Improved algorithm for stability of EFICA

作　　者：彭赛阳 王振华[1] 刘东青[1] 张泽龙[1] PENG Saiyang;WANG Zhenhua;LIU Dongqing;ZHANG Zelong(Air Force EarlyWarning Academy,Wuhan 430019, China)

机构地区：[1]空军预警学院,武汉430019

出　　处：《空军预警学院学报》2016年第6期422-425,共4页Journal of Air Force Early Warning Academy

摘　　要：针对EFICA算法稳定性会因随机选择初始迭代矩阵而受到影响的问题,提出使用最速下降法选出合适的迭代初始矩阵对EFICA算法进行了改进,给出了EFICA算法以及EFICA改进算法与Fast ICA算法的对比实验.仿真实验结果表明,对于通信信号,EFICA改进算法的平均串音误差在数值上小于原算法,其方差也由原来的0.740 8降低到0.008 6,有效提高了算法稳定性.Aimed at the problem that the stability of EFICA algorithm is affected by the random selection ofinitial iterative matrix, it is proposed that we should use the steepest descent method to select the appropriateiterative initial matrix for the improvement of the EFICA algorithm, the comparison experiment of the EFICAalgorithm and the improved EFICA algorithm and FastICA algorithm is given in this paper. The simulation resultsshow that for the communication signal, the average crosstalk error of EFICA improved algorithm is less than theoriginal algorithm’s in value, and its variance is reduced from 0.740 8 to 0.008 6 , thus improving the stability ofthe EFICA algorithm, effectively.

关 键 词：盲源分离 改进EFICA算法 最速下降法 

分 类 号：TN971.1[电子电信—信号与信息处理]