This work presents the development of an educational support system based on Convolutional Neural Networks (CNNs) applied to facial emotion recognition. The model was trained using public datasets of emotional expressions, enabling the real-time identification of affective states such as happiness, sadness, anger, surprise, and neutrality. From these detections, two dynamic indicators were defined: concentration and nervousness. Both were computed through a weighted mapping of emotions, where each recognized emotion contributed with specific coefficients to quantify levels of focus and stress. This methodology was inspired by studies in affective computing and educational psychology, which emphasize the influence of emotional states on attention and test anxiety.

The CNN model achieved an accuracy of approximately 70% during training and validation, ensuring reliable emotion detection for subsequent analysis. To determine readiness, a rule-based mechanism was applied: students were considered prepared when concentration reached at least 60 out of 100 while nervousness remained below 50. By combining these two indicators, the system provided an objective and interpretable evaluation of the student’s emotional readiness to answer questions or undertake an assessment.

The system was designed to support teachers in better understanding students’ emotional states during evaluative activities. By integrating emotional and cognitive factors, educators gain a more holistic view of the learning process, promoting fairer and more inclusive evaluation practices.

Experimental results confirmed consistent estimations of concentration and nervousness, with reliable classification of test readiness. These findings highlight the potential of artificial intelligence as an innovative tool in contemporary education.