Multi-Channel ANC with Adaptive Kernel Assisted on-Line Secondary Path Modeling

Conventional online secondary path modeling (SPM) for multi-channel active noise control (ANC) is often computationally intensive and unstable, particularly in large-scale systems. To address this, we propose a novel hybrid ANC method that leverages an adaptive kernel for SPM. The approach models a small subset of secondary paths in real-time and uses an adaptive kernel, whose hyperparameters are optimized via gradient descent, to accurately interpolate the remainder. Simulations in a multi-loudspeaker, multi-microphone setup demonstrate that the proposed method achieves steady-state performance comparable to conventional full-path modeling while exhibiting significantly faster and more stable convergence performance in dynamic acoustic environments.