Electronic Theses and Dissertations

Date of Award

5-1-2026

Document Type

Thesis

Degree Name

M.S. in Engineering Science

First Advisor

Lei Cao

Second Advisor

Hamid Bahrami

Third Advisor

Ramanarayanan Viswanathan

School

University of Mississippi

Relational Format

dissertation/thesis

Abstract

Binary hypothesis testing is a fundamental problem in statistical signal processing and decision theory, with applications in communications, radar systems, and distributed sensor networks. Classical approaches, particularly the likelihood ratio test (LRT), provide optimal detection performance when the underlying probability distributions are known. However, in many practical scenarios, the resulting decision rules can become analytically complex or computationally intensive, especially in multi-sensor and dependent data settings.

This thesis investigates the use of neural networks for binary hypothesis testing under diverse statistical conditions. A data-driven framework is developed to learn decision rules directly from observed data, enabling the approximation of likelihood ratio–based detectors without explicit evaluation of probability density functions. The approach is evaluated across multiple statistical models, including Gaussian, exponential, bivariate Gaussian, copula-based dependent distributions, and mixed Gaussian models.

Simulation results show that neural network-based detectors closely approximate the optimal LRT for Gaussian distributions while maintaining consistent performance in more complex settings where analytical solutions are difficult. Performance is evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC) metrics across different sensor configurations. Overall, this work bridges classical detection theory and data-driven learning, providing a flexible framework for multi-sensor detection problems.

Additionally, an AI-based encoder–detector framework is proposed, where a neural network is used to learn optimal signal transformations under power constraints. The performance of the learned encoder is compared with the identity encoder, demonstrating improvements in detection performance by using the learned encoder.

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.