Abstract

A number of children with Autism Spectrum Disorder (ASD) are growing to become a big problem in society at the present time. Autism is the neurological disorder that makes the children feel inferior in social communication. Identifying individuals with early autism (under 3 years of age) is difficult and there is still not any medical test for rapid detection of autism. Currently, the clinical diagnosis is mainly based on the observed behavior of the child, and besides, educational and psychological tests are also applied. Eye movements have an important role in an individual's perception and attention in social activities. Non-invasive detection and tracking techniques of the eye movement have been developed over many decades. Nowadays, the Eye Tracking is a technological process that enables the measurement of eye movements, eye positions, and points of gaze. In other words, eye tracking identifies and collects a person’s visual data of the eyes in terms of location, objects and duration. It is applied for a variety of different research methods to investigate human behavior. In particular, the Eye Tracking method makes an application for identifying people with Autism Spectrum Disorder (ASD), it has been not only a wide study but also a great interest in recent years. This article provided an overview of children's visual attention with Autism Spectrum Disorder and children with Typical Development (TD). The study has focused on exploring the difference in observed behavior of ASD children and TD children. This study used the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm for grouping a set of Eye Tracking data, which were collected by the test of children’s visual attention to a variety of images. The classification of Eye Tracking data was based on the feature parameters, which were extracted from analyzing data. Moreover, a model for classification and identification of Eye Tracking image data of children groups with ASD and TD was built on Deep Learning algorithms. In this study, the chosen algorithm was Multilayer Perceptron. The children's eye movements dataset was extracted from the paper "A dataset of eye movements for the children with autism spectrum disorder" by Huiyu Duan et al. The results showed that using the Eye Tracking data is highly promising in identifying children with ASD by analyzing the observed behavior of children on images, especially on human images.