一种基于LMS的改进变步长算法及其应用

A New LMS Algorithm of Variable Step Size and Its Application in System Identification

在分析定步长LMS算法和变步长LMS算法的基础上,提出了一种改进的LMS算法,改进的LMS算法是利用瞬时误差的绝对值的三次方和遗忘因子共同来调整步长的.理论分析和系统辨识的仿真结果均表明,新算法确实具有更快的收敛速度.

This paper presented a new LMS algorithm of variable step size after a brief discussion on traditional LMS and VSS-LMS algorithm, which controls step size according to the absolute value of third power of instantaneous error and forgetting factor. Theory analysis and computer simulations demonstrate that the presented algorithm has faster convergence.