Robust kernel recursive adaptive filtering algorithms based on M-estimate

Abstract

When coping with the large outliers in measurement caused by the non-Gaussian environmental noise, although the MCC criterion adopts the high-order statistics, the residual error for large outliers still exists. Considering that the M-estimate works well in minimum square error criterion and it can truncate the outliers and further improve the robustness, in this paper, we propose the robust kernel recursive least squares algorithms and the robust kernel recursive maximum correntropy algorithms based on three M-estimate methods. Then, numerical simulations verify that the M-estimates help the proposed algorithms have better performance than the conventional kernel recursive adaptive filtering against the non-Gaussian noise.