Optimal facial areas for webcam-based photoplethysmography

Resumo

For long-term pulse tracking, it is possible to use the photoplethysmography method, the essence of which is estimating the dynamics of optical properties of examined organ tissues. Since this method is not invasive and does not need complicated equipment, it can be used for long-term heart rate monitoring in a wide range of operating conditions. In several publications, it is shown that different variations of this method can be used for estimating the heart rate by means of a web camera shooting a face. Image analysis procedures are used for estimating the time dynamics of optical properties of local fragments of the epidermis. In the present work, we examine a problem on determining facial areas most suitable for pulse detecting. The ViolaJones algorithm is used for detecting facial areas. Coordinates of several regions where the color signal is measured are calculated with respect to the detected face. The color signal characterizes variation of color components for images of the epidermis. For estimating whether the image area is optimal, the corresponding criterion that characterizes the degree of frequency manifestation in the color signal spectrum close to the real value of heart rate is formulated. According to this criterion, the optimal regions are determined and they are as follows: the area near the nose and the area on the nose between the eyes.