一种稳定的变步长块LMS自适应算法

Stable Variable Step-Size Block LMS Adaptive Algorithm

针对LMS自适应滤波算法在输入信号高度相关时,收敛速度下降导致性能下降,本文从基本的块LMS算法开始,简要介绍了块LMS算法的实现方法,在此基础上重点分析了在变步长块LMS算法中,影响步长因子的要素。提出了一种新的变步长因子迭代算法(SVBLMS),该迭代算法充分考虑输入信号和误差信号对变步长因子的影响,并且迭代的结构简单,计算量小。通过Matlab仿真,仿真结果表明,该迭代算法较其它块LMS算法有更快的收敛速度,更稳定的收敛过程。当输入为有色信号或输入噪声较大时,本算法都能保持良好的性能。

This paper mainly focus on solving the problem of performance degradation of the LMS self-adaptation algorithm as the input signals highly correlated. This paper briefly introduces the realization method of Block LMS algorithm ever since the basic Block LMS self-adaptation algorithm and mainly analyzes the factors which influence step factor. A relatively new variable step size block LMS （SVBLMS） algorithm is proposed. This algorithm has fully considered the effect of input signals and error signals on Variable step factors and its iterative structure is quite simple, Matlab, the result shows that this iterative algo convergence process, compared with other block only with low computational complexity. Simulation with rithm has more rapid convergence speed and more stable LMS algorithm. When input signals are colored signals or input noise level is higher, this algorithm still can keep good performances.