Periodic update based estimates of subband system noise powers for subband adaptive filtering

摘要

Subband adaptive filters (SAFs) have been widely used due to their fast convergence rate for colored input signals. In order to meet the conflicting requirements of fast convergence rate and small misalignment of SAFs, a variable step-size matrix (VSSM) has been presented for the recently proposed normalized SAF (NSAF). The VSSM is derived by letting the powers of the subband a posteriori errors of the NSAF equal the powers of the corresponding subband system noises at each time instant and therefore the subband system noise powers should be estimated. In this paper, we propose to incorporate the concept of periodic update into the subband system noise power estimates to reduce the computational complexity of the VSSM-NSAF. Simulation results in a system identification context show that the proposed periodic update based method can achieve a good performance close to that of the full update one.