Manifold Separation-Based DOA Estimation for Nonlinear Arrays via Compressed Super-Resolution of Positive Sources

Manifold separation technique plays an important role in array modeling and signal processing for arbitrary arrays. By utilizing this technique, the atomic norm minimization (ANM) methods can be extended to the nonlinear arrays for DOA estimation via the generalized line spectral estimation approach. However, such an approach results in a high computational burden for the large-scale arrays. In this paper, a low-dimensional semidefinite programming (SDP) implementation to the coarray manifold separation atomic norm minimization (CMS-ANM) is proposed for a class of nonlinear arrays based on the compressed super-resolution of positive sources. The theoretical guarantee of the proposed SDP implementation for the exact recovery is presented, and the CMS-ANM-based low-complexity DOA estimation method is developed. The simulation results validate the theoretical analysis and demonstrate the satisfying trade-off for the performance and complexity.