Camera-Augmented Non-Contact Vital Sign Monitoring in Real Time

• Published in: IEEE sensors journal 2022-06, Vol.22 (12), p.1-1
• Main Authors: Shokouhmand, Arash, Eckstrom, Samuel, Gholami, Behnood, Tavassolian, Negar
• Format: Article
• Language: English
• Subjects: Cameras; Continuous radiation; Elevation; Estimation; FMCW Radar; Heart rate; Monitoring; Non-Contact Monitoring; Optimization; Radar; Radar beams; Real-time systems; Real-Time Vital Sign Monitoring; Respiratory rate; RGB-D Camera; Sensor Fusion; Sensors; Telemedicine; Torso
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2022.3172559

This study develops a camera-guided frequency-modulated continuous-wave (FMCW) radar to monitor vital signs. A red-green-blue-depth (RGB-D) camera estimates the human torso landmarks and a processing unit constantly adapts the radar beams to the direction of the subjects. To constantly optimize the regions of interest for monitoring respiratory rate (RR) and heart rate (HR), a novel method, coined "singular value-based point detection (SVPD)," is designed. Vital sign extraction is then followed as the last step. Experiments are conducted for the cases of single-subject (10 subjects, 31 scenarios, and 1550 repetitions) and dual-subject monitoring (6 subjects, 6 scenarios, and 90 repetitions). Average (RR, HR) accuracies of (97.68%, 85.88%), (90.02%, 86.05%), (96.71%, 89.50%), and (97.52%, 86.71%) are achieved for the range of distances (0.5-2.5 m), azimuth angles (0°-30°), elevation angles (-30°-+30°), and incident angles (-30°-+30°), respectively. The higher chest and upper abdomen are determined as the optimal regions for RR and HR estimation respectively, with average accuracies of 98.31% and 86.93%. Finally, the capability of dual-subject monitoring at various inter-subject distances (range of 20-70 cm) is confirmed with average accuracies of 92.26% and 73.23% for RR and HR respectively.