AI-powered voice analysis to recognize the various emotions of students at a university

Resumen

Abstract– This paper addresses the use of artificial intelligence in voice analysis to recognize the emotions experienced by university students. The objective of this project was to create a voice recognition prototype to improve students' emotional well-being and enrich their learning process. The study highlights the importance of AI in education for comprehensively evaluating student participation, intrinsic motivation, and emotional well-being—essential factors for their holistic development. The research was quantitative and experimental, conducted with first-year university students. A total of 4,723 audio recordings were collected and categorized into six emotions: happy, sad, angry, scared, surprised, and neutral. Various techniques were then applied to balance and expand the model's training dataset. The trained machine learning model performed excellently, achieving 98% effectiveness in training and 91% in validation. Neutrality and Sadness were found to be recognized with high consistency; in contrast, Surprise was the most difficult emotion to recognize. The findings highlighted the influence of these emotions on the learning process, providing evidence for the subsequent development of a Socio-emotional-Constructivist Pedagogical Model that integrates pedagogical strategies to develop more comprehensive, supportive, and effective learning environments.