S函数变步长LMS算法的一种L2范数修正

L2 Norm Correction of Variable Step Size LMS Algorithm Based on S Function

为了提高S函数变步长LMS算法的综合性能和抗干扰能力,引入L2范数控制步长更新.在L2范数中引入输入信号,对输入信号实施动态地跟踪,以提高算法的抗干扰能力;利用归一化输入信号代替原始信号,便于处理信号且可以降低输入信号动态范围对算法的影响.在高和低信噪比条件下的仿真结果表明:相较于其他的变步长LMS算法,所提出算法收敛速度更快和稳态误差更小.

In order to improve the comprehensive performance and anti-interference ability of S function variable step LMS algorithm,the paper used the L2 norm to control step update.The input signal is introduced into the L2 norm to track the input signal dynamically,so as to improve the anti-interference ability of the algorithm.In this study,original signals are replaced by normalized input signals,so that the signal could be processed conveniently and the influence of the dynamic range of the input signal on the algorithm could be reduced as well.Simulation results under high and low SNR conditions show that the proposed algorithm has the advantages of faster convergence rate and smaller steady-state error over other variable step size LMS algorithms.