Deep Learning-Driven Classification of 5G and LTE Signals for Next-Generation Wireless Networks

Cellular Radio Frequency (RF) spectrum monitoring and analysis are crucial for identifying and mitigating interfering and disruptive RF signals. In this study, we use Deep Learning (DL) models, including Convolutional Neural Networks (CNNs) and Transformer Networks (TNs), to classify and identify these signals. The primary objective is to classify cellular RF signals to aid in mitigating interference, thereby enhancing security and quality of service (QoS). Two models, ResNet50 [1] and ViT [2], are evaluated for their accuracy in classifying 5th generation New Radio (5G NR), 4th generation long-term evolution (4G LTE), and combined LTE-NR cellular signals. Both models demonstrate high true positive rates, exceeding $\mathbf{9 5 \%}$ across all classes. However, ViT consistently outperforms ResNet50, showcasing superior capability in capturing distinguishing features and more accurately classifying RF signals. Notably, ViT’s perfect true positive rates for the LTE and NR classes underscore its robustness and potential in classification and signal identification tasks, however ViT is much more resource intensive.