Effect of Different Lighting Conditions on Camera-based Non-Contact Vital Sign Monitoring

Wearables have been used to collect physiological signals for long-term pulse rate (PR) monitoring. Most of the wearables require skin contact for continuous monitoring. Recently, the potential of the contactless reflective photoplethysmogram (crPPG) has been extensively explored for non-contact sensing, particularly in geriatric patients and mobility tracking. Despite numerous studies, it has been recognized that the outcomes can be affected by ambient light conditions. In this work, an attempt has been made to explore the effects of varied light conditions for non-contact pulse rate monitoring using a laptop camera. The facial videos and ECG signals as Ground Truths are recorded from normal subjects (N=10) under three lighting conditions: white, shadow, and sunlight. The videos are applied to the face detection approach, and the regions of interest (forehead and cheek) are extracted. The mean pixel intensity of green channel frames and the corresponding Hue-channels signals are calculated. The hue signals are subjected to the fast Fourier transform (FFT) for the frequency spectrum and the peaks of the frequencies are used to compute the pulse rate in beats per minute (bpm). The results show that the proposed approach is able to extract bpm from extracted crPPG signals. The bpm exhibits high variation in sunlight conditions, whereas it is stable in shadow lighting conditions. The proposed algorithm shows neutrality in controlled and shadow conditions. Thus, the proposed approach could be useful for non-contact pulse rate monitoring in real-time conditions.