Robust Censored Regression M-Estimate Normalized Subband Adaptive Filter: Formulation and Analysis

To tackle with highly correlated input signals and impulsive interferences under censored regression (CR) model, this brief presents a novel CR-type algorithm called robust censored regression M-estimate normalized subband adaptive filter (R-CR-M-NSAF), which can achieve significant performance improvement compared to other robust CR-type algorithms with tolerable complexity requirements. Furthermore, the mean and mean-square convergence ranges along with steady-state learning behavior have been analyzed with the help of some commonly-utilized assumptions. Finally, numerical simulations corroborate that the developed R-CR-M-NSAF algorithm receives better convergence behavior in terms of estimation accuracy in comparison with other considering algorithms under different background noise environments with impulsive interferences for parameter estimation and acoustic echo cancellation scenarios.