基于Sigmoid二次型隶属度函数的改进LMS算法

来源期刊:中南大学学报(自然科学版)2014年第10期

论文作者:徐洋 徐松涛 马健 杨永建 肖冰松 向建军

文章页码:3470 - 3477

关键词:Sigmoid二次型函数;最小均方算法;步长因子;系统跟踪;系统辨识

Key words:Sigmoid quadratic membership function;LMS algorithm;variable step size;system tracking;system identification

摘    要:基于Sigmoid函数变步长最小均方(LMS)算法优点是计算量小、收敛速度快且对时变系统的跟踪性能好,但存在些许不足之处,如当误差信号较小时,步长因子变化过快,对于未知系统辨识速度较慢且可控参量过少。为克服上述缺点,更优化该算法性能,通过建立Sigmoid二次型函数,提出一种新的变步长LMS算法,在收敛过程中动态渐进调整步长大小,在获得较小的稳态误差同时,能够更快达到收敛。研究结果表明:改进算法收敛速度要比其他基于S函数改进算法的快;对时变系统跟踪的性能要优于归一化类的LMS算法。本文算法可在增加少量计算量的前提下,较好地克服误差信号与步长因子之间的矛盾,加快收敛速度,并引入新的可控变量,使调控更加灵活。

Abstract: The least mean square (LMS) algorithm based on S-function some advantages such as has a small amount of calculation, high convergence rate and good tracking performance for time-varying systems. But when the signal’s error is small, the step factor changes too fast, and system identification is not quick enough and the controllable variables are few. To solve the above shortcomings, an algorithm based on the sigmoid quadratic membership function was put forward. The results show that convergence rate of the algorithm is superior to other improved algorithms based on the S-function, and the tracking performance of the time-varying system is better than the improved normalized LMS algorithms. The algorithm put forward in this paper not only overcomes the contradiction between the signal’s error and step factor, but also makes the algorithm more flexible by introducing new controllable variable.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号