Method for improving RLS algorithms

Abstract

The recursive least-square (RLS) algorithm has been extensively used in adaptive identification, prediction, filtering, and many other fields. This paper proposes adding a second-difference term to the standard recurrent formula to create a novel method for improving tracing capabilities. Test results show that this can greatly improve the convergence capability of RLS algorithms.

Keywords

adaptive model algorithms / RLS / tracing capabilities

References

[1] Brockwell P. J., Davis R. A. Time series: theory and methods[M]. 1991, New York: Springer-Verlag Inc

[2] Sun J., Jin Lijun. Application of several improved RLS algorithms in adaptive filter system[J]. Journal of Chongqing University of Posts and Telecommunications, 2003, 15(3): 14-17

[3] Wu Q., Wang J., Shen L., et al. Robust RLS algorithm for adaptive arrays[J]. Acta Electronic Sinica, 2002, 30(6): 893-895

[4] Li J., Wang S., Wang Fei. Parameter estimation of adaptive chirp signal based on polynomial phase transforms[J]. Journal of Jinlin University, 2004, 34(4): 617-621

[5] Park D., Jun B. E. Self-perturbing recursive least squares algorithm with fast tracking capability[J]. Electronics Letters, 1992, 28(6): 558-559

[6] Jiang J., Cook R. Fast parameter tracking RLS algorithm with high noise immunity[J]. Electronics Letters, 1996, 28(22): 2043-2045

[7] Eom K., Park D. Fast tracking and noise-immunized RLS algorithm based on Kalman filter[J]. Electronics Letters, 1996, 32(25): 2311-2312

[8] So C. F., Ng S. C., Leung S. H. Gradient based variable forgetting factor RLS algorithm[J]. Signal Processing, 2003, 83: 1163-1175