Video-Based Stress Level Measurement Using Imaging Photoplethysmography

Abstract

In this paper, we propose a method for measuring stress levels using pulse rate variability features without any contact devices. Mental stress can be considered as a crucial factor which triggers mental diseases such as depression. In order to deal with the mental stress, many researchers tried to quantize the mental stress making full use of vital signs. However, contact sensors provides discomfort to people and the person who have skin irritation or damage are not able to attach the sensors. Instead of contact sensors, imaging photoplethysmography (iPPG) is growing field of interest. Despite some papers are demonstrated the effectiveness of pulse rate variability (PRV), almost papers did not investigate the effectiveness of PRV analysis. Moreover, PRV analysis provides only binary classes which are relaxed or stress conditions. Therefore, the contribution of our paper are follows. First, we investigate the effectiveness of PRV measurement in each iPPG method and proposed method from 78 videos for benchmark. Second, we extracted multiple features from PRV parameters for measuring the stress levels, not binary classes. Third, the combination method of optical-based and spatial subspace rotation methods indicates the best accuracy in all previous iPPG methods.