一种改进反正切函数的变步长LMS算法  

A variable step size LMS algorithm with improved arctangent function

作　　者：韩啸 陈钟荣[1] HAN Xiao;CHEN Zhongrong(School of Atmospheric Physics,Nanjing University of Information Science and Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学大气物理学院

出　　处：《中国科技论文》2019年第7期759-764,772,共7页China Sciencepaper

基　　金：国家自然科学基金资助项目(41075025)

摘　　要：针对已有的最小均方误差(least mean square,LMS)算法性能方面的缺陷,通过对步长选取问题的研究,提出了一种改进变步长LMS算法。基于原反正切函数引入输入信号反馈因子的二范数项,以实现对输入信号的实时跟踪,提高算法的跟踪能力。用误差相关值代替原算法中的误差平方项,提高其抗干扰能力,并加入补偿项,以补偿算法趋于稳态时的收敛速度。详细讨论了各个参数对算法收敛性、跟踪性、稳定性的影响及算法的抗干扰性。与已有算法进行的仿真结果对比表明,在高低信噪比情况下,改进算法在收敛速度、稳态误差,以及对时变系统跟踪能力等方面均呈现了优异的性能。Aiming at the defects of the existed least mean square(LMS)algorithm,an improved variable step size LMS algorithm was proposed by studying the step selection problem.Based on the original arctangent function,the new algorithm introduced the two norm terms of the input signal feedback factor to realize the real-time tracking of the input signal and improve the algorithm tracking ability.The error correlation value was used to replace the error squared term in the original algorithm to improve its anti-interference ability,and the compensation term was added to compensate the convergence speed of the algorithm tending to steady state.Meanwhile,various parameters on the convergence of the algorithm,tracking ability,impact and stability of the anti-jamming algorithm were discussed in detail.Compared with the simulation results of the existed algorithms,the proposed algorithm exhibits excellent performance in terms of convergence speed,steady-state error,and tracking ability of time-varying systems under high and low SNR.

关 键 词：自适应滤波 变步长LMS算法 反正切函数 反馈因子 稳态误差 

分 类 号：TN911.7[电子电信—通信与信息系统]