一种新的变步长LMS自适应滤波算法及其应用

A New Variable Step Size LMS Adaptive Filtering Algorithm and Its Application

针对变步长LMS(Least Mean Square)自适应滤波算法不能同时满足较高收敛速度以及较低稳态误差的问题,根据反馈理论提出了一种新的变步长LMS自适应滤波算法,在原有算法模型中通过引入反馈控制函数建立了一种新的步长与误差的非线性函数模型,使得当前的步长值跟当前误差与前一次误差比值的平方相关,通过MATLAB分析了新函数模型中关键参数对滤波性能的影响并确定了合理的关键参数.仿真结果表明:相比原有的算法,改进的新算法极大地提高了收敛速度,同时也降低了稳态误差.新算法性能良好,将其应用于超宽带无线电引信回波信号的滤波处理中,误差的抑制能力提高了4倍,滤波效果较佳.

For the problemthat variable step size LMS（Least Mean Square） adaptive filtering algorithm from a lot of literature could not meet the higher convergence rate and lower steady-state error, a new variable step size LMS adaptive filtering algorithm was proposed based on the feedback theory. A new nonlinear function model on step and error was created by introducing a feedback control function in the original algorithm model, making the current step value associated with the square of the current error and the previous error. The impact of new function model on the filtering performance was analyzed by MATLAB, and reasonable key parameters were determined. Simulation results show that： compared to the original algorithm, the new improved algorithm greatly improves the convergence rate, and the steady-state error is also reduced. The new algorithm has good performance it is applied to the filtering process of ultra-wideband radio fuze echo signal, the error suppression ability is improved by 4 times, and the filtering effect is better.