JLAMCNet: Joint Loss and Array Mismatch Compensation Network for Single Snapshot DoA Estimation

The article addresses the critical challenges in single-snapshot Direction of Arrival (DoA) estimation, a fundamental problem in array signal processing with wide-ranging applications. Traditional methods often struggle with insufficient data and array mismatches, leading to inaccurate DoA estimates. The paper introduces JLAMCNet, a novel network designed to overcome these limitations by incorporating a comprehensive array mismatch model that accounts for various imperfections, including timing errors, polarization mismatches, and pattern gain errors. This model provides a more accurate representation of real-world received signals, enhancing the network's ability to handle complex electromagnetic environments. One of the key innovations presented is the Direction-Aware Contrastive Loss (DACL) function, which maintains feature consistency under varying signal-to-noise ratio (SNR) conditions, thereby improving the discriminative power of the feature space. The dynamic loss weight adjustment mechanism ensures that the network adapts to different conditions, optimizing performance across various scenarios. Additionally, the Antenna-Aware Spatial Attention (AASA) mechanism dynamically adjusts the weights of feature channels for each antenna, compensating for array mismatches and enhancing feature extraction. The article also includes extensive simulation experiments that demonstrate JLAMCNet's superior performance compared to existing methods, showcasing its robustness and accuracy in DoA estimation under imperfect array conditions. The results highlight JLAMCNet's ability to achieve significant improvements in multi-DoA target estimation accuracy and its resilience in extreme mismatch scenarios, such as antenna failures. Overall, the paper provides a detailed exploration of advanced techniques for enhancing DoA estimation, offering valuable insights for researchers and practitioners in the field of array signal processing.