SPARSE DOA ESTIMATION BASED ON A DEEP UNFOLDED NETWORK FOR MIMO RADAR

Recently, deep learning has gained increasing popularity in array signal processing. In this paper, we estimate the direction of arrival (DOA) for the multiple-input and multiple-output (MIMO) radar system based on deep learning. First, we convert DOA estimation into a linear inverse problem with spatial sparsity, and construct a neural network based on the iterative shrinkage thresholding algorithm (ISTA) to improve the interpretability of the network. Then, a stacked denoising autoencoder (DAE) is employed to achieve data-driven denoising, which improves the anti-jamming ability of DOA estimation. Finally, a new deep unfolded network named denoising learned ISTA (Denoising-LISTA) is proposed for DOA estimation. Simulation results illustrate that the proposed method improves the robustness of DOA estimation with single snapshot sampling and keeps significant predominance in beam sharpening and sidelobe suppression.