Mohamed Othman Defense
August 22, 2014
UNIVERSITY OF UTAH
ELECTRICAL AND COMPUTER ENGINEERING DEPARTMENT
DISSERTATION DEFENSE FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Mohamed Abou Bakr Othman
Advisor: Behrouz Farhang
Detection of Nondeterministic Linear Chirps in non-Gaussian Noise Background
Chirp signals arise in many applications of digital signal processing. In this dissertation, we address the problem of detection of chirp signals that are encountered in a bistatic radar which we are developing for remote sensing of cosmic ray induced air showers. The received echoes from the air showers are characterized by their large Doppler shift (several tens of MHz), and very short sweep period (~10s). This makes our astrophysical problem a challenging one, since a very short sweep period is equivalent to a very low energy chirp signal. Furthermore, the related parameters of the received echoes are nondeterministic since they are tied to the physical parameters of the air showers that are stochastic in nature. In addition, our problem is characterized by the rarity of the expected chirp-echoes to be received, few events per week, and thus, background noise reception is the case most of the time. The primary focus of this research is to address these challenges and find an optimized detection approach under the existing receiver environment which contains non-Gaussian noise and characterized by low signal-to-noise ratio (SNR).
Matched filters are commonly used in radar systems when the chirp signal is known. In our first method, we revisit this context and use a “matched-filter” as a basis of building a rake-like receiver which consists of a set of filters matched to quantized chirp rates, logarithmically distributed within the chirp-rate interval of interest. We examine the detection capability of the proposed structure through extensive theoretical and numerical analysis. Theoretical analysis and simulation results prove that the proposed detector has high detection capability for a range of chirp slopes in a low SNR environment.
A major source of false-alarms found to be due to sudden noise spikes that cover wide frequency bands. These transient signals have high amplitudes and occur at random time instants. This leads to erroneous detection decision. We study the influence of amplitude limiting the noisy signal on reducing the received false alarms and enhancing the detection performance of the proposed rake-like receiver.
In our second method, we use Hough transform (HT), which is widely used in the area of image processing for the purpose of finding parameterized patterns, as a basis of building a robust detection technique. We examine the detection capability of the proposed structure through theoretical and numerical analysis. Our results prove that the proposed detector has high detection capability for a range of chirp slopes in a low SNR environment.
The introduced detection algorithms are implemented over a Virtex-5 FPGA. National Instruments modules are used as a high-performance custom hardware. Due to rarity of received echoes, we emulate the expected radar echoes to evaluate the system performance. The detection performance of the emulated echoes is examined using the implemented receiver at the field. Also, we compare the performance of both detectors.
Wednesday, August 27, 2014
Chemical Engineering Conference Room
Merrill Engineering Building (MEB) Room 3291
The public is invited