Design of a Behavior Classification System for Monitoring Autism Spectrum Disorder using Deep Learning

A developmental disorder known as an autism spectrum disorder (ASD) is characterized by unusual social, communication, and behavioral situations. According to experts, children with ASD ranging from 2-16 years old need constant monitoring because of their unpredictable behavior. Meltdown and seizure are two of the most important aspects that the parents need to monitor because of their aftermath on children with ASD. This paper is aimed at designing a behavior classification system for monitoring children with ASD that classify the activity such as shaking to seizures and self-injurious behavior to meltdown, classifies the audio of the child, such as screaming and crying to meltdown, and designs a mobile application that has recommendations on the proper response of the parent or guardian. The system is built based on the SqueezeNext and Transformer algorithms, which utilize two Raspberry Pi 4B, equipped with a Web Camera responsible for the behavior classification and Re Speaker 2-Mics Pi HAT is responsible for the audio classification of the child with ASD. The audio and video classification system had 98.33% accuracy in classifying the child’s behavior, and the device’s alert notification and recommendation system had 100% system accuracy. The Behavior Classification System’s design has proven to be effective in assisting parents and guardians of children with ASD.