Towards a tool for identification of emotions and integration with serious game learning.

The use of interactive systems in the school context, such as educational games and virtual learning environments, has become increasingly common, especially due to their playful nature, which leads to a more attractive and interesting learning process for the student. However, the development of these systems needs adequate metrics for their evaluation so that their problems are explicit, clear, and objective. Therefore, observing the user’s emotional states during an interactive tool is highlighted as an important methodology for classifying facial expressions. In view of the above, the present work aims to develop a convolutional neural network model capable of classifying emotions in order to, in future works, create an application capable of recognizing the emotions of a user through a webcam. Thus, a model was developed which, through tests, presented the best accuracy of 94.69%, a value that, when compared with the best value obtained in related works (97%), indicates the quality of the model’s architecture the relevance of its use in possible future work.