Improving Coverage of Linear Antenna Arrays in Crowded Wireless Channelsunder Low SNR Conditions

Accurate Angle of Arrival (AOA) estimation based on linear antenna array isvital in several applications such as radar, sonar, and wireless communications.However, operating in practical conditions faces significant challenges such as lowSignal-to-Noise Ratio (SNR), dense environments, and limited practical Field ofView (FOV) of linear arrays. These issues restrict the performance of conventionalAOA estimation methods, especially when signal sources are positionedoutside the practical FOV near the array end-fires. This paper introduced anAOA estimation approach that addresses these challenges together through atwo-stage pre-processing unit integrated with a sparse-based Direction Finding(DF) estimator. The first stage uses Oversampling and Averaging (OSA) processfor enhancing the effective SNR and improving estimation performance underlow-SNR conditions. The second stage applies Array Steering (AST) process, aphase shift for steering AOAs that exist outside the limited linear-array FOV intoits practical range, for enabling accurate estimates near the array end-fires. Integratingthis pre-processing unit with the sparse-based DF estimator expands thepractical linear-array FOV while maintaining robust performance in dense environmentsand low SNR scenarios. Extensive theoretical analysis and MATLABsimulations demonstrate that the proposed approach significantly outperformsconventional sparse-based estimators, achieving higher Probability of Correct Detection (PCD) and improving estimation performance. The performance of theconventional estimator gradually improves with increasing SNR, it only reaches acomparable level to the proposed method at much higher SNR values. In contrast,the proposed approach consistently maintains its superior performance across theentire tested SNR range,-20:20 dB , ensuring robust and accurate estimation evenunder low SNR conditions. At SNR= -15 dB, the conventional sparse-based estimatorachieves a PCD of 0.81, whereas the proposed approach achieves a superior PCD of 0.907.