Identification and Spectrum Occupancy Analysis of 4G/5G Signals Using Deep Learning Techniques

This paper introduces a deep learning framework for the classification and spectrum occupancy analysis of 4 G and 5 G signals based on spectrograms as input data. A spectrogram, that visually shows the spectrum of frequencies contained in a signal over time, is a good input for the identification of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4 \mathrm{G} / 5 \mathrm{G}$</tex> signals. The model is based on the modified U-Net architecture and trained on a set of labeled spectrograms of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4 \mathrm{G} / 5 \mathrm{G}$</tex> resource channels such as PDSCH, PDCCH, PSS, SSS, and others are mapped. The goal of this project is to give a more accurate determination of spectral occupancy for <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$4 \mathrm{G} / 5 \mathrm{G}$</tex> signals. Existing methods classify spectrograms, and our approach uses labeled spectrograms during training to enhance the model's ability to classify signals and more importantly, map spectrum usage more effectively. The paper details the data preprocessing steps, model architecture, training process, and evaluation metrics, demonstrating the model's capability in improving spectrum occupancy analysis and signal identification.