Détection d’activité spectrale à l’aide de RTL-SDR et de l’IA

Abstract

Modern information processing systems have undergone significant evolution driven by digital transformation and the integration of intelligent technologies into data analysis. Physical signals are now converted into digital data to be processed using advanced mathematical algorithms, enabling the extraction of relevant information. Artificial intelligence plays a central role in improving classification, analysis, and automated decision-making performance. These technologies are widely applied across diverse fields such as telecommunications, surveillance, and complex data analysis, offering a deeper understanding of dynamic environments and enhanced system performance optimization. In this work, we developed an experimental approach based on the use of the RTL-SDR receiver for the acquisition of real radio frequency signals, followed by preprocessing and modeling using supervised learning algorithms including Random Forest, K-Nearest Neighbors (KNN), and XGBoost. The experimental results demonstrated a robust spectral activity detection capability, validated through confusion matrices and a comparative performance analysis of the models. Furthermore, a web application was designed to make spectral activity detection accessible and interactive. It incorporates a real-time radio signal capture interface, a CSV file import and analysis feature, and a dashboard for visualizing analysis results. This platform represents a practical contribution, providing a flexible tool for spectral analysis and detection in complex radio environments.