【摘要】 在分析传统LMS算法及其改进算法的基础上,提出了一种新的可变步长LMS算法,新算法采用梯度矢量的平均值的平方值调整步长.计算机仿真表明,与传统LMS算法和文献[1]中的VSS-LMS算法相比,新算法有更快的收敛和跟踪速度,并能收敛到更小且更稳定的均方误差.更多还原

【Abstract】 Following a brief discussion on traditional LMS and its improved algorithm ,a new LMS algorithm of variable step size is proposed. The new algorithm make use of the square of average grads to produce the new variable step size. Computer simulations demonstrate that the proposed algorithm accomplishes faster convergence ,tracing speed and steady smaller MSE than the traditional LMS and VSS-LMS algorithm proposed in literature.更多还原