自适应步长加权正交约束自然梯度ICA算法

›› 2012, Vol. 25 ›› Issue (12): 7-.

自适应步长加权正交约束自然梯度ICA算法

唐兴佳,张秀方

(西安电子科技大学理学院,陕西西安 710071)

出版日期:2012-12-15 发布日期:2013-02-28
作者简介:唐兴佳(1987—),男,硕士研究生。研究方向:信息与盲信号处理。张秀方(1988—),女,硕士研究生。研究方向:概率模型优化。
基金资助:
国家自然科学基金资助项目(61075117/F030402)

Natural Gradient ICA Algorithm Based on Adaptive Step-size and Weighted Orthogonal Constraint

TANG Xin-Jia, ZHANG Xiu-Fang

(School of Science,Xidian University,Xi'an 710071,China)

Online:2012-12-15 Published:2013-02-28

摘要/Abstract

摘要：

针对传统自然梯度ICA算法的不稳定和分离结果不准确,提出一种自适应步长加权正交约束自然梯度ICA算法。首先,基于分离矩阵所满足的正交性约束,引入一种单步正交性修正方法。然后,根据相邻迭代结果之差可用于平滑构造每步迭代结果与最优值的距离,设计出一种单步误差估计函数。最后,据此误差估计函数引入一种自适应调整的步长。仿真实验表明,自适应步长加权正交约束自然梯度ICA算法,相比于传统的自然梯度ICA算法具有更快的收敛速度,且算法的稳定性和分离结果的准确性都得到了较大提高。

关键词: ICA, 最速下降, 加权正交约束, 自适应步长

Abstract:

For the instability of the traditional natural gradient ICA algorithm and the inaccuracy of separation results,a natural gradient ICA algorithm combining adaptive step-size and weighted orthogonal constraint is put forward.Firstly,based on the orthogonal constraint of separation matrix,a single-step orthogonal correction method is introduced.Then,because the difference of adjacent iteration results can be used to construct smoothly the distance from iteration result to optimal value,a single-step error estimation function is proposed.Finally,according to this error estimation function,an adaptive step-size is raised.Simulation experiments show that the rate of convergence of natural gradient ICA algorithm combining adaptive step-size and weighted orthogonal constraint is higher than that of the traditional natural gradient ICA algorithm,and at the same time the stability of the algorithm and the accuracy of the separation results have been improved greatly.

Key words: ICA;steepest descent;weighted orthogonal constraint;adaptive step-size

中图分类号:

TN911.7

唐兴佳, 张秀方. 自适应步长加权正交约束自然梯度ICA算法[J]. , 2012, 25(12): 7-.

TANG Xin-Jia, ZHANG Xiu-Fang. Natural Gradient ICA Algorithm Based on Adaptive Step-size and Weighted Orthogonal Constraint[J]. , 2012, 25(12): 7-.

参考文献

［1］ANDRZEJ C,SHUN-CHI A.自适应盲信号与图像处理［M］.吾正国,唐劲松,章林柯,译.北京:电子工业出版社,2004.

［2］HYVARINEN A,KARHUNEN J,OJA E.Independent component analysis［M］.Newyork:Wiley Press,2001.

［3］CARDOSO J F,LAHELD B.Equivariant adaptive source separation［J］.IEEE Transactions on Signal Processing,1996,45(2):434-444.

［4］AMARI S.Natural gradient works efficiently in learning［J］.Neural Computation,1998(10):251-276.

［5］BELOUCHRANI A,MERAIM K,CARDOSO J F,et al.A blind source separation technique using second-order statistics［J］.IEEE Transactions on Signal Processing,1997,45(2):434-444.

［6］ZHANG W T,LIU N,LOU S T.Joint approximate diagonalization using bilateral rank-reducing householder transform with application in blind source separation［J］.Chinese Journal of Electronics,2009,18(3):471-476.

［7］DEGERINE S,KANE E.A comparative study of approximate joint diagonalization algorithms for blind source separation in presence of additive noise［J］.IEEE Transactions on Signal Processing,2007,55(6):3022-3031.

［8］ZHU Xiaolong,ZHANG Xianda,DING Zizhe,et al.Adaptive nonlinear PCA algorithms for blind source separation without prewhitening［J］.IEEE Transactions on Circuits and Systems,2006,53(3):745-752.

［9］YE Jimin,HUANG Ting.New Fast-ICA algorithms for blind source separation without prewhitening［J］.Communications in Computer and Information Science,2011,255(2):579-585.

［10］HYVARINEN A.Fast and robust fixed-point algorithms for independent component analysis［J］.IEEE Transactions on Neural Network,1999,10(3):626-634.

［11］YANG H H,AMARI S.Adaptive on-line learning algorithms for blind separation:maximum entropy and minimum mutual information［J］.Neural Computation,1997,9(5):1457-1482.

［12］刘宁,张伟涛,楼顺天.可变步长正交性约束的自然梯度盲信号分离算法［J］.中国铁道科学,2010,11(6):98-102.

［13］THOMAS P,ALLEN G,AUGUST N.Step-size control in blind sourse separation［C］.Helsinki,Finland:International Workshop on Independent Component Analysis and Blind Sourse Separation,2000:509-514.

［14］ZHU Xiaolong,ZHANG Xianda.Blind source separation based on optimally selected estimating functions［J］.Journal of Xidian University,2003,30(3):335-339.

[1]	田谨,应捷,朱丹丹. 一种新的箭头形交通信号灯识别方法[J]. , 2015, 28(11): 150-.
[2]	张庆锐. 一种新的自适应步长ICA算法[J]. , 2013, 26(1): 123-.