Time Lab's Approach to the Challenge on Computer Vision for Remote Physiological Measurement

Abstract

Computer vision for remote physiological measurement is novel and uniquely challenging task, which enables non-contact monitoring of the blood volume pulse (BVP) using a commonly accessible camera. This paper introduces Time Lab’s approach presented at the 2nd challenge on Remote Physiological Signal Sensing (RePSS) organized within ICCV2021. We propose an end-to-end rPPGNet for remote photoplethysmographyraphy (rPPG) signals estimation. A improved design of spatial-temporal map is also made, which is an efficient representation of the rPPG signal by removing most of the irrelevant background content. Furthermore, our approach achieved first place on the 2nd RePSS Challenge Track 1 and has outperformed the methods of other participants as we have achieved M_IBI = 117.25(4.51% improvement compared to the challenge top-2 result), R_HR = 0.62(8.77% improvement). The codes are publicly available at https://github.com/yuhang1070/2nd_RePSS_Track1_Top1_Solution.