A neural reference projection-based method for multi-reference active noise control (L)

In this Letter, a neural reference projection-based method is proposed for compressing multichannel reference signals in feedforward active noise control (ANC). The method is integrated with the filtered-x affine projection algorithm to form a neural reference projection-based filtered-x affine projection algorithm, thereby improving convergence speed and noise reduction performance. Using a real-vehicle road noise dataset, simulation results demonstrate that, in multi-reference ANC, faster convergence and higher broadband noise attenuation are achieved compared with conventional approaches.