On the Effect of Face Detection on Heart Rate Estimation in Videoplethysmography

Videoplethysmography (VPG) is a rapidly developing optical sensing technique that provides an attractive alternative to traditional pulse oximetry methods. In its core, VPG relies heavily on face detection and tracking algorithms to extract a useful signal. Existing work mainly take advantage of ready to use solutions and treat face detection as a reliable given tool. In this study, we provide initial investigation of the effect of face detection frequency on spectral heart rate estimation. We collect the data using an off-the-shelf tablet and demonstrate how periodicity of face detection deteriorate the quality of the extracted VPG signal. Both frequency and time domain are considered. We provide important insights into this problem and reveal that under relatively stationary conditions, using static face region is more beneficial.