Biometric Facial Recognition System and Expression Classifier Using Deep Learning

Espinoza P.; Sinche E.; Ticona W. · Proceedings of the 14th International Conference on Cloud Computing, Data Science and Engineering, Confluence 2024 (2024) · DOI: 10.1109/confluence60223.2024.10463221

Resumen

The recognition of emotions through facial expressions is difficult for computer systems, unlike humans who can easily do it in various contexts, such as interaction with computers. Studies indicate that the most effective approach to automatic emotion recognition is machine learning, and deep learning in particular offers greater accuracy. This research focuses on determining the level of stress and fatigue based on the facial features predicted by a proposed model, grouping indicators such as anger, sadness, and fear as signs of stress. A model was trained using convolutional networks to analyze facial patterns and relate them to emotions. A Kaggle dataset containing various facial expressions was used for testing and training. Special attention was paid to extracting and processing the captured video camera images to remove noise, which allowed for accurate classification of many facial reactions. This, in turn, helped predict burnout indicators such as stress and fatigue, with greater than 90% accuracy.

Instituciones: Universidad ESAN; Universidad Tecnológica del Perú