ScholarWorks@숙명여자대학교

초록

In this paper, we propose Proxi-rPPG - a contactless remote photoplethysmography (rPPG) sensor mounted on a head-mounted display (HMD) - to reliably obtain heart rate data in a virtual reality (VR) environment where users can move freely. We also develop and evaluate an emotion-recognition model using the Proxi-rPPG approach. Specifically, we attach a small camera module to the outside of an HMD to capture the user's cheek-ear region at close range, then extract heart rate and heart rate variability via an rPPG algorithm. While traditional reflective PPG requires precise skin contact - and wrist-worn sensors can be highly susceptible to noise from arm movements - conventional rPPG methods also face limitations imposed by user motion. Proxi-rPPG addresses these drawbacks and demonstrates its potential to quantify emotional changes arising from VR content experiences. In a user study with 20 participants, we compared waveform data from the Proxi-rPPG sensor to that of a reflective wrist-PPG device. We found that the Proxi-rPPG approach yielded a lower noise ratio than the wrist sensor in dynamic VR conditions. It also achieved 65 % accuracy in a four-quadrant emotion classification task (High/Low Arousal × Positive/Negative Valence) using a machine-learning model. These findings suggest that Proxi-rPPG can be effectively applied to various domains, including emotion monitoring in VR.

키워드