Towards Early Detection of Autism Spectrum Disorder through Culturally Grounded, Multimodal Behavior Detection

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by difficulties in social communication, interaction, and adaptive behavior. Early identification is critical, yet most children are diagnosed well after the optimal intervention window because of limited clinical resources and the lack of scalable, naturalistic assessment tools. This doctoral research proposes a culturally grounded multimodal framework for ASD assessment using semi-structured triadic play interactions among a child, caregiver, and examiner. Built on a novel Play Behavior protocol tailored to the Indian context, the framework investigates facial, social-communicative, and play-based behaviors through statistical analysis and automated multimodal video understanding. It further integrates a two-stage pipeline for behavior recognition and ASD prediction, producing interpretable clinician-facing behavioral profiles. By combining culturally grounded behavioral assessment with automated analysis, the proposed framework aims to support objective, low-cost, and scalable ASD screening in resource-constrained settings.