A Study on Impacts of Ambient Lights and Skin Tone for Contactless Blood Oxygen Measurement using Visible Spectrum Facial Videosaper

V. Yuvaraj,N. G. Vidya,V. Jeyakumar,R. Arulselvi
DOI: https://doi.org/10.1109/DELCON57910.2023.10127508
2023-02-24
Abstract:Blood oxygen saturation also termed SpO2, is a measure of oxygen-carrying capacity of the hemoglobin in the blood as a percentage. Normal arterial blood oxygen saturation level among humans varies from 97 to 100 percent (+/- 5%). SpO2 less than 90% is low, so-called Hypoxemia. Today, COVID patients need frequent and precise SpO2 assessments. Current pulse oximeters demand a contact-based measurement setup employing clips or sensors, which raises the possibility of skin irritation and infection. Additionally, depending on the patient’s age and finger size, pulse oximeters don’t always fit comfortably on everyone’s fingertips, which might result in misleading findings. Other restrictions of the current devices include errors brought on by interference, venous pulsations, skin problems, insufficient perfusion, light interference, and the presence of nail polish. This work offers a successful technique for non-contact SpO2 measurement from facial recordings. The videos show 31 participants doing three activities in various lighting situations. Each activity must be recorded with a camera mounted to a tripod. The obtained films go through pre-processing procedures include converting video to frames, discarding the background, finding faces, choosing regions of interest (RoI), and isolating the RGB channels. The SpO2 level is computed using the red and blue channels, which correspond to the red and infrared light used in contact pulse oximeters. In order to determine the pulsatile absorbance rate, the AC and DC components of the Red (R) and Blue (B) channels are separated. SpO2 will be predicted from the R-R ratio and its related parameter using Linear regression. Root mean squared error (RMSE) is the error between estimated and actual SpO2’s that is calculated for various skin tone participants. From the experiment, it is inferred that medium lighting with a lux range of 200-500 is a minimum requirement to estimate SpO2 which offers low RMSE for all skin types.
Engineering,Medicine
What problem does this paper attempt to address?