Two-dimensional subband adaptive filters

The well-known two-dimensional least-mean-square (LMS) adaptive filter has been widely used in many applications. However, it suffers from slow convergence for correlated input signals. To address this problem, this paper introduces the normalized subband adaptive filter (NSAF) into two-dimensional adaptive filtering and then develops a two-dimensional subband adaptive filter (TD-SAF). Moreover, in order to further improve the performance of the TD-SAF, a two-dimensional variable step-size subband adaptive filter (TD-VSS-SAF) is proposed. It is derived by minimizing the sum of the subband mean-square errors in all subbands at each iteration. Simulation results in the two-dimensional system identification context show that the proposed TD-VSS-SAF can obtain both fast convergence rate and small steady-state misalignment.