Advancements in Automated Assessment and Diagnosis of Autism Spectrum Disorder Through Multimodality Sensing Technologies: Survey of the Last Decade

Abstract

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder characterized by difficulties in social interaction, communication, and repetitive behavior patterns. Traditional research approaches have primarily focused on studying autism using single-modal data analysis, such as relying solely on audio, video, or neuro signals. However, recent advancements in technology, cognitive science, and artificial intelligence (AI) have provided opportunities to explore the potential benefits of multi-sensory integration and fusion of modalities in understanding autism patterns. This survey makes three key contributions to advancing the future of ASD diagnosis and intervention. First, it provides a comprehensive review of recent advancements in multi-modal sensing technologies, detailing primary modalities, data cleaning and synchronization techniques, feature extraction, and fusion methodologies to integrate diverse sensory data. Second, it classifies assistive technologies into three major categories: (i) computer-based systems, (ii) virtual reality simulations, and (iii) robotic interactions, analyzing their applications for cross-referencing symptoms and enabling real-time interventions in skills assessment and therapy. Third, it identifies critical challenges related to data collection, sensor synchronization, standardizing assessment paradigms, and real-time processing demands, proposing actionable future directions to improve diagnostic precision, scalability, and adaptability. These contributions underscore the transformative potential of multi-modal sensing systems to revolutionize ASD assessment and diagnosis by enabling comprehensive, objective, and tailored solutions for diverse individuals across the autism spectrum.