Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Non-Contact Health Monitoring During Daily Personal Care Routines

Xulin Ma, Jiankai Tang, Zhang Jiang, Songqin Cheng, Yuanchun Shi, Dong LI, Xin Liu, Daniel McDuff, Xiaojing Liu, Yuntao Wang

TL;DR

This work tackles non-contact health monitoring during daily personal care where lighting variations and occlusions hinder standard rPPG. It introduces LADH, the first long-term multi-modal (RGB+IR) rPPG dataset with ground-truth PPG, SpO2, and RR collected from 21 participants over 10 days across five activities. The authors propose FusionPhysNet, a modality-fusion two-stream network with a joint HR–SpO2–RR loss $\text{Loss} = \text{MSE}_{\text{BVP}} + \text{MSE}_{\text{RR}} + 0.002 \times \text{MSE}_{\text{SpO2}} \times (100 - \text{Mean}(\text{SpO2}))$. Results show that RGB+IR fusion improves HR accuracy (e.g., MAE 4.99 BPM in day-wise training) and that multi-task learning enhances SpO2 and RR estimates, establishing a robust benchmark for real-world non-contact vitals monitoring.

Abstract

Remote photoplethysmography (rPPG) enables non-contact, continuous monitoring of physiological signals and offers a practical alternative to traditional health sensing methods. Although rPPG is promising for daily health monitoring, its application in long-term personal care scenarios, such as mirror-facing routines in high-altitude environments, remains challenging due to ambient lighting variations, frequent occlusions from hand movements, and dynamic facial postures. To address these challenges, we present LADH (Long-term Altitude Daily Health), the first long-term rPPG dataset containing 240 synchronized RGB and infrared (IR) facial videos from 21 participants across five common personal care scenarios, along with ground-truth PPG, respiration, and blood oxygen signals. Our experiments demonstrate that combining RGB and IR video inputs improves the accuracy and robustness of non-contact physiological monitoring, achieving a mean absolute error (MAE) of 4.99 BPM in heart rate estimation. Furthermore, we find that multi-task learning enhances performance across multiple physiological indicators simultaneously. Dataset and code are open at https://github.com/McJackTang/FusionVitals.

Non-Contact Health Monitoring During Daily Personal Care Routines

TL;DR

This work tackles non-contact health monitoring during daily personal care where lighting variations and occlusions hinder standard rPPG. It introduces LADH, the first long-term multi-modal (RGB+IR) rPPG dataset with ground-truth PPG, SpO2, and RR collected from 21 participants over 10 days across five activities. The authors propose FusionPhysNet, a modality-fusion two-stream network with a joint HR–SpO2–RR loss . Results show that RGB+IR fusion improves HR accuracy (e.g., MAE 4.99 BPM in day-wise training) and that multi-task learning enhances SpO2 and RR estimates, establishing a robust benchmark for real-world non-contact vitals monitoring.

Abstract

Remote photoplethysmography (rPPG) enables non-contact, continuous monitoring of physiological signals and offers a practical alternative to traditional health sensing methods. Although rPPG is promising for daily health monitoring, its application in long-term personal care scenarios, such as mirror-facing routines in high-altitude environments, remains challenging due to ambient lighting variations, frequent occlusions from hand movements, and dynamic facial postures. To address these challenges, we present LADH (Long-term Altitude Daily Health), the first long-term rPPG dataset containing 240 synchronized RGB and infrared (IR) facial videos from 21 participants across five common personal care scenarios, along with ground-truth PPG, respiration, and blood oxygen signals. Our experiments demonstrate that combining RGB and IR video inputs improves the accuracy and robustness of non-contact physiological monitoring, achieving a mean absolute error (MAE) of 4.99 BPM in heart rate estimation. Furthermore, we find that multi-task learning enhances performance across multiple physiological indicators simultaneously. Dataset and code are open at https://github.com/McJackTang/FusionVitals.

Paper Structure

This paper contains 13 sections, 1 equation, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Physiological Sensing in Daily Personal Care Scenarios
  4. Physiological Sensing Datasets
  5. Dataset
  6. Data Collection
  7. Method
  8. Input embedding
  9. Neural Network Model
  10. Joint-Training
  11. Results and Findings
  12. Limitations
  13. Conclusion

Figures (3)

  • Figure 1: The experimental setup of data collection while participants are brushing teeth.
  • Figure 2: A visual illustration of our daily data collection protocol. Participants have different activities across states.
  • Figure 3: FusionPhys Model with Input frames of facial RGB and facial IR. PPG, RR and SpO2 estimation tasks are trained simultaneously with a combined loss.