自适应LMS算法的寻优步长及其优化选择

SELECTION AND OPTIMIZATION OF THE STEP SIZE OF ADAPTIVE LMS ALGORITHM

本文给出了LMS算法收敛特性的统计分析,推导出了超额均方误差的递推公式,进而给出了保证均方误差均匀收敛的充分条件,然后又以超额均方误差下降最快为准则,推出了最优收敛步长。计算机模拟验证了上述结果。

Statistical analysis of the least mean—squares(LMS)adaptive algorithm is presented. Iterative relation of excess mean—squares error is derived, and then a sufficient condition for steady convergence of mean—squares error is given. At last the optimal step size ensuring fast convergence is derived. Some computer simulation is presented to conform the analysis.