Identification et Classification d’un signal FM et GSM à l’aide du Machine Learning

Abstract

This thesis focuses on the automatic identification and classification of radio frequency signals, specifically FM (Frequency Modulation) and GSM (Global System for Mobile Communications) signals. The goal is to develop an efficient system capable of distinguishing between FM, GSM, and background noise using supervised Machine Learning techniques. To achieve this, signals were captured using an RTL-SDR (Realtek Software Defined Radio), a low-cost yet powerful device capable of receiving a wide range of frequencies from VHF to UHF. The acquisition and processing were conducted through Python scripts in Google Colab. After capturing the signals, several preprocessing steps were applied (filtering, decimation, segmentation) to extract relevant features for training machine learning models. Multiple supervised algorithms were tested. The results show a high classification accuracy, confirming the robustness of the approach. This work demonstrates the potential of combining software-defined radio with artificial intelligence techniques for practical applications such as spectrum monitoring, interference detection, and cognitive radio systems.