Multi-Source 2D-AOA Estimation via Deep Learning

In recent years, deep learning (DL) methods have received considerable attention for one-dimensional (1D) angle of arrival (AOA) estimation, yet fewer studies have explored their application to two-dimensional (2D) AOA estimation. This paper presents a novel, low-complexity DL-based approach designed to independently estimate both the elevation and azimuth AOAs of sources using an L-shaped array, along with their pairing. Our proposed 2D-AOA estimator scheme is validated through numerical simulations, showcasing its superior performance compared to existing 1D and 2D DL methods. Notably, our method approaches the Ziv-Zakai bound closely, especially in scenarios with low signal-to-noise ratio and low-angle differences, even amidst multiple highly correlated signals. Furthermore, complexity analysis confirms the superiority of our proposed method.