Active Impulsive Noise Control With Adaptive Robust Kernels

Traditional active noise control (ANC) algorithms suffer significant performance degradation when addressing impulsive noise. Recent advancements in robust ANC algorithms have focused on the design and application of effective cost functions. A class of ANC algorithms based on the maximum correntropy criterion (MCC), which uses entropy as a robust adaptive cost function, has shown promising capability in handing heavy-tailed impulsive noise. However, the performance of these algorithms heavily relies on empirically manual parameter tuning, which poses challenges for practical applications. To address this, this paper proposes a filtered-x generalized adaptive robust function (FxGARF) algorithm. The robustness is quantified by a continuous parameter that can adaptively adjust based on the residual distribution. Additionally, to enhance the control precision of the algorithm, the scale parameter involved in the generalized robust kernel family is determined by the mean value of the noise signal filtered through Hampel filter. Thus, the proposed algorithm can adapt to noise characteristics, eliminating the need for manual kernel parameter tuning. Moreover, the mean convergence and mean-square performance of the FxGARF algorithm are analyzed. Finally, simulation results demonstrate that the proposed algorithm adaptively selects kernel parameters for impulsive noise of varying characteristics and achieves superior noise reduction performance compared to benchmark algorithms.