Vital signs estimation in elderly using camera-based photoplethysmography

In the context of Ambient Assisted Living, the demand for healthcare technologies development has been increased in the last two years by the current global health situation. The contagious nature of COVID-19 warrants inevitability to the thought of easing a continuous measurement of vital signs such as Heart Rate, Breath Rate, and SpO2 in a non-invasive manner, contributing of fundamental importance in the assessment of health status especially among frail and/or elderly individuals. A widely diffused method for contactless vital signs monitoring is remote photoplethysmography from facial video streams that, contrary to traditional contact measurement techniques, allows the measurement of vital parameters without the need for wearable sensors (generally considered uncomfortable, especially by the elderly), even with commercial and low-cost digital cameras. This paper proposes the design and implementation of a new pipeline for estimating Heart Rate, Breath Rate, and SpO2 values, and its integration on Raspberry Pi 4 as an elaboration unit. The pipeline provides specific algorithmic blocks to improve vital signs estimation in elderly subjects as it is made more difficult by the skin tone and the presence on the face of wrinkles, folds, and moles. Quantitative evaluations on our dataset containing the acquisition of only elderly older than 65 years of age demonstrate the validity of the proposed pipeline. For validation against the state of the art, tests were also conducted on three standard benchmark datasets containing video with subjects of varying ages. Again, the pipeline proved to be robust concerning the estimation of vital signs considered in the present work.