A band-dependent variable step-size sign subband adaptive filter

• We propose a band-dependent variable step-size sign subband adaptive filter using mean-square deviation minimization.
• By assigning the different step size in each band, the performance of the sign subband adaptive filter is improved.
• The proposed algorithm includes the estimation method of the measurement-noise variance for an impulsive-noise environment.
• The proposed algorithm includes the reset algorithm for a system change suddenly.
• The proposed algorithm performs better than the existing algorithms in aspects of the convergence rate and the steady-state error.

This letter proposes a band-dependent variable step-size sign subband adaptive filter using the concept of mean-square deviation (MSD) minimization. Since it is difficult to obtain the value of the MSD accurately, the proposed step size is derived by minimizing the upper bound of the conditional MSD with given input. By assigning the different step size in each band, the filter performance can be improved. Moreover, we suggest the estimation method of the measurement-noise variance in an impulsive-noise environment, because the proposed algorithm needs the measurement-noise variance to calculate the step size. The reset algorithm is also applied for maintaining the filter performance when a system change occurs suddenly. Simulation results demonstrate that the proposed algorithm performs better than the existing algorithms in aspects of the convergence rate and the steady-state estimation error.