The normalized subband adaptive filter under variable kernel width maximum correntropy criterion for sparse system

This paper presents a novel subband adaptive filter (SAF) based on the maximum correntropy criterion (MCC) with correntropy-induced matric (CIM) penalty for sparse system, which is called CIM-MCCSAF. The CIM-MCCSAF is obtained by incorporating the correntropy-induced matric penalty into MCCSAF. It outperforms MCCSAF in sparse system identification. While occurring the impulsive interfere, the step size of CIM-MCCSAF is sharply dropped. Thereby, the CIM-MCCSAF has good robustness. However, the CIM-MCCSAF still have a tradeoff between steady-state error and the converging speed. Therefore, we proposes a band-dependent variable kernel width CIM-MCCSAF (VKW-CIM-MCCSAF) algorithm where the kernel width is obtained by maximizing the subband scale function. Finally, simulations verify the merit of the VKW-CIM-MCCSAF and CIM-MCCSAF.