摘要: 基于概率密度非参数估计的]’一义高斯密度(GGD)核估计和线性独立成分分析(ICA)神经网络,木文提出了一种新的非参数lCA算法,实现了对源信号分布的全自要求.该算法直接由观测信号样本出发,对分离信号评价函数直接估计,可以只用一种灵活评价函数分离任意的杂系混合信号,并且GGD核可以根据源信号的高阶统计性质自适应的改变以适应不同的要求,从而在一定程度上解决了ICA算法中选取估计信号评价函数的问题.模拟实验说明了所提算法的性能优越性.
[1]Hyvarinen A,Karhunen J,Oja E, Independent component analysis. New York; John Wiley and Sons,2001,476. [2]Wang F S, Zhang L R, independent subspace analysis for blind signal separation using natural gradient algorithm. Journal of Nanjing Univer- sity(Natural Sciences),2011,47(4):410一415.(土法松,张林让.基于自然梯度的独立子空间自信号处理方法.南京大学学报(自然科学),2011,47(4):410~415. [3]Comon P, .Jutten C. Handbook of blind source separation;Independent component analysis and applications. Oxford:Elsevier, 2010,772 [4]Comon P. independent component analysis:A new concept? Signal Processing, 199,36(3): 287一314. [5]Bell A,Scjnowski T.An information maximiza tion approach to blind separation and blind de convolution. Neural Computation, 1995,7(6) 1129一1159. [6]Amari S, Cichocki A,Yang H H. A new learn- ing algorithm for blind signal separation. Neural information Processing System, 1996,8:757一 763. [7]Cardoso J F. High-order contrasts for inde pendent component analysis. Neural Computa tion, 1999,11(1):157一19 2. [8]Lee T W,Girolami M, Sejnowski T.Independ ent component analysis using an extend infomax algorithm for mixed sub-Gaussian and super Gaussian sources. Neural Computation, 1998,11(2):417~441 [9]Zhang X D, Zhu X L,Bao Z. Grading learning for blind source separation. Science in China (Series F),2003,46(1):31~44 [10] Wang F S, Li H W, Li R. Unified nonparamet- rie and parametric ICA algorithm for hybrid source signals and stability analysis, Interna- tional Journal of Inovative Computing, Infot- mation and Control,2008,4(4)933一942. [11]Amari S, Cardoso J. Blind source separation scmiparametric statistical approach. IEEE hransactions of Signal Processing, 1997,45 (11)2692一2700. [12]Silverman B W. Density estimation for statistics and data analysis. Ncw York; Chapman and Hall. 1985. [13]Tesei A,Regazzoni C S. HOS-based general ized noise pdf models for signal detection opti- mization. Signal Processing, 1998,65(2): 267一281. |
No related articles found! |
|