一种基于相关函数的LMS算法

LMS Algorithm Based on Correlation Function

从 Asharif定义的相关函数均方误差 (Mean squares error,MSE)准则 Jr(n)出发 ,在利用牛顿梯度法时用当前时刻的梯度估计代替前一时刻的梯度估计 ,并由矩阵求逆定理导出了一种新的相关函数自适应滤波算法。理论分析表明 ,新的算法对步长因子 μ>0是无条件收敛的 ,该算法应用于回波消除中的计算机数值仿真 ,结果显示其收敛性能良好 ,优于 CLMS算法。

Based on correlation function mean squares error criterion J r(n ) proposed by Asharif, a new correlation function adaptive filtering algorithm is proposed by replacing the previous gradient vector with the current gradient vector in the Newton gradient method and using the matrix inversion lemma. Results of theoretical analysis show that the algorithm is stable for step size factor μ>0 . Simulation results indicate that the algorithm′s convergence performance is better than that of CLMS algorithm.