一种改进的归一化LMS算法

An Improved Normalized LMS Algorithm

传统的归一化最小二乘(LMS)算法可以解决算法收敛速度和采样信号收敛误差之间的矛盾。通过对收敛步长的归一化处理,从而减小收敛过程的稳态误差。但是,这种算法却将收敛的步长变为固定,不利于算法的快速收敛。为此,提出一种改进的归一化LMS算法,将Sigmoid函数变步长最小均方(SVSLMS)算法与归一化LMS算法结合,提高了归一化LMS算法的收敛速度。仿真结果表明,该算法性能较之归一化LMS算法、SVSLMS算法和普通LMS算法均更优异。

The traditional normalized least mean square （LMS） algorithm can solve the contradiction between convergence speed of the algorithm and convergence error of the sampled signals. By nor- malizing the convergence step-size, the steady-state error of convergence process is reduced. But this algorithm turns the convergence step-size into fixed value, which is unfavorable to fast conver- gence of the algorithm. Thus this paper proposes an improved normalized LMS algorithm, which combines the Sigmoid variable step-size least mean square （SVSLMS） algorithm with thenormal- ized LMS algorithm,improves the convergence speed of normalized LMS algorithm. Simulation re- suits show that the performance of the proposed algorithm is better than that of normalized LMS algorithm,SVSLMS algorithm and ordinary LMS algorithm.