Variable Step Size Technique for Adaptive Blind Decorrelation

Blind decorrelation is a related task to blind source separation (BSS) which is applicable to numerous problems. A critical challenge in adaptive blind decorrelation (ABD) is the choice of step size to achieve fast initial convergence speed and low steady state error in time-varying systems. Unfortunately, unlike some supervised training, the error factor of ABD is inaccessible in practice, and so the relation between step size and error factor can not be utilized to adjust the step size. In this paper, we first present an algorithm to restructure the error factor by adopting an auxiliary decorrelation system with some restriction, and then based on a nonlinear updating rule of step-size in the light of the error factor descending in an exponential form, we propose a novel variable step size algorithm for ABD. Simulation results indicate the good performance of our proposed method.

[1]  Wenwu Wang,et al.  Variable step-size sign natural gradient algorithm for sequential blind source separation , 2005, IEEE Signal Processing Letters.

[2]  F. M. Silva,et al.  On the stability of symmetric adaptive decorrelation , 1994, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).

[3]  Xie Sheng-li.  A Variable Step Size LMS Adaptive Filtering Algorithm and Its Analysis , 2001 .

[4]  Andrzej Cichocki,et al.  Neural networks for blind decorrelation of signals , 1997, IEEE Trans. Signal Process..

[5]  Andrzej Cichocki,et al.  Adaptive blind signal and image processing , 2002 .

[6]  A. Cichocki,et al.  Blind separation of nonstationary sources in noisy mixtures , 2000 .

[7]  S.C. Douglas,et al.  Adaptive step size techniques for decorrelation and blind source separation , 1998, Conference Record of Thirty-Second Asilomar Conference on Signals, Systems and Computers (Cat. No.98CH36284).