Real-Time Emotion-Based Concentration Estimation: An Educational Framework with Lightweight Neural Networks

This work presents a facial emotion-based concentration monitoring system, designed with a didactic focus for students beginning their studies in Artificial Intelligence. The application uses the OpenCV library for real-time video capture and face detection, combined with a simple neural network model trained on facial expression images to classify seven basic emotions: anger, disgust, fear, happiness, neutral, sadness, and surprise. The key innovation of this project lies in its simple and accessible structure, allowing students to grasp core concepts of computer vision and machine learning through practical experimentation. The model was intentionally trained using a reduced and simplified dataset, emphasizing how basic neural architectures can still produce functional results in real-world scenarios. Based on the detected emotion, the system applies straightforward rule-based logic to estimate the user’s concentration level (high, medium, or low), offering an intuitive application of AI in educational or interactive environments. In addition to promoting technical learning, the codebase is modular, well-documented, and easily encourages students to explore extensions such as model refinement, data preprocessing, or alternative AI approaches. This work bridges theoretical learning and hands-on AI application, highlighting how even minimal resources and simple neural models can serve as powerful tools for understanding intelligent systems and human–computer interaction.