An Efficient DOA Estimation Method in Smart Antenna Array Using Attention based GRU-CNN Architecture and Mutual Information Based Enhanced NCA Algorithm

Accurate Direction of Arrival (DOA) estimation plays a crucial role in array signal processing, particularly in noisy and multipath environments where classical methods often underperform.This study proposes a novel deep learning-based framework that combines Convolutional Neural Networks (CNNs) and Attention-Based Gated Recurrent Units (AB-GRU) for robust spatial and temporal feature extraction.A bi-objective Improved Neighborhood Component Analysis (INCA) is introduced for feature selection, integrating mutual information with classification accuracy to identify the most discriminative features.Experimental results, conducted across a wide range of signalto-noise ratio (SNR) levels, demonstrate the superiority of the proposed model .The method achieves a minimum average RMSE of 0.19, significantly outperforming benchmark techniques.The model maintains high estimation accuracy even at low SNR levels and reduced snapshot counts, confirming its robustness and practical applicability.These results underscore the effectiveness of the proposed approach in enhancing DOA estimation under real-world constraints.