Stress levels estimation from facial video based on non-contact measurement of pulse wave

In this paper, we propose a method to measure multiple stress levels using pulse wave extracted from a facial video. Stress is a crucial factor which triggers mental illnesses. For the prevention, it is necessary to measure and handle with stress. Many researchers have attempted to measure stress using with a contact sensor. However, contact sensors can be uncomfortable or unsafe depending on the person. Recently, imaging photoplethysmography (iPPG), which measures pulse wave without contact sensors has been gathering interest. However, the estimatability of pulse rate variability (PRV) on stress condition, which include information of stress has not been investigated in most papers. In addition, the stress measurement based on PRV only detect relaxation and stress. Therefore, in this study, the following contributions is made. First, we investigate the effectiveness of the PRV of the several iPPG methods on stress condition. It is noted that comparing several iPPG methods on the stress condition is not achieved in the previous works. Second, we propose a method to estimate 4 stress levels based on PRV features based on a facial video with the iPPG methods. It is noted that “multiple” stress levels estimation based on ‘a facial video’ has not been accomplished in the previous works in the iPPG method.