Direction-finding based on the theory of super-resolution in sparse recovery algorithms

The problem of recovering directions-of-arrival in the sparse signal model with multiple snapshots is considered. Based on the theory of super resolution, multiple snapshots are used to jointly estimate directions-of-arrival in the continuous domain. Instead of uniformly discretizing the search range, interpolation preprocessing on the estimated super-resolution directions is suggested leading to a sparse convex optimization formulation. Moreover, a first order iterative algorithm is employed to reduce the computational time. A good selection of regularization parameter is guaranteed via the modified generalized cross validation (GCV). Numerical results demonstrate the performance of the proposed methods.