Device Agnostic Measurement of Blood Pressure From RPPG Signals

Recent advances in sensing technologies enable the measurement of Blood Pressure (BP) on a smart phone by analyzing the patterns in a remote Photoplethysmograph (rPPG). In this paper, we propose a device agnostic method to non-invasively measure BP using rPPG signals. We achieve state-of-the-art results on 1276 rPPG samples collected from 18 different smartphone models by training a sequence model on the longitudinal features extracted from the signal. These features are invariant to changes in the sampling rate and are closely related to the underlying physiological processes. The results conform to the criteria established by the American and British measurement standards for BP measurement in certain ranges. Clinical validation of the algorithm on 111 rPPG signals using 4 different smartphones resulted in an average RMS error of 17.79 mmHg and 10.83 mmHg for systolic and diastolic BP measurements, respectively.