Robust DOA Estimation Through Flexible Sparsity-Aware Modeling in Low SNR

Direction-of-arrival (DOA) estimation is crucial for underwater acoustic applications, such as target detection, localization, tracking and identification. Underwater array signal processing is characterized by its low signal-to-noise ratio (SNR). Conventional estimation methods and learning-based techniques both exhibit significant performance degradation in such scenarios. To improve performance at low SNR, inspired by recent advances in flexible sparsity-aware priors in Bayesian framework, we propose a hierarchical Bayesian probabilistic model through Generalized Hyperbolic (GH) distribution prior. Based on the probabilistic model, closed-form inference updates are derived by the variational inference (VI) technique. Numerical results demonstrate that our method achieves the superior performance, especially under the low SNR conditions.