Enhancing Remote Monitoring of Emotional Health in Children Using Facial Emotion Recognition

In recent years, the importance of monitoring emotional health of children has gained significant attention, particularly in the context of cyberspace and remote environments. This research focuses on enhancing the real-time monitoring of the emotional health of children using facial emotion recognition (FER) through a deep learning-based Convolutional Neural Network (CNN) model. A custom dataset was created by compiling child images from well-established emotion recognition datasets such as FER-2013, CK+, JAFFE, and AffectNet, categorizing them into four primary emotion classes: Happy, Sad, Neutral, and Angry. The model was trained and evaluated on this dataset, achieving a training accuracy of $\mathbf{9 4. 9 0 \%}$ and a validation accuracy of $\mathbf{8 8. 1 2 \%}$. The system is designed to assist psychiatrists in understanding the emotional states of children, allowing them to provide timely recommendations or instructions to parents. The approach has significant potential in teletherapy and e-therapy applications, especially in monitoring and addressing children’s emotional well-being remotely.