A Novel Off-Grid Model and Fast Iterative Method for DOA Estimation With Arbitrary Arrays

High accuracy and low computational complexity are always contradictory for any grid-based DOA estimators. To overcome such difficulty, we establish a novel off-gird model based on approximated linear representation of continuous steering vector space, and propose an estimator based on grid location and adaptive step-size iteration (GLASI). The model is available for arbitrary array geometry as well as real-world arrays with imperfection, and allows optimal estimation through effective but compact approximation. Either MUSIC-type or sparse-based algorithm can be applied for coarse grid location, which enables GLASI to work regardless of source coherence and the number of snapshots. Simulation results verify GLASI enjoys remarkable performance superiority over its on-grid counterparts, with low grid density and thereby high computational efficiency.