Kernel Normalized Least Mean High Order Error Algorithm for Sub-Gaussian Noise

Robust kernel adaptive filtering (KAF) has been extensively studied to identify nonlinear system corrupted by the heavy-tailed impulsive noise, whereas few KAF algorithm is reported to concern on the light-tailed sub-Gaussian noise. This paper proposes a new kernel normalized least mean high order error (KNLMHOE) algorithm in the presence of sub-Gaussian noise. More importantly, the convergence condition on the mean stability of KNLMHOE is also provided. The KNLMHOE algorithm is simple to implement and exhibits significant performance improvement compared with several state-of-the-art KAF algorithms, which is confirmed by simulations.