Method for Classifying Emotions in Older Adults Playing eSports

In recent years, extending the healthy lifespan of older adults has become increasingly important; however, dementia is a major hindrance to achieving this goal. Electronic sports (eSports) are receiving increasing attention owing to their potential to prevent dementia. People experience various emotions while participating in eSports. When quantified, these emotions can serve as an indicator of how much participants enjoy the game and can be used to evaluate eSports. This study investigates methods for estimating emotional arousal from videos of participants playing eSports and classify these emotions into positive and negative categories. We used changes in the saturation in both cheek areas of the face during emotional events as inputs for a machine learning model. This model combined a convolutional neural network and a long short-term memory network. The results suggest that using saturation information from both cheeks as features in machine learning can potentially classify emotions into two types: positive and negative. We believe this method can be a key technology for extending the healthy lifespan of older adults and supporting remote communication.