The Random Data-Reusing GMCC Adaptive Filtering Algorithm for System Identification Under Impulsive Noise Environments

Abstract

The generalized maximum correntropy criterion (GMCC) has been widely applied for robust adaptive filtering (AF) algorithm. The gradient-based GMCC (GB-GMCC) algorithm realizes good filtering performance for system identification under impulsive noise environments. However, the highly colored input signal can damage the convergence rate of GB-GMCC. Therefore, based on the data-reusing method, we propose a robust AF algorithm, called as data-reusing GMCC (DR-GMCC) algorithm, which uses the information of the latest $K$ input data to remedy the convergence limitation of GB-GMCC. In addition, to enhance the filtering performance of DR-GMCC, we use a random strategy to select the past $K$ input data leading to a new algorithm, named as random DR-GMCC (RDR-GMCC). Furthermore, for RDR-GMCC, we also analyze the mean-square convergence and computational complexity. Compared with existing algorithms, simulation results verify that RDR-GMCC achieves better filtering accuracy and faster convergence rate.