A variable step-size adaptive system identification algorithm based on modified decorrelation principle

DOI

10.2991/icsmim-15.2016.78How to use a DOI?

Keywords

decorrelation; variable step-size LMS algorithm; adaptive system identification; Labview

Abstract

To overcome the defects of traditional adaptive Least-mean-square(LMS) algorithm, figure out the problem that the existing algorithms’ system identification abilities would decline while their input signals are correlated with each other, a novel variable step-size and variable parameters adaptive system identification algorithm based on modified decorrelation principle is proposed. The design theories and process, the selection principles of the adjusting parameters and the performance analysis of the algorithm are all given below. In the algorithm, the rate of its convergence is improved by the modified decorrelation principle. Variable step size in the algorithm is based on hyperbolic tangent function, it is adjusted by the mean value of the gradient vector, so the system can modify it step size by the energy of the input signals and the real-time error. The variable step size also introduce the self-correlation of the real-time error to retrain the influence of the independent noises. Theory and simulation results verified the system identification abilities and the superiorities of the algorithm under different circumstances.

Copyright

© 2016, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Huamin Ge
AU  - Qing Wan
AU  - Weiwei Zhao
PY  - 2016/01
DA  - 2016/01
TI  - A variable step-size adaptive system identification algorithm based on modified decorrelation principle
BT  - Proceedings of the 2015 4th International Conference on Sensors, Measurement and Intelligent Materials
PB  - Atlantis Press
SP  - 414
EP  - 426
SN  - 2352-538X
UR  - https://doi.org/10.2991/icsmim-15.2016.78
DO  - 10.2991/icsmim-15.2016.78
ID  - Ge2016/01
ER  -