System-informed neural networks for broadband Direction-of-Arrival estimation

This study presents a high-resolution Direction-of-Arrival (DOA) estimation method designed for noise-dominant environments. By integrating physical insights into a machine learning framework, the proposed approach overcomes the limitations of existing methods, including high computational costs and performance degradation in extreme noise conditions. Utilizing a linear system model, incoming signals are expressed as combinations of basis vectors, with the typically complex-valued representation converted into the real domain to ensure compatibility with standard neural network architectures. The approach extends the linear system model from narrowband to broadband signals, further enhancing DOA estimation accuracy. This work offers a practical and efficient solution for DOA estimation in challenging sonar applications. [This work was supported by the Korea Research Institute for defense Technology planning and advancement (KRIT)—Grant funded by the Korea government (DAPA—Defense Acquisition Program Administration) (No. KRIT-CT-23-026, Integrated Underwater Surveillance Research Center for Adapting Future Technologies, 2025)]