Predictive Directional Selective Fixed-Filter Active Noise Control for Moving Sources via a Convolutional Recurrent Neural Network

A team of researchers from Singapore, led by Boxiang Wang, has introduced Predictive Directional Selective Fixed-Filter Active Noise Control (PD-SFANC), a novel method that leverages a Convolutional Recurrent Neural Network (CRNN) to anticipate noise direction changes from moving sources. Traditional Directional Selective Fixed-Filter Active Noise Control (D-SFANC) selects pre-trained filters based on the current Direction-of-Arrival (DoA) of noise, but it fails to adapt quickly when the noise source moves, such as a passing vehicle or a drone. PD-SFANC addresses this by using a CRNN to capture the temporal dynamics of moving noise and predict future control filters, effectively canceling noise before it arrives at the listener.

Numerical simulations confirm that PD-SFANC significantly improves noise-tracking ability and dynamic noise-reduction performance compared to several representative ANC baselines. The method was tested across various movement scenarios, demonstrating its superiority in handling non-stationary noise. This breakthrough could revolutionize active noise cancellation in applications like smart headphones, automotive cabins, and industrial environments, where noise sources are often in motion. The paper is available on arXiv (2604.23144) and has been submitted to the Audio and Speech Processing category.