Frequency-Domain Adaptive Filter Algorithm with Switching Step-Size

Frequency-domain adaptive filter (FDAF) algorithms have been widely used in many fields in virtue of its fast convergence and low computational complexity. However, FDAF algorithms with fixed step-size cannot balance convergence rate and steady-state misadjustment. In this paper, we propose the switching step-size based FDAF (SSS-FDAF) algorithm that selects the optimal step-size at each iteration for the FDAF update by comparing the mean-square deviation (MSD) trends with different step-sizes, to obtain fast convergence and low steady-state misadjustment at the same time. Furthermore, a novel reset strategy is designed for guaranteeing the tracking capability of the proposed algorithm. Computer simulations of colored signals and real-world signals have demonstrated the effectiveness of our algorithm.