Distribution-Aware, Risk-Sensitive (DA-RS-FxNLMS) Active Noise Control for Non-Gaussian Acoustic Environments

Active noise control (ANC) in real-world acoustic environments frequently faces impulsive and heavy-tailed noise disturbances, which degrade the performance significantly and lead to slow convergence. This work proposes a dynamically adaptive distribution-aware risk-sensitive filtered-x normalized least mean square (DA-RS-FxNLMS) method for efficient ANC under a non-Gaussian and impulsive scenario. The proposed ANC framework employs a correntropy-based risk-sensitive exponential cost function, which incorporates higher-order statistics and adapts to the error distribution. Further, an adaptive and dynamically adjusted kernel width mechanism tracks the time-varying noise characteristics. The normalized filtered-x structure provides stability under secondary path uncertainty. Simulation is carried out by applying α-stable noise to create an impulsive noise environment, which is produced by the Chambers-Mallows-Stuck method. The outcomes of the proposed method are compared with the baseline methods, showing that the proposed method achieves a noise reduction of 7.02 dB, a significant 40% faster convergence, and improved robustness under strong impulsive noise conditions with α = 1.2. The outcome confirms that the proposed method efficiently delivers a promising solution for the ANC system. To the best of our knowledge, for the first time, a unified ANC framework integrates distribution-aware, risk-sensitive learning, adaptive correntropy kernel estimation, and filtered-x normalization for non-Gaussian acoustic environments.