Signal Processing 154 (2019) 174–181
Contents lists available at ScienceDirect
Signal Processing
journal homepage: www.elsevier.com/locate/sigpro
Target detection exploiting covariance matrix structures in MIMO
radar
Jun Liu
a , b , ∗
, Jinwang Han
a
, Zi-Jing Zhang
a
, Jian Li
c
a
National Laboratory of Radar Signal Processing, Xidian University, Xi’an 710071, China
b
Department of Electronic Engineering and Information Science, University of Science and Technology of China, Hefei 230027, China
c
Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL 32611 USA
a r t i c l e i n f o
Article history:
Received 4 May 2018
Revised 17 July 2018
Accepted 18 July 2018
Available online 15 August 2018
Keywords:
Adaptive detection
MIMO radar
Adaptive matched filter
Persymmetry
Toeplitz structure
a b s t r a c t
We aim to address the problem of target detection in collocated multiple-input multiple-output (MIMO)
radar where the disturbance covariance matrix is unknown. In practice, the disturbance covariance ma-
trix has a persymmetric or Toeplitz structure, when the MIMO radar receiver is equipped with a sym-
metrically or uniformly spaced linear array. We exploit the persymmetric or Toeplitz structure of the
disturbance covariance matrix to design two adaptive detectors that do not require training data. Analyt-
ical expressions for the probability of false alarm and detection probability are derived for the proposed
detector that takes the persymmetry into account. Numerical examples are provided to show that the
proposed detectors outperform the conventional counterparts.
© 2018 Elsevier B.V. All rights reserved.
1. Introduction
Multiple-input multiple-output (MIMO) radar is currently an
active research area [1–6] . There are two main classes of MIMO
radar, based on their antenna configurations. The first one uses
widely distributed antennas, which is referred to as distributed or
statistical MIMO radar [7–10] . It employs spatial diversity to allevi-
ate the scintillations of radar cross sections (RCSs) of targets, bring-
ing in performance gains in several aspects, such as higher detec-
tion probability [11–13] and improved estimation accuracy [14,15] .
The other uses collocated antennas [16–20] , which is also called
coherent MIMO radar. MIMO radar with collocated antennas has
become a standard in diverse applications, including the automo-
bile radar industry [21] . It exploits waveform diversity to provide
higher spectral resolution with much lower hardware count [16] ,
design flexible spatial transmit beampattern [22,23] , improve de-
tection performance [24] , and strengthen interference rejection ca-
pability [25,26] .
Target detection is one of the primary functions of radar. It has
been extensively studied for distributed MIMO radar (see [8,11–
13,27–34]
, and the references therein). However, less work has
been conducted on target detection with collocated MIMO radar
[24,35–39] . We focus herein on target detection via collocated
∗
Corresponding author.
E-mail address: junliu@ustc.edu.cn (J. Liu).
MIMO radar, which for brevity is referred to as MIMO radar here-
after.
In [24] , target detection via MIMO radar is considered under
the assumption that the disturbance covariance matrix is known.
A clairvoyant detector is proposed to show that the exploitation of
orthogonal transmit signals leads to a gain in the target detection
performance. A general signal model for MIMO radar is provided
in [40] , where subspace detectors and a double subspace detector
are designed by using a set of homogeneous training data to esti-
mate the unknown disturbance covariance matrix. In [35] , the dis-
turbance is assumed to have compound-Gaussian distribution with
unknown covariance matrix. A two-step detector is developed in
[35] , where the unknown covariance matrix is estimated by us-
ing a set of homogeneous training data. In many practical scenar-
ios, it is difficult to collect a large number of independent identi-
cally distributed (IID) target-free training data, due to many factors
including variation in terrain and nonlinear array responses [41] .
The detection performance of the existing detectors is significantly
degraded when the number of training data is small. A Bayesian
approach was proposed in phased-array radar for detecting dis-
tributed targets without resorting to training data in [42] , where
the prior distribution of the covariance matrix is required.
Different from phased-array radar, MIMO radar makes it possi-
ble to directly use adaptive techniques [36] . Therefore, it is possi-
ble for MIMO radar to detect a target or estimate its parameters
without the training data or even range compression. In [25] , the
problem of parameter estimation for a point-like target is studied
https://doi.org/10.1016/j.sigpro.2018.07.013
0165-1684/© 2018 Elsevier B.V. All rights reserved.