一种改进的低复杂度变步长LMS算法

An Improved Low Computational Complexity Variable Step Size LMS Algorithm

改进型的变步长LMS算法在有效抑制瞬时噪声对经典的变步长LMS算法影响的同时,也增加了算法的计算复杂度,提高了其硬件实现的难度。为降低变步长LMS算法的计算复杂度,提出了一种步长改变因子与前后两个时刻误差的乘积成正比的新的变步长LMS改进算法,在不增加计算复杂度的条件下,有效地抑制了瞬时噪声对迭代步长的影响。仿真结果表明,提出的算法和现有的变步长LMS算法收敛速度相当,但其稳态误差更小,计算复杂度也更低,有利于算法的硬件实现。

Although the modified variable step size LMS algorithm can effectively suppress the instant noise interferences on the step size, its computational complexity is increased. Consequently, the difficulty for the hardware realization is increased. In order to reduce the complexity of the variable step size LMS algo-rithm, a novel modified variable step size algorithm with step variation proportional to the product of cur-rent error and previous error is proposed in this paper. The proposed algorithm can well restrain the instant noise interferences without increasing the computational complexity. The simulation results indicate that compared with the existing variable step size of LMS algorithms, the convergence speed of the proposed al-gorithm is about the same, but its steady-state error is much smaller. Meanwhile, its computational com-plexity is lower, and favorable to hardware realization.