Publication

Abstract : Blood pressure (BP) is a primary vital sign that provides critical insights into an individual’s overall health and potential underlying conditions. Abnormal BP levels, whether elevated (hypertension) or lowered (hypotension), can indicate serious cardiovascular issues, metabolic disorders, or other systemic problems. Monitoring BP is thus essential for early detection and management of these conditions. Traditional contact-based BP measurement techniques like arm cuffs and arterial lines are accurate but obtrusive, limiting mobility and risking dermatological issues from prolonged skin contact. While smartwatches enable more convenient tracking, they still require direct skin contact which can be cumbersome and cause irritation over extended wear. There is a need for entirely non-contact, unobtrusive BP monitoring solutions to avoid constraints of contact-requiring methods. This research investigates the potential of using remote photoplethysmography (rPPG), independent component analysis (ICA), pulse transit time (PTT), and regression techniques to estimate BP values from facial video recordings. By analyzing subtle color variations in the face caused by changes in blood volume, rPPG allows contactless extraction of physiological signals like heart rate. Combining this with PTT calculated from the video can enable BP derivation via regression modeling. However, rPPG is highly susceptible to factors like ambient lighting conditions, motion artifacts, and camera distance/angle. To assess feasibility, 40 PTT readings were collected from 20 individuals aged 21−45 years where 2 recordings were captured per individual which lasted 10 seconds long. Testing on 6 of these individuals achieved 95.6% accuracy compared to a contact reference, indicating potential but also underlying challenges. The results suggest that more advanced rPPG extraction and processing techniques may be needed to obtain BP values robustly and accurately from facial videos, especially with subject...