Analysis of Art Therapy for Children with Autism by Using the Implemented Artificial Intelligence System

Abstract

Autism is a disease that manifests as social communication disorders and repetitive sensory movements. The current incidence of autism continues to rise globally, and the number of children with autism is also increasing. The treatment of autistic children in China is mainly intervention, but experienced autism instructors are lacking. Here, machine learning and artificial intelligence (AI) algorithms are adopted as the technical foundation to build a system that measures the emotional performance of autistic students in the classroom based on actual application scenarios. By analyzing the classroom learning data of autistic children, the students’ emotions can be effectively judged to help teachers evaluate and track their classroom performance and reduce teachers’ burden. Results demonstrate that when the matrix composed of frame sequences and key point coordinates is used as the input, the spatio-temporal graph convolutional network is determined as the principal model of action recognition, with an accuracy of 90%, and the participation score can be obtained by calculating the action response time. In the experimental process of facial expression recognition, the random forest’s classification accuracy of the feature point sequence based on images can reach 99%. Therefore, the random forest is determined as the principal classifier for facial expression recognition. After the relationship between expression intensity, pleasure, and expression category is analyzed, the scoring method is designed. The experiment also discovers that painting can be a rehabilitation therapy for children with autism. The above results can provide a theoretical foundation for the treatment of autistic children.