Phase-shifted remote photoplethysmography for estimating heart rate and blood pressure from facial video

Human health can be critically affected by cardiovascular diseases, such as hypertension, arrhythmias, and stroke. Heart rate and blood pressure are important physiological information for cardiovascular monitoring, which have recently been advanced by camera-based remote photoplethysmography (rPPG). Deep learning-based heart rate estimation models have achieved substantial progress by learning periodic information from videos. However, blood pressure estimation is challenging to restore detailed rPPG waveform morphology under motion and illumination artifacts, requiring additional physiological information. To address these limitations, this paper proposes a two-stage deep learning framework for estimating heart rate and blood pressure from facial video. The proposed algorithm consists of a dual remote photoplethysmography network (DRP-Net) and bounded blood pressure network (BBP-Net). DRP-Net infers rPPG signals at acral and facial sites, and these phase-shifted rPPG signals are utilized to estimate heart rate. BBP-Net integrates temporal features and analyzes phase discrepancy between the acral and facial rPPG signals to estimate systolic blood pressure and diastolic blood pressure. Moreover, We augmented facial videos in temporal aspects by utilizing a frame interpolation model to increase bradycardia and tachycardia data. Experiments were conducted on MMSE-HR and V4V datasets to demonstrate the effectiveness of the proposed method. Our method achieved the state-of-the-art performance for estimating heart rate and blood pressure with significant margins compared to previous methods. Our code is available at https://github.com/gyutaehwang/phase_shifted_rppg.