LilyDoll: Designing a System for the Detection of Autism Spectrum Disorder

Autism Spectrum disorder (ASD) is a neurodevelopmental disorder characterized by communication difficulties, limited interests, and repetitive behaviors. Kids with ASD struggle with various life skills, including personal hygiene, emotional regulation, and social interactions. Traditional interventions may be valuable but do not always cater to these individuals' unique needs and preferences. This research aims to bridge this gap by introducing an innovative, technologydriven approach to facilitate skill development and learning for children with ASD. A system named “LilyDoll” is presented, which contains several key elements, including ASD application development, autism detection research, and modular learning support for children. This study proposes an architecture for the LilyDoll project, which outlines the key elements, including various users, application development design, and other associated technological details. In addition, a machine learning (ML) based autism detection framework is proposed to be incorporated within the architecture of this research project. Through the incorporation of ML-based autism detection, the project can continuously evolve and improve its detection performance through enhancements to the framework and by feeding more training data to the system. The proposed framework uses machine learning for training on autism screening data and provides an accuracy of 100% and an F1 score of 0.99. In conclusion, this research project aims to solve the challenges of autism diagnosis and education while also creating effective solutions for children and their caregivers.