ORCID Identifier(s)


Graduation Semester and Year




Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering


Electrical Engineering

First Advisor

Saibun Tjuatja


Embedded in each observation of a RADAR system is an amalgam of scattered signals. An important function of RADAR signal processing is to resolve or separate this mixture of target signals. Imaging using RADAR data is a well-established area of research. Conventional RADAR imaging methods use a two-dimensional Fourier transform to back-project RADAR measurements to scattering sources [1]. The Fourier transform based imaging techniques have resolution limited by the bandwidth and spatial diversity of available data samples. Super-resolution methods such as MUSIC overcome the resolution limitation by employing an alternate model for the measured data [2]. These techniques are capable of enhancing RADAR images through an increase in the resolvability of scattering produced at particular spatial locations. This research focuses on enhancing RADAR imaging techniques through a concept of isolating and localizing scatterers from within a mixture. The isolation step is accomplished through the notion of separating signals based on their non-Gaussianity. Finite sized radar targets have non-Gaussian probability densities which enables the use of this measure in the distinction of scatterers [3]. The second attractive feature of non-Gaussianity as a measure of distinction is that a mixture of sources with non-Gaussian probabilities tends to Gaussian as the number of independent sources increases. The approach of this manuscript uses a kurtosis maximization algorithm to search for the most non-Gaussian components of a mixture. Once these elements are isolated, an attempt is made to generate an accurate localization of the individual scatterers through a composite maximum view of the components. This research employs simulated and measured RADAR data. The measured RADAR data was captured on the turntable ISAR system at the Wave Scattering Research center of the University of Texas at Arlington. Results show that there is a strong correlation between leptokurtic scattering data and the unique scattering centers in the radar field of view. The results of the testing, on synthetic target sets, yielded a 52% probability of detection of the known scatterers with a localization error of zero in most cases.


RADAR, Scattering, Statistical signal processing


Electrical and Computer Engineering | Engineering


Degree granted by The University of Texas at Arlington