一种新的变步长最小均方算法

A new variable step size LMS adaptive algorithm

针对以往变步长最小均方算法形式过于复杂、鲁棒性和通用性较差的问题 ,根据变步长自适应滤波算法的步长调整原则 ,分析了已有的变步长最小均方算法的步长与误差关系曲线特点 ,提出了一种新的变步长最小均方算法 .经过比较知 ,新算法形式简单、收敛性能强 ,通用性好 .此外 ,提出了变步长公式相似时应满足的条件 :e(n)始终位于exp(·)函数之中 ,且当e(n) =0时 ,exp(·) =1.

The previous variable step size LMS algorithms usually have complex form ,robust of a sort and singleness of use. According to the principle of variable step size and two kind of variable step size LMS algorithms, the paper puts forward a new variable step size LMS algorithm which is simpler in form with a better performance in convergence and general use. Furthermore , the paper puts forward a condition about the analogical form of the variable step size formulae: e(n) is always in the function of exp(·) and e(n)=0 when exp(·)=1.