ESPRIT matching pursuit algorithm for DOA estimation with single snapshot

Abstract

For a uniform linear array, classical DOA estimation methods, such as ESPRIT etc., with a space smoothing preprocessing, can be applied to the case of single snapshot. Due to the spatial sparsity, this DOA estimation problem can be translated into a sparse signal recovery problem. Then we can use some sparse recovery methods to resolve DOAs. As a highly efficient method among those, the orthogonal matching pursuit (OMP) algorithm is considered at first. It has been observed that OMP can offer DOA estimates with smaller mean-square errors than ESPRIT at low SNRs; however, an opposite conclusion will be drawn at high SNRs. Combining the advantages of the both methods, we present a novel ESPRIT matching pursuit (EMP) algorithm, which takes a full utilization of ESPRIT and a maximum correlation method in a process of greedy pursuit. Simulation results indicate that the presented method outperforms both OMP and ESPRIT at different SNRs.