基于梯度向量的变步长LMS算法

A New Improved Variable-step-size LMS Algorithm Based on the Gradient Vector

传统的最小均方误差(Least Mean Square)算法难以同时获得较快的收敛速度和较小的稳态误差。在对传统LMS算法、变步长LMS算法及其改进算法分析的基础上,通过建立步长参数与梯度向量之间的一种新的非线性函数关系,提出一种改进的变步长LMS算法。分析和仿真表明,改进后的算法收敛速度更快,均方误差更小;同时也具有良好的抗噪性能。

The traditional Least Mean Square （LMS） algorithm has difficulty in gaining the fast convergence speed and low steady state error at the same time. Based on a brief analysis of traditional LMS, variable-step-size LMS algorithm and its improved algorithm, establishing a new nonlinear functional relationship between the step-size and the gradient veetor,a new improved variable-step-size LMS algorithm is proposed. The theoretical analysis and computer simulation prove that this algorithm converges faster, with smaller mean square error and good anti-noise performance.