Super-resolution direction finding with far-separated subarrays using virtual array elements

In this study, a novel two-stage virtual array element construction procedure is proposed for super-resolution direction finding of multiple narrowband coherent point sources with a sparse array consisting of two or more far-separated subarrays. An all-pole model is used to fit to the array snapshot and the corresponding model parameter estimation methods are discussed. Virtual elements between and outside the subarrays are then constructed in sequence using different approaches, namely, the semi-parametric and non-parametric techniques, to increase the effective array aperture size. As a result, a great improvement in angular resolution is achieved. Compared with the existent procedure using minimum weighted norm (MWN) only, the proposed procedure is superior in sidelobe artefact reduction and weaker adjacent source detection as a result of its inherent low sidelobe level. Numerical simulations also demonstrate that under lower signal-to-noise ratio or with fewer snapshots, the proposed procedure has better performance than the one using MWN only in both resolution and accuracy, as well as the conventional estimation of signal parameters via rotational invariance techniques in accuracy, although such advantages tend eventually to disappear with reduced target angular spacing.