Electronic Theses and Dissertations

Date

2025

Document Type

Thesis

Degree Name

Master of Science

Department

Electrical & Computer Engineering

Committee Chair

Eddie Jacobs

Committee Member

Aaron Robinson

Committee Member

Madhusudhanan Balasubramanian

Abstract

The application of neural networks has demonstrated significant potential in the field of acoustic source localization. This thesis investigates the deployment of such advanced neural network models for the specific challenge of uncrewed aerial vehicle (UAV) localization. A persistent obstacle in training deep learning models for this purpose, however, is the considerable effort required to collect and curate sufficiently large real-world datasets, which is both time consuming and resource-intensive. To address this limitation, this research incorporates synthetic data augmentation using a synthetic source sound generation script and the pyroadacoustics simulation environment. These synthetic sources, when processed through the pyroadacoustics simulator, can provide a diverse and physically meaningful set of training examples. This thesis then systematically evaluates whether the integration of such synthetic data into the training process enhances the localization performance of the neural network, potentially reducing the burden of real-world data collection while maintaining, or even improving, model accuracy and generalizability.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest.

Notes

Open Access

Share

COinS