一种改进的步长自适应EASI算法

An Improved Adaptive Step-size EASI Algorithm

首先定义了描述信号分离状态的一种测度,并在认真分析相关固定步长和变步长EASI算法的基础上,将其加以运用,提出一种新的步长自适应EASI算法。由于该算法步长是基于分离状态的,其学习速率由信号的分离程度自适应的选取,因而能很好地解决收敛速度和稳态误差之间的矛盾。计算机仿真结果与理论分析相一致,证实了该算法明显优于其它的EASI算法。

This paper proposes to use separating state to control the step- size of EASI algorithm for the first time. After detailed analyzing relevant Exed stepsize and variable step- size EASI algorithms, the paper presents a new adaptive step- size EASI algorithm. Because the variability of the new algorithm＇ s step- size is based on separating state, its learning ratio is chosen adaptively according to separating degree, therefore it can improve convergence speed and reduce the misadjustment error in the steady state simultaneously. Computer simulations confirm the theoretical analysis and show the algorithm performance is superior to other EASI algorithms.