Off-Grid DOA Estimation for Colocated MIMO Radar via Reduced-Complexity Sparse Bayesian Learning

Abstract

Recent advance on signal processing has witnessed increasing interest in machine learning. In this paper, we revisit the problem of direction-of-arrival (DOA) estimation for colocated multiple-input multiple-output (MIMO) radar from the perspective of machine learning. The reduced-complexity transformation is first applied on the array data from matched filters, thus eliminating the redundancy of the array data for the relief of calculational burden. Furthermore, the pre-whitening is followed to obtain a simplified noise model. Finally, the DOA estimation is linked to off-grid sparse Bayesian learning (OGSBL), which does not require to update the noise hyper-parameter, and a block hyper-parameter is utilized to accelerate the convergence of the OGSBL algorithm. The proposed estimator provides better DOA estimation accuracy than the existing peak searching algorithm. The effectiveness of the proposed algorithm is verified via numerical simulation. INDEX TERMS Array signal processing, MIMO radar, DOA estimation, off-grid, sparse Bayesian learning.

Sparse Principal Component Analysis via Fractional Function Regularity

“…to relax the l 0 -norm in compressed sensing [15,16,23,24] to get the sparsity signal, where a > 0 is a parameter and…”

“…. , r), y * j is the same as in formula (24), by using the iterative formula (40), (β * j ) n+1 is obtained. If B is fixed, according to (21), it is only try to minimize n i�1 ‖x i − AB ⊤ x i ‖ 2 � ‖X − XBA ⊤ ‖ 2 with respect to A, where A ⊤ A � I r .…”

Fast Angle Estimation and Sensor Self-Calibration in Bistatic MIMO Radar With Gain-Phase Errors and Spatially Colored Noise

“…Till now, various estimation strategies have been proposed for MIMO radars, such as multiple signal classification (MUSIC) [5], Capon, subspace fitting, estimating signal parameters via rotational invariance technique (ESPRIT) [6], least squares (LS) [7], maximum likelihood (ML) [8], tensor approaches [9]- [12], optimization-aware algorithms [13]- [15]. With respect to tensor approaches, there are two main tensor decomposition frameworks, called higher-order singular value decomposition (HOSVD) and parallel factor analysis (PARAFAC) decomposition.…”

“…In [19], a reduceddimensional MUSIC method has been given. It is suitable for arbitrary sensor manifold, but it needs exhaustive spectrum grid search, which means it is computationally inefficient and cannot avoid the drawback of off-grid problem [8]. In [20], an ESPRIT-like algorithm has been proposed, which is capable to provide closed-form solution for joint DOD and DOA estimation.…”

Target Localization in Bistatic EMVS-MIMO Radar Using Tensor Subspace Method

“…Generally, target localization in a bistatic MIMO radar involves the estimation of directionof-departure (DOD) and direction-of-arrival (DOA). So far, a lot of excellent estimation algorithms have emerged, such as multiple signal classification (MUSIC) [8], estimation of signal parameters via rotational invariance techniques (ESPRIT) [9]- [11], propagator method (PM) [12], [13], maximum likelihood (ML) [14]- [17], higher order singular value decomposition (HOSVD) [18]- [20] and parallel factor (PARAFAC) [21]- [25]. Generally, most of the existing algorithms are transferred from the traditional spectrum estimation algorithms.…”