基于修正反正切函数的变步长LMS算法

Variable step-size LMS algorithm based on modified arctangent function

为了能够减少算法运算时间、减小稳态误差、提高收敛速度、增强跟踪性能以及增加抗噪性能,提出了1种变步长最小均方误差(least mean square,LMS)算法。针对现有LMS类算法在低信噪比下性能不佳、人为设定参数较多等缺点,基于反正切函数,且利用误差的相关函数动态调整步长。理论上分析了该算法复杂度、稳态失调、收敛速度、跟踪性能以及抗噪性能,并分别设计高信噪比和低信噪比的条件下进行实验仿真比较。理论分析结合实验仿真验证:该算法在高低信噪比时均具有较快的收敛速度和跟踪速度,能获得小的稳态误差和稳态失调,且需要设定的参数变量个数少。

In order to balance the computational time, steady-state error, convergence rate, tracking performance and anti-noise performance,a new variable step-size LMS algorithm is proposed. It is a solution to solve the problem of poor performance under low SNR and many artificial parameters. exploiting the arctangent function, and dynamically adjusts step-size based on the error correlation function The theoretical analysis are conducted,including the complexity of the algorithm, the steady state misad- justment, convergence rate and tracking performance and anti-noise performance, and design experimental simulation comparison under the condition of the high signal-to-noise ratio and low signal-to-noise ratio. The theoretical analysis and simulation results show that this algorithm can achieve better convergence rate, tracking speed, lower steady state error and misadjustment under high and low SNR conditions.