Reproduce “Directional Selective Fixed-Filter Active Noise Control Based on a Convolutional Neural Network in Reverberant Environments” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Boxiang Wang, Zhengding Luo, Haowen Li, Dongyuan Shi, Junwei Ji, Ziyi Yang, Woon-Seng Gan
Researchers from Nanyang Technological University and Northwestern Polytechnical University developed a directional Selective Fixed-Filter Active Noise Control system that integrates a lightweight Convolutional Neural Network for robust Direction-of-Arrival estimation in reverberant environments. This system demonstrates improved noise reduction and faster response compared to adaptive algorithms, and avoids the noise amplification observed in non-directional fixed-filter methods.
Key insights
- Deep learning, specifically a lightweight CNN, can accurately estimate the Direction-of-Arrival (DoA) of noise sources in complex, reverberant acoustic environments, a significant challenge for traditional methods.
- Incorporating dynamic DoA information through intelligent filter selection dramatically improves the robustness and noise reduction performance of fixed-filter ANC systems when noise source directions vary.
- A multi-task learning approach for azimuth and elevation classification allows the CNN to learn highly generalizable spatial features from multi-channel audio, even with a low parameter count.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.