TY  - JOUR
T1  - Class Imbalance in UWB-Based Backward Driving Detection
AU  - Lee, Kyungbo 
AU  - Ko, Young-Bae 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.11.1683
KW  - UWB CIR
KW  - backward driving
KW  - detection
KW  - class imbalance
KW  - DSR
KW  - 
KW  - deep learning
AB  - Backward driving in highway Hi-Pass lanes poses serious safety risks due to limited space and sudden maneuvers. This study proposes a deep learning framework that classifies forward and backward driving using Ultra-Wideband (UWB) Channel Impulse Response (CIR) data. Experiments conducted in a real tunnel environment show that a narrow segment of the CIR sequence was found to contain the most discriminative features for direction detection. To address this limitation, the concept of the Discriminative Signal Ratio (DSR) and analyze the impact of class imbalance. CNN, LSTM, and Transformer models are compared under varying class ratios, and results show that oversampling the minority class at a 1:6 ratio yields optimal performance. The findings demonstrate the feasibility of CIR-based direction detection and provide insights for future deployment in intelligent transportation systems.