Remote measurement of heart rate from facial video in different scenarios

Abstract

Vital signs, such as heart rate (HR), can be remotely measured from facial video recorded by a consumer-level digital camera. However, the accuracy of heart rate estimation is typically affected by ambient light and body motions. Hence, we propose an anti-disturbance method that integrates Eulerian video magnification (EVM), signal quality assessment (QA), and adaptive chirp model decomposition (ACMD) to measure HR in the distance. Next, we evaluate and validate the performance of the proposed method in five different scenarios, including low-illumination, normal-illumination, high-illumination, unbalanced-illumination, and head-motion. The experimental results demonstrated that the estimations of HR using the proposed method in different scenarios have a high consistency with the corresponding ground truths. Moreover, comparing with the methods based on empirical mode decomposition (EMD) or variable mode decomposition (VMD) algorithms, the proposed method can provide a robust solution to eliminate disturbance and obtain a precise HR measurement from the facial videos in different scenarios.