Convolutional Neural Network-Based Recognition of Children's Facial Expressions in Response to Gaming

This study explores the use of Convolutional Neural Network (CNN) algorithms for the purpose of recognizing children's facial expressions during gaming activities, with a focus on understanding the emotional consequences of gaming. This study intends to build a robust model and assess the accuracy of CNN in detecting six basic emotions among children aged between 6 and 13 years using our dataset that we collected from children in Timor Leste as many as 600 images and the Children's Real-World Facial Expressions (CFEW) dataset of more than 11,000 images for training data. Then we also use our video data and the LIRIS-CSE dataset from the internet as test data as many as 180 videos and images. The data we obtained were images of children when not playing games and playing games consisting of facial expressions, especially those showing anger, happiness, sadness, fear, surprise, and neutral. This methodology consists of various processes, including data collection, preprocessing, augmentation, model training, and evaluation, with the main goal of identifying patterns and trends in children's emotional responses to games. The results of this study indicate that the final accuracy of detecting children's faces when playing games is 96.78% and the validation data accuracy value is 95.32%. It is proven that the CNN architecture or model used in this research dataset is optimal.