Machine Learning-Based Vehicle Classification Using Electromagnetic Leakage Signals

초록

Recent studies have reported classification tasks using electromagnetic (EM) leakage signals emitted from various electronic devices. However, there have been no attempts to utilize EM leakage signals from vehicles for classification. In this study, we simulated EM leakage signals emitted from the engines of three types of vehicles and conducted vehicle classification experiments using two machine learning algorithms: random forest (RF) and LightGBM. The experiments utilized fast Fourier transform (FFT) and short-time Fourier transform (STFT) as input features. The experimental results showed that STFT features outperformed FFT features in both models and achieved the highest accuracy of 84.7% when using STFT features with the RF model. This demonstrates the feasibility of vehicle classification based on EM leakage signals. © 2024 IEEE.

키워드