Optimal source selection for image photoplethysmography

This paper presents an optimal selection on the region of interest (ROI) and color spaces to extract photoplethysmography signals from facial videos. The study is carried out under two sections; ROI selection and color space selection. We compared five different ROI selection for different regions of the face. We also investigated three color spectrums namely additive, perceptual and orthogonal color spaces. The study experimented on the publicly available human-computer interaction (HCI) database. When evaluated on 30 subjects, the results showed that the ROI selection on the forehead and the green spectrum of the additive color space to provides higher accuracy for heart rate measurement.