In-Vehicle Seat Occupancy Detection Using Ultra-Wideband Radar Sensors

In this paper, we propose a method for identifying the seats in which people are sitting using ultra-wideband radar sensor data. First, we install two radar sensors in locations suitable for monitoring people inside the vehicle and then collect reflected radar signals. Based on the acquired radar sensor data, we design a convolutional neural network (CNN)-based classifier for seat occupancy detection. As input data for training the CNN, the radar signals accumulated over time acquired from two radar sensors are synthesized in parallel and used. As a result, our proposed method can determine the seats occupied by people with more than 93% accuracy.