Electronic Theses and Dissertations

Date of Award

1-1-2019

Document Type

Thesis

Degree Name

M.S. in Engineering Science

First Advisor

Ramanarayanan Viswanathan

Second Advisor

Lei Cao

Third Advisor

John Daigle

School

University of Mississippi

Relational Format

dissertation/thesis

Abstract

Signal detection in cognitive radio involves the determination of presence or absence of a primary user signal so that the secondary user may opportunistically gain access when the spectrum is unoccupied. In decentralized sensing scheme, two or more secondary users sense the spectrum, process individual observation and then pass the quantized data to a fusion center where a decision with regard to which hypothesis being true, that is, a signal being present or absent, is made. In the second part of the thesis, we study Bayes error performance of two-sensor tandem network designed to detect the presence or absence of deterministic signals in correlated Gaussian noise. Hence, the correlation coefficient remains identical under both hypotheses. Specifically, we address the question of which sensor ought to serve as the fusion center for optimal detection performance. In the process of this query, we draw some inference parallel to the “Good, Bad and Ugly’’ signal regions formulated originally for the two-sensor one-bit-per-sensor parallel fusion network by Willet,et.al. In the tandem “Good” region, numerical results conclusively show that the strategy of placing better sensor, i.e the sensor with higher signal to noise ratio, serving as the fusion center is preferred for better detection performance. In the first part of thesis, we study the error performance in a parallel network consisting of two sensors. In the parallel configuration, each sensor quantizes it's own observation into a single-bit and transmits them to the fusion center. At the fusion center, the performance of AND and OR rules are examined by assuming the observations at the two sensors are jointly Gaussian, with specific means, variances and correlation coefficient, under hypothesis H1, whereas the observations under H0 are still Gaussian with specific means and variances but are statistically independent. The optimum quantizers at each sensor are found by minimizing the probability of error at the fusion center. We use a genetic algorithm (GA) to find a sub-optimal solution. It was observed that, when prior probabilities of hypotheses are equal, AND performs at least as well as OR.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.