一种改进的解相关LMS自适应算法

A Modified Uncorrelated LMS Adaptive Algorithm

针对变步长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的问题,提出了一种改进的解相关LMS自适应算法,该算法引入解相关原理和归一化处理,用输入向量的正交分量来更新滤波器权系数,有效加快了算法的收敛速度,且稳态误差小,使得算法在有色输入和大范围的动态输入下都能保持良好性能.

We discussed and analyzed the Variable Step-Size Adaptive Algorithm （VSSLMS）, in which the descent of its convergence speed led to the reduction of performance when the input signals were heavily correlated . As a result, we proposed a modified uncorrelated LMS adaptive algorithm. We substituted the orthogonal components of the input signals for input vectors to update coefficients of the adaptive filter. Due to the uncorrelation principle and the process of normalization, the convergence speed could be quickened and the misadjustment could be smaller, so the algorithm could achieve good performance even for colored inputs and with a large dynamic input range.