Facial Expression Recognition (FER) of Autism Children using Deep Neural Networks

Children with Autism Spectrum Disorder (ASD) have difficulty interacting with their parents, teachers, and classmates because they lack the required social skills and often engage in repetitive behavior. In this research, a deep learning method was used to identify facial emotion in children with autism while interacting with a computer or mobile screen, which helps to provide them with assistance in a timely manner. An optimized Visual Geometry Group (VGG) model has been proposed for deep convolutional neural network (CNN) expression recognition. The CK+48 and JAFFE databases were used to analyze and train facial expression data using the VGG-16 deep CNN. The overall results showed that the proposed model achieved 99.09% accuracy for the CK +48 group and 95.23% accuracy for the JAFFE group.