Denoising Autoencoder-Driven Direction-of-Arrival Estimation with V-Shaped Coprime Antenna Array

Abstract:
High-resolution Direction of Arrival estimation is critical in radar-based target detection, particularly with limited snapshots. Our approach utilizes efficient computational techniques for precise 2D-paired-DOA estimation, employing a sparsely arranged V-shaped coprime array for its sensor element efficiency. To enhance reliability under low snapshot conditions, we integrate a Denoising Autoencoder to correct errors from noise and snapshot limitations. Subsequently, the DAE’s output is fed into an SS-MUSIC algorithm, yielding an impressive over 80 percent DOA estimation accuracy improvement. This method offers a robust solution for high-resolution DOA estimation in scenarios constrained by limited snapshots.