Off-Grid DOA Estimation for Block Orthogonal Matching Pursuit

In the direction-of-arrival (DOA) estimation problem with multiple incident signals and multiple snapshots, an off-grid Block Orthogonal Matching Pursuit (OG-BOMP) algorithm is proposed to address the performance instability of the Orthogonal Matching Pursuit (OMP) algorithm. Under multiple snapshots conditions, the vectors of signal sources to be estimated from different snapshots exhibit joint sparsity. By converting the multiple measurement vectors (MMV) model into a block single measurement vector (SMV) model, this joint sparsity can be fully exploited. After further considering off-grid errors, more accurate off-grid parameters are obtained, leading to more precise estimation results. Simulation results demonstrate that the proposed method significantly improves estimation accuracy compared to traditional OMP and BOMP algorithms under multiple snapshots and exhibits strong multi-target resolution capability.

I. Introduction

Direction of arrival (DOA) estimation is a primary means for spatial localization of targets or signal sources and an important research direction in fields such as communication, radar, reconnaissance, and electronic countermeasures [1], [2]. Over the past few decades, maximum likelihood-based (ML) DOA estimation algorithms and subspace-based estimation algorithms have become relatively mature, but they each have their limitations [3]. For example, algorithms such as Multiple Signal Classification (MUSIC) experiences significant performance degradation or even failure under low SNR and few snapshot conditions [4]. Maximum likelihood methods perform better than subspace methods in scenarios with a known count of sources, small samples, and low SNR [5], but their computational complexity is high, and some simplified methods rely on high-precision DOA estimates as a prerequisite, making the practical application difficult [6].