Evaluation of video-based rPPG in challenging environments: Artifact mitigation and network resilience

Nhi Nguyen ¹, Le Nguyen ¹, Honghan Li ²

⁰Center for Machine Vision and Signal Analysis (CMVS), University of Oulu, Oulu, Finland.

Computers in Biology and Medicine +

|

July 25, 2024

View abstract on PubMed

Summary

This study enhances remote photoplethysmography (rPPG) for reliable vital sign monitoring in real-world conditions. It introduces practical strategies to overcome video quality issues, improving accuracy for diverse applications.

Area Of Science

Biomedical Engineering
Signal Processing
Medical Imaging

Background

Video-based remote photoplethysmography (rPPG) offers non-contact vital sign monitoring but struggles with real-world artifacts.
Challenges include videocodec artifacts, low light, degradation, occlusions, and hardware/network constraints.

Purpose Of The Study

To systematically investigate real-world challenges affecting rPPG accuracy.
To propose and evaluate practical mitigation strategies for robust vital sign monitoring.

Main Methods

Comprehensive analysis of artifacts' detrimental effects on rPPG measurements.
Development of biosignal recovery, denoising, and inpainting techniques for video integrity.
Evaluation across diverse scenarios with varying data quality.

Main Results

Demonstrated improved accuracy across various rPPG methods under challenging conditions.
Validated the effectiveness of proposed strategies in enhancing rPPG measurement reliability.
Showcased enhanced generalizability for practical, diverse applications.

Conclusions

Proposed strategies significantly improve the dependability and resilience of video-based rPPG systems.
This work contributes to more reliable remote vital sign monitoring technologies.
The findings emphasize the importance of addressing data quality issues for practical rPPG deployment.

Keywords:

Artifact reduction Denoising techniques Inpainting techniques Network resilience Video-based remote photoplethysmography Vital sign monitoring