Detecting Pulse from Head Motions Using Smartphone Camera

This paper proposes a touchless system for estimating heart rates using smartphone cameras. It is based on computing subtle and periodic head motions by tracking facial feature points in camera images. However, it suffers from the problem with the shaking movements of the hand that holds the smartphone. To resolve the problem, the proposed system tracks background feature points together with facial feature points. This is based on that the movements of the background feature points are mainly caused by hand shaking. The movements of background feature points are then used to correct locations of the facial feature points. In experiments, the correction improved the accuracy of heart rate estimation by more than 15%. As a result, the proposed system could estimate heart rates with high accuracy.