Continuous and Non-Contact Extraction of Respiratory Signal and Rate from Thermal Video

Achieving a simple, non-contact, non-invasive, and non-disruptive method for measuring and monitoring vital signs, including respiratory signals, is crucial in this field of research. This paper aims to implement algorithms for extracting respiratory signals and breathing rates from thermal facial images. Unlike many previous studies, where the algorithm was validated under resting conditions, this paper designs a protocol to capture a diverse range of respiratory activity datasets in varying conditions. Data were collected from six healthy individuals, with thoracic and abdominal contact-based respiratory signals used as reference signals. Processing steps, including robust facial tracking in thermal images and precise selection of target areas around the nose is implemented to extract respiratory signal and rates from the thermal video. The effectiveness of the proposed computational method is assessed through both quantitative and qualitative metrics, considering a variety of breathing patterns beyond just resting conditions