Presentation Information
[A-11-04]Evaluation of Relaxation States Based on Facial Expressions Using Image Recognition Technology
◎Nao Sakai1, Mahisa Hoshino1, Takahiro Baba1 (1. Kurume Institute of technology)
Keywords:
Image Recognition,Facial Expression Recognition,Psychological State Estimation,Relaxation Assessment,Machine Learning
In recent years, increasing levels of stress and mental workload have become significant social issues, leading to growing interest in technologies for estimating users' psychological states. This study investigates a method for evaluating relaxation states based on facial expressions using image recognition technology. An experiment was conducted with ten participants aged 18–23 years. The participants performed typing tasks under a silent condition and while listening to music from nine different genres. After each trial, a psychological-state questionnaire was administered, and facial images obtained during the task were labeled as "Relaxed," "Neutral," or "Stressed" according to the questionnaire results. An image recognition model was trained using 80 of the collected 100 facial images, while the remaining 20 images were used for evaluation. Classification performance was assessed using evaluation data excluding the neutral class. The results showed an overall classification accuracy of 81%, with accuracies of 85% for the relaxed state and 50% for the stressed state. These findings suggest that psychological states can be estimated from facial expressions. Future work includes increasing the amount of training data and improving learning methods to enhance estimation accuracy.