A Variable Step-Size NLMS Algorithm with Adaptive Coefficient Vector Reusing

In adaptive filtering, there is usually a trade-off between the speed of convergence and the accuracy of the learning procedure. Recently, variable step-size algorithms and coefficient vector reusing schemes were proposed to solve this trade-off. This paper presents a new adaptive filtering algorithm that combines both strategies to achieve fast convergence speed and low steady-state misadjustment simultaneously. In the proposed algorithm, the error signal is used to dynamically adjust the step-size and the reusing order in each iteration. Simulation results demonstrate better performance of the proposed algorithm when compared to previously proposed approaches.