Noncontact sensing of heart rate variability from facial video using the topology algorithm

This study proposes a noncontact, accurate sensing method for heartbeat R-R intervals and heart rate variability from the video of a target person’s face. The proposed method comprises three steps: (1) automatic face recognition and tracking, (2) the detection of skin-tone pixels in the face region, and (3) heart rate estimation using the topology algorithm. We applied the proposed method to facial video recordings of 55 participants and evaluated its accuracy. The results demonstrate that the root-mean-square error in estimating the heart rate variability was 2.7 ms using the proposed method, demonstrating the effectiveness of the proposed approach.