Real-time Stress Detection through Facial Expressions Using a Vision Transformer
Keywords:
Stress Detection; Facial Expressions; Vision Transformer; Deep LearningAbstract
Stress is a common issue in today’s society, impacting emotional health, decision-making, and work performance. This research introduces a real-time, non-contact method for detecting stress by analyzing facial expressions using deep learning. A Vision Transformer (ViT) model is fine-tuned to recognize facial features linked to stress and non-stress states. Unlike conventional techniques that rely on wearable sensors, this method uses a standard webcam, offering greater ease of use and accessibility. The model is trained on two publicly available datasets and assessed through accuracy, precision, recall, and loss metrics. It achieves an 86% accuracy rate in identifying stress in real time. This approach eliminates the need for physical devices, enhancing comfort and usability. It provides quick and automatic stress detection, making it suitable for use in practical settings such as schools, workplaces, and public areas. The system presents a scalable, user-friendly solution that supports emotional awareness and contributes to mental health monitoring.
References
[1] “Stress Levels rise around the world”, WIN/Gallup International. (2022), https://winmr.com/stress-levels-rise-around-the-world/
[2] Part, C., Rosenblatt, J. D., Lee, Y., Pan, Z., lacobucci, M., Ong, A., & McIntyre, R. S. “The association between stress and severe mental illness: A meta-analysis”. Psychiatry Research, 289, 113091, 2020
[3] Chew, W.T., Chong, S. C., Ong, T.S & Chong, L.Y. (2021). “Facial Expression Recognition via Enhanced Stress Convolution Neural Network for Stress Detection”. VISAPP 2021.
[4] Ogasawara, R., Bouazizi, M., & Ohtsuki, T., “Camera-Based Stress Detection Using Face-Related and Emotion-Related Features”. IEEE International Conference on E-health Networking, Application, and Services (HealthCom), 2024
[5] Lek, J. X. Y., & Teo, J., Academic Emotion Classification Using FER: A Systematic Review. Human Behaviour and Emerging Technologies, 2023, Article ID 9790005, 2024
[6] Zhang, J., Mei, X., Liu, H., Yuan, S., & Qian, T. (2019). Detecting Negative Emotional Stress Based on Facial Expression in Real Time. 2019 IEEE 4th International Conference on Signal and Image Processing.
[7] Jeon, T., Bae, H., Lee, Y., Jang, S. (2023). Stress Recognition Using Face Images and Facial Land Marks. Proceedings of the IEEE Conference
[8] Sriramprakash, S., Vadana, D. P., & Murthy, O. V. R. (2017). Stress Detection in Working People. Procedia Computer science 115, 359-366
[9] Voleti, S., NagaRaju, M. S., Kumar, O. V., & Prasanna, V. (2024). Stress Detection from Facial Expressions Using Transfer Learning Techniques. 2024 International Conference on Distributed Computing and Optimization Techniques (icdcot)
[10] Almeida, J., & Rodrigues, F. (2021). Facial Expression Recognition System for Stress Detection with Deep Learning. Proceedings of the 23rd International Conference on Enterprise Information Systems (ICEIS 2021), 256-263
[11] Zhang, J., Yin, H., Zhang, J., Yang, G., Qin, J., & He, L. (2022). Real-time Mental Stress Detection using Multimadality Expressions with a Deep Learning Framework. Frontiers in neuroscience, 16, 947168
[12] Hills, P. J., Dickinson, D., Daniels, L. M., Boobyer, C. A., & Burton, R. (2019). Being Observed Caused Physiological Stress Leading to Poorer Face Recognition. Acta Psychologica, 196, 118-128
[13] Jebelli, H., T., Kamat, V. R., & Ahn, C.R. (2021). A machine learning approach for real-time stress recognition using wearable sensors. IEEE Transactions on Automation Science and Engineering, 18(2)
[14] Shaees, S., Naeem, H., Arslan, M., Naeem, M. R., Ali, S. H., & Aldabbas, H. (2021, September). Facial emotion recognition using transfer learning. In 2021 International Conference on Computing and Information Technology (ICCIT-1441) (pp. 1-5). IEEE
[15] Oh, S., & Kim, D. K. (2022). Comparative analysis of emotion classification based on facial expression and physiological signals using deep learning. Applied Sciences, 12(3), 1286
