Two-dimensional multi-snapshot Newtonized orthogonal matching pursuit for DOA estimation

Abstract

Estimating the azimuth and elevation angles of targets is referred to as two-dimensional direction of arrival (2D-DOA) estimation problem, which is of vital importance in array signal processing and multiple input multiple output (MIMO) millimeter wave communication fields. Inspired by the Newtonized orthogonal matching pursuit (NOMP) for line spectrum estimation (LSE), this paper proposes the two-dimensional multi-snapshot NOMP (2D-MNOMP) to deal with the 2D-DOA estimation. Specifically, two-dimensional fast Fourier transform (FFT) is utilized to significantly reduce the computation complexity, and a Newton refinement step and feedback strategy are proposed to improve performance of DOA estimation. Based on the generalized likelihood ratio test (GLRT), the stopping criterion is established. The near-optimal performance of 2D-MNOMP is also demonstrated by comparing against other methods and the Cramér-Rao bound (CRB), both in terms of simulations and real data.