Video-based physiological measurement using 3d central difference convolution attention network

Yu Zhao, Bochao Zou, Fan Yang, Lin Lu, Abdelkader Nasreddine Belkacem, Chao Chen

Research output: Chapter in Book/Report/Conference proceedingConference contribution

17 Citations (Scopus)

Abstract

Remote photoplethysmography (rPPG) is a non-contact method to measure physiological signals, such as heart rate (HR) and respiratory rate (RR), from facial videos. In this paper, we constructed a central difference convolutional attention network with Huber loss to perform more robust remote physiological signal measurements. The proposed method consists of two key parts:1) Using central difference convolution to enhance the spatiotemporal representation, which can capture rich physiological related temporal context by gathering time difference information 2) Using Huber loss as the loss function, the gradient can be smoothly reduced as the loss value between the rPPG and ground truth PPG signal is closer to the minimum. Through experiments on multiple public datasets and cross-dataset evaluation, the good performance and robustness of the rPPG measurement network based on central difference convolution are verified.

Original languageEnglish
Title of host publication2021 IEEE International Joint Conference on Biometrics, IJCB 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665437806
DOIs
Publication statusPublished - Aug 4 2021
Event2021 IEEE International Joint Conference on Biometrics, IJCB 2021 - Shenzhen, China
Duration: Aug 4 2021Aug 7 2021

Publication series

Name2021 IEEE International Joint Conference on Biometrics, IJCB 2021

Conference

Conference2021 IEEE International Joint Conference on Biometrics, IJCB 2021
Country/TerritoryChina
CityShenzhen
Period8/4/218/7/21

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Biomedical Engineering

Fingerprint

Dive into the research topics of 'Video-based physiological measurement using 3d central difference convolution attention network'. Together they form a unique fingerprint.

Cite this