Facial Expression Recognition as markers of Depression

• Published in: 2023 Asia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) pp. 674 - 680
• Main Authors: Gue, Jia Xuan, Chong, Chun Yong, Lim, Mei Kuan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 31.10.2023
• Subjects: Asia; Depression; Face recognition; Focusing; Image recognition; Information processing; Mental health
• ISSN: 2640-0103
• DOI: 10.1109/APSIPAASC58517.2023.10317295

With research done to determine depression with different types of inputs, such as text, images, and mobile sensing, numerous distinct approaches have been devised and enhanced for identifying depression across diverse backgrounds. However, there is a significant lack of research in making use of only facial expressions to detect depression. Introducing such methods could be beneficial due to the large amount of research done to develop models capable of performing facial expression recognition. This project aims to utilize pre-existing facial expression models to determine depression based on the expression recognition capabilities, and to also determine the reliability and effectiveness of using only facial expression to detect depression. We have chosen to use EfficientNet, a state-ofthe-art facial expression recognition model, to experiment with the different facial expressions that are shown based on images of clinically proven to be depressed patients. The findings from our work can further improve depression detection research, specifically focusing on making use of only information from shoulder up of patients, while also helping the community to improve mental health.