[Paper] Multimodal Stress Estimation Using Multibiological Information: Towards More Accurate and Detailed Stress Estimation

抄録

In this paper, we propose a method for estimating stress levels using multibiological information obtained from red-green-blue (RGB) facial videos. The most popular method for estimating stress is by questionnaire, which is not highly objective or accurate. To estimate stress levels more accurately, it is desirable to establish the method which can estimate with objective and accurate information. In recent years, many methods for using biological information have been proposed. In this study, we used three types of biological information to estimate stress: pulse rate variability, blink rate, and pupil size, which were captured using an RGB camera. Results showed that the use of these biological signals improved the accuracy of stress estimation. Moreover, using biological information in a multimodal manner enabled the most accurate and detailed stress estimation.