A Deployable Sensor Platform for Contactless Health Observations in Uncontrolled Environments

Abstract

This work aims to investigate and improve the capabilities of contactless vital sign monitoring systems (CVSMs) in environments with increased variance in lighting and subject pose using a deployable hardware platform designed for use in the DARPA Triage Challenge (DTC). Current research into CVSMs primarily focuses on their use in controlled environments, like sitting at a desk or driving a vehicle. The deployable platform, henceforth referred to as the Sensor Package for Updated Diagnosis (SPUD), is designed to operate in a disaster-response triage scenario as represented by the DTC, and consists of a Raspberry Pi along with a near-infrared time-of-flight (TOF) camera used to obtain intensity and depth images. Heart rate (HR) and respiratory rate (RR) are estimated from the datastream via a ROS2 pipeline that can then be integrated into a larger network of autonomous systems for potential use as part of the DTC. We analyze how the system performs under a variety of subject positions. We struggled to replicate the success of HR CVSM techniques, but research is continuing to improve the accuracy of this system. For an ideal subject seated at a desk, we find an RR estimation to be within 10% of ground truth in 85% of tests. In other testing configurations, RR estimate accuracy stays within 10% of ground truth in a minimum of 55% of tests.