摘要
根据几种变步长最小均方误差(VSLMS)算法的特性,把它们分为2类.把2类VSLMS算法的步长公式结合起来,提出一种改进的VSLMS算法.对改进前后的算法进行仿真实验,实验结果表明,改进算法同时获得了2类算法的优良性能,在收敛性、稳态误差及跟踪性方面都有显著提高.
According to the features about several variable step-size least mean square (V^LM~) algorithms, they were divided into two categories. Then, after combining the step-size formulas of two algorithms, an improved VSLMS algorithm was proposed. Finally, the improved algorithm and the original algorithms were simulated, the experimental results showed that the improved algorithm obtained all good performance of two different algorithms at the same time, and it improved obviously on convergence, steady-state error and traceability performances.
出处
《安徽大学学报(自然科学版)》
CAS
北大核心
2013年第4期68-72,共5页
Journal of Anhui University(Natural Science Edition)
基金
国家自然科学基金资助项目(61076086)
高等学校博士学科点专项科研基金资助项目(103401110008)
安徽大学"211工程"基金资助项目
关键词
LMS
变步长
学习曲线
LMS
variable step-size
learning curve