Deep Unfolding Network for Near-Field Source Localization via Symmetric Nested Array

We exploit the deep unfolding networks (DUN) to achieve localization of near-field (NF) sources via symmetric nested array (SNA). Firstly, we calculate the vectorization of fourth-order cumulant (FOC) matrix which is only related to direction of arrival (DOA). Then, the vectorization of FOC matrix is considered as the input of DUN to obtain DOA spatial spectrum. Next, we calculate the range vector with the estimated DOA. Finally, the range vector is considered as the input of DUN to estimate the range parameter of NF sources. Simulation demonstrate that the DUN is robust to off-grid parameter estimation and can improve the resolution of DOA spatial spectrum. Moreover, the DUN has faster convergence speed and higher estimation accuracy.