The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Sparsity based methods for target localization in multi-sensor radar
Author:	Kim, Haley H.
View Online:	njit-etd2017-015 (xi, 93 pages ~ 2.2 MB pdf)
Department:	Department of Electrical and Computer Engineering
Degree:	Doctor of Philosophy
Program:	Electrical Engineering
Document Type:	Dissertation
Advisory Committee:	Haimovich, Alexander (Committee chair) Simeone, Osvaldo (Committee member) Kliewer, Joerg (Committee member) Abdi, Ali (Committee member) Govoni, Mark A. (Committee member)
Date:	2017-01
Keywords:	Compressive sensing GMTI MIMO CFAR Interference cancellation Random arrays
Availability:	Unrestricted
Abstract:	In this dissertation, several sparsity-based methods for ground moving target indicator (GMTI) radar with multiple-input multiple-output (MIMO) random arrays are proposed. MIMO random arrays are large arrays that employ multiple transmitters and receivers, the positions of the transmitters and the receivers are randomly chosen. Since the resolution of the array depends on the size of the array, MIMO random arrays obtain a high resolution. However, since the positions of the sensors are randomly chosen, the array suffers from large sidelobes which may lead to an increased false alarm probability. The number of sensors of a MIMO random array required to maintain a certain level of peak sidelobes is studied. It is shown that the number of sensors scales with the logarithm of the array aperture, in contrast with a ULA where the number of elements scales linearly with the array aperture. The problem of sparse target detection given space-time observations from MIMO random arrays is presented. The observations are obtained in the presence of Gaussian colored noise of unknown covariance matrix, but for which secondary data is available for its estimation. To solve the detection problem two sparsity-based algorithms, the MP-STAP and the MBMP-STAP algorithms are proposed that utilizes knowledge of the upper bound on the number of targets. A constant false alarm rate (CFAR) sparsity based detector that does not utilize any information on the number of targets referred to as MP-CFAR and MBMP-CFAR are also developed. A performance analysis for the new CFAR detector is also derived, the metrics used to describe the performance of the detector are the probability of false alarm and the probability of detection. A grid refinement procedure is also proposed to eliminate the need for a dense grid which would increase the computational complexity significantly. Expressions for the computational complexity of the proposed CFAR detectors are derived. It is shown that the proposed CFAR detectors outperforms the popular adaptive beamformer at a modest increase in computational complexity.