基于模糊推理的变步长LMS自适应滤波算法

Variable Step Size LMS Adaptive Filtering Algorithm Based on Fuzzy Inference

LMS算法是一种基于最速下降法的最小均方误差自适应滤波算法。为了提高LMS 算法的收敛速度,依据模糊控制原理,推导出一种结构简单的步长与误差的非线性函数关系, 进而得出一种新的变步长LMS自适应滤波算法(FVSLMS),该算法结构简单,易于实现。在理论上,根据万能逼近定理,用FVSLMS算法可以以任意精度逼近步长与误差的非线性函数关系,因此它可以作为以误差调节步长的变步长LMS算法的一类统一形式。最后,通过计算机仿真说明了FVSLMS算法具有较好的收敛性能。

In order to upgrade convergence rate of the LMS adaptive filtering algorithm, a variable step size LMS adaptive filtering algorithm based on fuzzy inference （FVSLMS） is proposed. The FVSLMS is derived from fuzzy control theory. It has a simple function structure, with a non-linear functional relationship between step size and error. By the universal approximation theorem, the functional relationship between step size and error could be given by FVSLMS in arbitrary accuracy. So FVSLMS could be a uniform algorithm for variable step size LMS adaptive filtering algorithms. The simulation shows that FVSLMS has very good convergence property.