An affine-matrix tracking method for heart rate monitoring using imaging photoplethysmography

Imaging photoplethysmography (iPPG) can be obtained on face with only a mobile phone or webcam, which greatly reduces the discomfort of wearable devices. The noise of iPPG mainly comes from two sources. One is the light change. The other is the head movement. Many researchers have contributed a lot on how to remove motion noise and extract clean iPPG. In this paper, we propose an affine-matrix tracking method using Shi-Thomas corner points, Kanade-Lucas-Tomasi tracking algorithm and affine transformation matrix to update the region-of-interest (ROI). We also compare it with direct ROI-vertex tracking method. Experimental results demonstrate that the affine-matrix tracking method is more suitable for rotation movements while ROI-vertex tracking method fits more for panning movements.