Robust DOA Estimation Based on Subarray Learning Selection
We propose a learning-based DOA framework that selects subarray elements dynamically to address array distortion and path-dependent loss. A lightweight classifier is trained to activate subarrays based on spatial features, and a physicsinformed loss improves generalization. Simulations using Bellhop show accuracy gains over classical baselines in low SNR and distortion settings.
Authors
- Yudong Wang (ORCID: https://orcid.org/0000-0002-0156-1294)
- Guiling Wu (ORCID: https://orcid.org/0000-0002-4334-7704)
- Xinhao Cai (ORCID: https://orcid.org/0009-0009-0467-6597)
- Ruifeng Liu (ORCID: https://orcid.org/0000-0002-3063-7673)
- Cheng Li (ORCID: https://orcid.org/0000-0003-1963-8234)
- Liang Qian (ORCID: https://orcid.org/0009-0000-7831-3785)
Institutions
- Shanghai Jiao Tong University (CN)
- Changhong (China) (CN)
Publication Details
- Published
- 2025-07-12
- DOI
- https://doi.org/10.1109/icsip65915.2025.11171575
- Primary Topic
- Image and Signal Denoising Methods
- Type
- article
- Field-Weighted Citation Impact
- 0.00