一种改进变步长LMS自适应滤波算法

An Improved LMS Adaptive Filtering Algorithm with Variable Step Size

步长在LMS算法迭代过程中属于一个矛盾量,控制算法的收敛速度和稳态误差,为了同时获得较小的稳态误差和较快的收敛速度,相关学者提出了多种变步长LMS算法,采用时变的步长来提高算法的瞬时收敛速度并减小稳态误差。在传统LMS算法及多种变步长LMS算法的研究基础上,提出一种新的改进变步长LMS算法。算法通过构建一个新的非线性函数表达式,替代G-SVSLMS算法中的常值参数,使算法在收敛速度较快的情况下,获得较小的稳态误差,克服了G-SVSLMS算法在稳态阶段,受噪声影响,步长变化过快的缺点。仿真结果表明,相比原有算法,改进算法的性能更优,且在低信噪比的环境下,取得了良好的效果。

The step size is to a contradiction in the iterative process of the LMS algorithm,which controls the convergence speed and steady-state error of the control algorithm.In order to obtain smaller steady-state error and faster convergence speed,related scholars have proposed a variety of variable step size LMS algorithms,using time-varying step size to improve the instantaneous convergence speed of the algorithm and reduce the steady-state error.This paper proposes an improved variable step size LMS algorithm based on the research of traditional LMS algorithm and a variety of variable step size LMS algorithms.The algorithm constructs a new nonlinear function expression to replace the constant parameter in the G-SVSLMS algorithm,so that the algorithm obtains a small steady-state error in the case of faster convergence.and overcomes the shortcomings of the G-SVSLMS algorithm in the steady state phase,which is affected by noise and the step size changes too fast.The simulation results show that the improved algorithm has better performance than the original algorithm,and has achieved good results in a low SNR environment.