A new prediction algorithm: flexible switch for combining LMS and RBF adaptive filters (CLMF)

  • Authors:

  • M. Reza Pooshideh;Javad Haddadnia;Tahere Royani

  • Affiliations:

  • Department of Electrical Engineering, Tarbiat Moallem University of Sabzevar, Sabzevar, Iran;Department of Electrical Engineering, Tarbiat Moallem University of Sabzevar, Sabzevar, Iran;Department of Electrical Engineering, Tarbiat Moallem University of Sabzevar, Sabzevar, Iran

  • Venue:
  • ISCGAV'10 Proceedings of the 10th WSEAS international conference on Signal processing, computational geometry and artificial vision
  • Year:
  • 2010

Quantified Score

Hi-index 0.02

Visualization

Abstract

Estimation and prediction are important tasks in the communication system. The LMS and RBF are suitable algorithms for prediction of linear and nonlinear model respectively. The LMS (Least Mean Square) algorithm is a linear adaptive filter, which has properties of slow convergence and good tracking in low SNR and RBF (Radial Basis Function) adaptive filter is a nonlinear filter that predict nonlinear behavior of signals. The LMS algorithm can predict each part of signal which has linear form better than the RBF adaptive filter. However, if linear combining of buffered signal can predict subsequent sample, the RBF adaptive filter have not results as well as the LMS algorithm. Suitable soft or flexible switch is needed for this purpose. In this paper, a stochastic gradient based switch is proposed based on error of the LMS and RBF algorithms. The proposed algorithm is configured for prediction of sinusoidal signal and chirp tracking problem. Experimental results show better performance compared to both the RBF and LMS algorithms in prediction problem and noisy chirp tracking.