Enhancing Remote-PPG Pulse Extraction in Disturbance Scenarios Utilizing Spectral Characteristics

Kai Zhou, Simon Krause, Timon Blocher, Wilhelm Stork; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 280-281

In recent years, several approaches for remote Photoplethysmography (rPPG) have been proposed, and the recently proposed methods have achieved substantial improvement in measurement accuracy. However, none of the methods has investigated the possibility of using the spectral characteristics for the design of rPPG signal extraction algorithms. In this paper, we propose a new rPPG measurement method which exploits the spectral characteristics of rPPG signals. We validated the freshly proposed method on a benchmark dataset including seven scenarios and 26 participants. The results of the validation experiment demonstrates the feasibility to use spectral characteristics to extract rPPG signal. By combining with the constraint plane, the new proposed method provides better overall performance.