Privacy-Preserving Digital Health Biomarkers for Elderly People
Overview
The project’s main goal is to develop an autonomous camera system (Heart Vision) that employs facial recognition to monitor the heart rate of elderly individuals in a non-intrusive and privacy compliant manner. Using a Machine Learning framework named pyVHR, we can determine a human heart rate from video input alone via a technology known as remote photoplethysmography (rPPG). At any given moment, our facial regions undergo a variety of color changes as a result of blood flow which are imperceptible to the human eye – but not to a camera! By analyzing this series of color changes over time, this cloud-based solution can automatically determine one’s heart rate without the need for cumbersome diagnostic gadgets. To completely preserve one’s privacy, each video recording is locally processed in order to extract only the necessary color values from select facial regions. These values are then sent to the cloud, while the rest of the video is completely discarded, thereby ensuring no identifiable information ever enters the webspace. Users can log into our application to see a variety of comprehensive heart rate charts and infographics, and may also choose to add contact information for their healthcare providers so that urgent notifications can automatically dispatch should a health anomaly be detected.
Community Benefit
The advent of modern digital technologies have brought about unprecedented medical opportunities from the collection of individuals’ biometric data. Preserving privacy of digital health biomarkers is imperative to not only ensure that individuals' personal data is protected against misuse, but to build community trust in healthcare technology. Heart Vision offers a non-invasive, hands-free solution to heart rate monitoring which aims to dramatically improve the quality of care, reduce hospitalizations, and increase the lifespan of elderly people.
Team Members
Sponsors
FAU Multimedia Lab