A kernel recursive minimum error entropy adaptive filter

Abstract

The minimum error entropy, a currently useful alternative criterion, is widely adopted in the signal processing domain against impulsive noise. In this brief, we propose a novel algorithm to blend the advantages of both the kernel recursive least squares algorithm and the minimum error entropy criterion, called kernel recursive minimum error entropy algorithm. The proposed new algorithm achieves better recovery performance in predicting the Mackey–Glass time series, equalizing the nonlinear channel under heavy tailed alpha-stable environments and processing EEG data.