一种新的变步长变换域自适应滤波算法

A New Variable Step-size Transform Domain Adaptive Filtering Algorithm

变换域LMS算法能通过正交变换有效降低输入信号自相关矩阵特征值的分散程度,可提高算法的收敛速度;变步长LMS算法可以克服固定步长因子所导致的算法在较快收敛速度和较小稳态误差之间存在的矛盾,从而获得较快的收敛速度和较好的收敛结果。将二者相结合,提出了一种新的变步长变换域自适应滤波算法。计算机仿真结果表明该算法具有更快的收敛速度和更小的稳态误差,并且运算量较少,具有良好的实用性能。

The transform domain LMS algorithm can reduce the cross-correlation of input signals effectively through orthogonal transforms, so the convergence rate will be improved;the variable step-size LMS algorithm can overcome the conflict between high convergence rate and low steady-state error which is caused by fixed step-size, so higher convergence rate and better convergence result can be acquired. The transform domain LMS algorithm is integrated with the variable step-size LMS algorithm, from which a new variable step-size transform domain adaptive filtering algorithm is presented. The computer simulation results indicate that the new algorithm has higher convergence rate, lower steady-state error and less computation, so it has favorable practical performance.