一种改进的变步长ELMS算法

A Refrained Variable Step-Size ELMS-Like Algorithm

在简单讨论基小最小均方(LMS)算法的基础上,引入了扩展的最小均方(ELMS)算法,并分析说明了该算法能达到更小的稳态MSE。改进的变步长ELMS算法是在对有用信号的预测中采用了自适应为归一化的的最小均方(NLMS)预测估计器,步长的迭代中引入遗忘因子λi,利用其与误差信号的加权和来产生新的步长参与迭代。理论分析与计算机仿真结果表明,该算法有较好的收敛性能和较小的稳态失调。

Following a brief discussion on basic LMS algorithms, the ELMS algorithms with steady small MSE is introduced. The new algorithm using the adaptive NLMS signal-estimator to predict the signal s（k）. Forgetting factor λi and error signal are used to control the step size update for iteration. Theoretics analysi and computer simulations demonstrate that the presented algorithm has good performance both in convergence properties and steady small misadiustment.