Robust active noise control: an information theoretic learning approach

Abstract

Nonlinear active noise control (ANC) systems, which employ a nonlinear filter as the adaptive controller is not robust when the primary noise to be mitigated has a non-Gaussian distribution. The algorithm which updates the weights of the controller may even diverge for some higher magnitude primary noise signals. With an objective to improve the robustness of nonlinear ANC systems, a correntropy based nonlinear ANC system is developed in this paper. The proposed ANC scheme uses an information theoretic learning approach and has been shown to provide robust noise mitigation even for non-Gaussian primary noise signals.