动量滤波-eLMS算法及其性能分析

针对有源噪声控制中滤波-e LMS(最小均方算法)算法收敛速度慢,收敛步长取值范围小及受参考信号自相关矩阵特征值分散程度影响较大的缺点,提出一种改进的滤波-e LMS箅法一动量滤波-e LMS算法.算法在滤波-e LMS算法的基础上,结合动量LMS算法,在权系数更新迭代时引入一个动量项,此动量项包含了先前梯度的估计值.理论推导证明算法不仅可以加快系统的收敛速度还可以扩大收敛因子的取值范围.仿真结果表明,动量滤波-e LMS算法具有收敛速度快、稳态误差小的优点.还讨论了算法中不同动量因子对算法收敛性能的影响,确定了它们的最优取值范围.

Momentum Filtered-e LMS Algorithm and Its Performance Analysis

In view of the shortcomings of low convergence of filtered-e LMS algorithm,and the small range of the convergence step size affected by the input reference signal's auto-correlation matrix eigenvalue dispersion,an improved filtered-e LMS algorithm,i.e.the momentum filtering-e LMS algorithm is presented,which combines the filtered-e LMS with the momentum LMS,by using a momentum term which includes the previous estimate of gradient for updating the weight coefficients. Theoretical derivations show that the al...