Validation of an Emotion Recognition System for People With Down Syndrome

This research presents the validation of a novel software application designed to recognize primary emotions in individuals with Down syndrome (DS), to support therapeutic interventions through artificial intelligence. The research addresses the need for innovative technological tools that assist therapists in real-time emotional assessment during therapy sessions. The study’s objective was to validate the application’s effectiveness, reliability, and therapeutic usefulness in recognizing five spontaneous emotions (happiness, anger, sadness, surprise, and neutrality) in individuals with DS attending a specialized care institution. The study followed ethical protocols approved by the Ethics Committees in Colombia and Ecuador to achieve this. Data was collected during therapy sessions, and three research hypotheses were formulated to evaluate the application’s performance. Structural Equation Modeling (SEM), using SmartPLS, was employed to analyze the relationships between observed emotional responses and the system’s feedback. The results demonstrated that the application accurately identified the targeted emotions in real time, and 94% of participating therapists positively assessed its usefulness in clinical settings. This validation confirms that the software can provide valuable input for therapists, enabling the design of tailored strategies that address the emotional needs of individuals with DS. The findings support the growing evidence supporting integrating machine learning and deep learning technologies into therapeutic tools for vulnerable populations.