BSENSE: In-vehicle Child Detection and Vital Sign Monitoring with a Single mmWave Radar and Synthetic Reflectors
Mingyue Tang, Pranshu Teckchandani, Jizheng He, Hongzhou Guo, Elahe Soltanaghai
Abstract
Recent regulations on monitoring infants and children in vehicle cabins have spurred interest in using Millimeter-wave (mmWave) radars due to their reliability in various lighting conditions and privacy benefits. However, existing radar-based vital sign detection solutions fail in car settings with abundant occlusions or closely-seated multi-person scenarios. To resolve these limitations, we introduce BSENSE, a joint occupancy and vital sign monitoring system using a single radar that is robust to occlusion and varying seating arrangements and number of occupants in vehicle cabins. BSENSE incorporates synthetic wireless reflectors positioned in car corners to redirect radar signals toward blind spots, enabling Non-Line-of-Sight (NLoS) vital sign detection while maintaining sensing performance in Line-of-Sight (LoS) areas. The proposed system employs a hybrid architecture combining signal processing and a deep learning pipeline that can detect the car seating layout and jointly learn occupied seats and signatures of breathing to distinguish adults from children and infants, and monitor their vital signs over time. Our extensive evaluations with 120,000 radar data points, 400 different experimental scenarios, a mix of 10 adults, 5 children of age 1--11, and two programmable infant and child simulators demonstrate BSENSE's capability in child detection with over 97% accuracy and estimating their breathing rate within 6 BPM error, even in multi-person and NLoS scenarios, and across different car models.