An investigation of upper limb motor task based discriminate for high risk autism

Abstract

Early diagnosis of children with autism means early intervention which is very important to increase child treatment outcomes. One of these signs is based on examination of child upper limb movements. This paper presents an evaluation of tasks methods for investigation of upper limb motor in children at High Risk (HR) for autism. This study examined three tasks and finds the upper limb motor type that discriminates the high risk of autism. These three tasks are: the first task is a throw a small ball into a transparent plastic followed by insert the ball into a clear tube open at both sides. In the second task, place a block into a large open box then place four similar blocks on a target block to make a tower and in the third task is put in a shape into a small fund with removable lids. The paper introduces using machine learning to discriminate between the children with high risk for autism. The feature extracting techniques Linear Discriminant Analysis (LDA) is used. After generating feature vectors, Support Vector Machine (SVM) and Extreme Learning Machine (ELM) are used for classification step. The results are very encouraging the maximum classification accuracy for a task that inserts the ball into a clear tube open at both sides with mean accuracy 75.0% and 81.67 with SVMs and ELM respectively.