Abstract

As modern ML techniques have become increasingly advanced, they have begun to be integrated into wireless RF systems for classification, identification, and spectrum management. Deep Neural Networks (DNNs) enable RF system operators and designers to design more flexible systems with greater robustness to errors and attacks. However, neural networks require significant amounts of properly annotated data to train. Current data labeling methods lack the ability to obtain reliable true labels for circuit properties such as carrier frequency offset (CFO), power amplifier (PA) non-linearity, and in-phase/quadrature (IQ) imbalance. This thesis investigates the efficacy of a novel architecture, RF-Diffusion, for generating high-quality synthetic data conditioned on circuit-specific parameters. Using mathematical models of circuit properties, a dataset of synthetic examples was created to test the architecture’s ability to learn and implement these parameters.

The metrics of Error Vector Magnitude (EVM), Adjacent Channel Leakage Ratio (ACLR), and Frechet distance were used to evaluate performance. It was found that the architecture effectively learns these parameters and can reproduce high-quality synthetic samples. These findings indicate that this type of generative model could be used to supplement real datasets with synthetic examples with labels that would otherwise be impossible to obtain.

Committee Chair

Roger Chamberlain

Committee Members

Michael Brent,Yingying Fan

Degree

Master of Science (MS)

Author's Department

Computer Science & Engineering

Author's School

McKelvey School of Engineering

Document Type

Thesis

Date of Award

Spring 5-2026

Language

English (en)

Download

Share

COinS