Robust Constrained Normalized M-Estimate Subband Adaptive Filter: Algorithm Derivation and Performance Analysis

This brief presents a constrained normalized M-estimate subband adaptive filter (CNMSAF) algorithm that is robust to impulsive noise, which utilizes subband decomposition technique to whiten the colored input signals thereby achieving fast convergence speed and low computational complexity. In addition, the stability and theoretical mean square deviation performance of the algorithm are analyzed. In order to solve the constrained filtering problem with sparsity, the L1 norm of the filter weight vector is used as an additional constraint, and we also propose the L1-CNMSAF algorithm. Computer simulations verify the accuracy of the theoretical analysis and the effectiveness of the proposed algorithms.