Performance Comparison of Variable Step-Size NLMS Algorithms

The purpose of a variable step-size normalized LMS filter is to solve the dilemma of fast convergence rate and low excess MSE. In the past two decades, many VSS-NLMS algorithms have been presented and have claimed that they have good convergence and tracking properties. This paper summarizes several promising algorithms and gives a performance comparison via extensive simulation. Simulation results demonstrate that Benesty's NPVSS and our GSER have the best performance in both time-invariant and time-varying systems.

[1]  Junghsi Lee,et al.  The Generalized Square-Error-Regularized LMS Algorithm , 2008 .

[2]  Danilo P. Mandic,et al.  A generalized normalized gradient descent algorithm , 2004, IEEE Signal Processing Letters.

[3]  Iain B. Collings,et al.  Interference-Normalized Least Mean Square Algorithm , 2007, IEEE Signal Processing Letters.

[4]  Danilo P. Mandic,et al.  Collaborative Adaptive Learning using Hybrid Filters , 2007, 2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP '07.

[5]  Young-Seok Choi,et al.  Robust Regularization for Normalized LMS Algorithms , 2006, IEEE Transactions on Circuits and Systems II: Express Briefs.

[6]  Masayuki Kawamata,et al.  A new variable step size LMS algorithm-based method for improved online secondary path modeling in active noise control systems , 2006, IEEE Transactions on Audio, Speech, and Language Processing.

[7]  Young-Seok Choi,et al.  Adaptive Regularization Matrix for Affine Projection Algorithm , 2007, IEEE Transactions on Circuits and Systems II: Express Briefs.

[8]  Ali H. Sayed,et al.  Variable step-size NLMS and affine projection algorithms , 2004, IEEE Signal Processing Letters.

[9]  Tyseer Aboulnasr,et al.  A robust variable step-size LMS-type algorithm: analysis and simulations , 1997, IEEE Trans. Signal Process..

[10]  Jacob Benesty,et al.  A Nonparametric VSS NLMS Algorithm , 2006, IEEE Signal Processing Letters.