Detection for Joint Attention Based on A Multi-sensor Visual System

Abstract

Autism Spectrum Disorder (ASD) is one of the most common mental disorders in childhood, with a wide range and high risk. At present, there is no cure for autism. The symptoms associated with autism can be improved through early diagnosis and intervention. Joint Attention is an important paradigm in the diagnosis and intervention of ASD, the JA performance of child refers to the skill of following the eyes and fingers of others. This paper proposes an algorism based on a multi-sensor visual system, which gains the gaze of the child and transforms the Joint Attention detection into a geometric problem and proposes a solution. We conducted 20 rounds of Joint Attention testing on 10 non-ASD adults through this system and algorism, and achieved an accuracy of 97.94%.