Non-contact measurement of pulse rate variability using webcam and application to mental illness screening system


This article was originally published here

Annu Int Conf IEEE Eng Med Biol Soc. November 2021; 2021: 7016-7019. doi: 10.1109 / EMBC46164.2021.9630038.

ABSTRACT

The COVID-19 pandemic is a global health crisis. Mental health is essential in such uncertain situations, especially when people have to severely restrict their movement and change their lifestyle. As a result, many countries have turned to telemedicine to strengthen and expand mental health services. Our research group previously developed a mental illness screening system based on Heart Rate Variability Analysis (HRV), allowing objective and easy self-monitoring of mental health. This screening system cannot be used for telemedicine because it uses electrocardiography (ECG) and contact photoplethysmography (PPG), which are not widely available outside of a clinical setting. The aim of this study is to allow the extension of the aforementioned system to telemedicine by the application of contactless PPG using an RGB webcam, also called imaging-photoplethysmography (iPPG). IPPG measurement errors occur due to changes in the relative position between the camera and the target, and due to changes in light. Conventionally, in image processing, the pixel value of the entire region of the face is used. We offer skin pixel extraction to eliminate blinking, eye movements, and changes in light and shadow. In signal processing, the green channel signal is conventionally used as a pulse wave due to the absorption characteristics of blood flow. Taking advantage of the fact that the red and blue channels contain noise, we provide a signal reconstruction method to remove noise and strengthen the signal in the pulse rate variability (PRV) frequency band by weighting the three signals of the RGB camera. We carried out an experiment with 13 healthy subjects and showed that the PRV index and pulse rate (PR) errors estimated by the proposed method were lower than those of the conventional method. The correlation coefficients between the values ​​estimated by the proposed method and the reference values ​​of LF, HF and PR were 0.86, 0.69 and 0.96, respectively.

PMID: 34892718 | DOI: 10.1109 / EMBC46164.2021.9630038


Source link

Previous Competition to name eagles with webcam nest on Hilton Head SC
Next Razer Kiyo Pro: our favorite webcam is on sale for $ 100 off