Title: Utilising variable vector step-size in normalised LMS algorithm

Authors: Ning Li, Yonggang Zhang, Yanling Hao

Addresses: College of Automation, Harbin Engineering University, Harbin, 150001, China. ' College of Automation, Harbin Engineering University, Harbin, 150001, China. ' College of Automation, Harbin Engineering University, Harbin, 150001, China

Abstract: A new variable step-size normalised least mean square (NLMS) algorithm is proposed in this paper, which is designed for the applications where the unknown filter has an exponential decay impulse response. In the proposed algorithm, a variable vector step-size is used on the basis of maximising the decrease of the mean square deviation (MSD). Theoretical analysis is given to show the validity. Compared with other variable step-size NLMS algorithms, the proposed method has a faster convergence rate and smaller steady state mean square error (MSE). Simulation results show the improved performance of the proposed algorithm. The same idea of utilising variable vector step-size can be expanded to other variable step-size NLMS algorithms.

Keywords: adaptive filtering; least mean squares; normalised LMS; NLMS; variable step-size; mean square deviation; simulation.

DOI: 10.1504/IJCAT.2011.042244

International Journal of Computer Applications in Technology, 2011 Vol.41 No.1/2, pp.137 - 142

Published online: 01 Sep 2011 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article