Author: Mavlonbek Khomidov, Jong-Ha Lee 👨🔬
Affiliation: Department of Biomedical Engineering, Keimyung University, Department of Computer Engineering, Keimyung University 🌍
Purpose: In this research, we aim to estimate blood pressure using remote photoplethysmography (rPPG) signal extracted from facial video. This method provides non-invasive and contactless, continuous blood pressure monitoring without using traditional cuffs or physical sensors. By analyzing subtle changes in facial skin color caused by blood flow, this approach aims to provide an affordable, convenient solution for real-time blood pressure monitoring.
Methods: We collected 27 face video recordings using ordinary web camera captures at 30 frames per second and as a reference device we used an automatic blood pressure monitor (OMRON HEM-790IT). We selected five regions of interest on the face — forehead, right cheek, left cheek, nose, and chin — to extract rPPG signals for blood pressure estimation. From the video frames, we analyzed the average intensity of the green channel and the HSV (hue, saturation, value) components. The green channel is commonly chosen for its higher signal-to-noise ratio in rPPG applications, while the HSV color space provides additional insight into color variations by separating luminance from chromatic components. After extracting these features, we applied a deep neural network to learn the complex relationships between the obtained signals and actual BP measurements.
Results: Our proposed method demonstrated promising results, achieving an overall accuracy of 92% in estimating blood pressure. In particular, the mean absolute error (MAE) for systolic BP was 6.4 mmHg, while the MAE for diastolic BP was 5.6 mmHg.
Conclusion: This study highlights the effectiveness of a camera-based algorithm for monitoring blood pressure. Using a non-contact approach, the proposed method eliminates the need for traditional cuff or wearable devices that require physical contact, making it particularly useful for people who experience discomfort, skin sensitivity, or irritation when using such devices. Additionally, the proposed method reduces the risk of cross-contamination and prevents the spread of infectious diseases.