Facial Emotion Recognition System Using CNN-Mini-Xception

Facial emotion recognition (FER) in children with Down syndrome poses unique challenges due to their distinctive facial features and potential cognitive differences. This study explores the feasibility and efficacy of employing a convolutional neural network (CNN) based approach, specifically utilizing the mini-Xception architecture, to address these challenges that arose from the data collected for individuals with Down syndrome captured spontaneous emotions in uncontrolled environments. The CNN model achieves promising results, initially attaining an accuracy of 0.85 in recognizing primary emotions. Through meticulous hyperparameter tuning and introduction of the MiniXception model, the accuracy improves significantly to 0.92, demonstrating the effectiveness of the proposed approach. This research contributes to enhancing our understanding of FER in children with Down syndrome and offers a valuable tool for supporting their social and emotional development.