Graduation Semester and Year
Summer 2026
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Electrical Engineering
Department
Electrical Engineering
First Advisor
Dr. Qilian Liang
Second Advisor
Dr. Yijing Xie
Third Advisor
Dr. Chenyun Pan
Abstract
The rapid evolution of wireless communication imposes stringent requirements for ultra-reliable, low-latency transmission in dynamic, interference-prone environments. Traditional model-driven signal processing struggles to adapt to nonlinear hardware effects, time-varying channels, and complex interference patterns. Deep learning (DL) offers a transformative, data-driven alternative, enabling end-to-end optimization and robust adaptation under uncertain propagation conditions.
This dissertation investigates deep learning architectures for intelligent and resilient wireless communication through three complementary contributions. The first introduces a Vision Transformer (ViT)-based modulation classification framework that leverages self-attention to capture local and global dependencies in spectrogram representations of Quadrature Amplitude Modulation (QAM) signals. The ViT achieves superior classification accuracy across signal-to-noise ratios (SNRs) and remains robust under Rician fading and Doppler shifts, outperforming convolutional neural network (CNN) and support vector machine (SVM) baselines, particularly for high-order modulations; computational complexity analysis confirms practical viability.
The second contribution proposes a Variational Autoencoder (VAE)-based generative framework for probabilistic signal reconstruction and classification over Rician fading channels. The VAE learns compact, channel-aware latent embeddings that encode modulation semantics and mitigate distortion. Bit error rate (BER) and cosine similarity analyses show performance approaching theoretical limits and enhanced low-SNR robustness relative to classical detection schemes and CNN baselines, with robustness further evaluated under line-of-sight variation, Doppler shift, and latent dimensionality changes.
The final contribution introduces latent diffusion for automatic modulation classification (LD-AMC), adapting the latent diffusion architecture to a discriminative role. A VAE trained with auxiliary classification, center, and inter-class separation losses compresses each pilot-equalized I/Q frame into a class-structured latent vector, and a class-conditioned denoising diffusion probabilistic model (DDPM) is trained in this latent space. At inference, the diffusion model acts as a bank of per-class denoisers: each latent is refined under every class hypothesis, and the refinement pattern augments the raw latent as a discriminative feature. Evaluation covers accuracy across M-QAM orders, comparisons with VAE, DDPM, and CNN baselines, robustness and ablation studies, and complexity analysis.
Together, these contributions integrate discriminative and generative learning, providing practical solutions to channel and interference challenges and a foundation for intelligent, self-optimizing communication systems.
Keywords
Deep Learning, Wireless Communications, Automatic Modulation Classification (AMC), Vision Transformer (ViT), Variational Autoencoder (VAE), Denoising Diffusion Probailistic Model (DDPM), Latent Diffusion, Signal-to-Noise Ratio (SNR) Robustness, M-QAM, Rician Fading Channels
Disciplines
Signal Processing | Systems and Communications
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Kulkarni, Swarada Ajit, "Deep Learning for Wireless Communications" (2026). Electrical Engineering Dissertations. 3.
https://mavmatrix.uta.edu/electricaleng_dissertations2/3