Transferring a facial depression model to estimate mood in a natural web browsing task

Abstract

Because people are living in a stressful era, they are prone to common mental health problems, which cause them to experience low mood and loss of interest or pleasure. Although many suffer from depression/low mood, they hesitate to undergo clinical check-ups. Therefore, a systematic and efficient web-based system that automatically detects emotions is necessary. The purpose of this study was to design and develop a system that can automatically detect negative and positive mood states and to investigate the relationship between the depression and mood states of individuals. User’s facial expressions features are detected and analyzed in real-time and after the completion of using the system, the determined emotion is hence provided to the users. A facial depression model constructed from a dataset obtained in a human–agent interaction (HAI) experiment was applied to a general human-computer interaction (HCI) situation to classify negative and positive mood states. The model exhibits the highest accuracy rate for classifying mood states. These findings suggest that faces provide strong evidence of mood induction to depression and guidance to construct the automatic mental health care web-based system to know the preliminary mental state.