Title: Early Autism Screening Using an Intelligent Toy; Sorour Ekhtiari, Hadi Moradi, Hamidreza Pouretemad, Babak Araabi. 1391
Introduction: Detecting autism in early ages is essential for its treatment, but in many regions there is no doctor with adequate expertise to detect autism in children.Therefore there is a great demand for an automatic method, which can detect or screen autistic children.
An automatic method needs to rely on behaviors detectable in an autistic child. One of the most important behaviors in autistic children is the repetitive behavior which, may appear when an autistic child plays with toys. Thus this behavior along with a few other behaviors and feature, which are related to movement, could lead us to the design of an automatic method for screening autistic kids.
As children spend lots of time playing with toys and the majority of autistic children are male, a toy car is a suitable object for our purpose of analyzing a child’s behavior while playing with it.
Objective: This paper describes the design of an intelligent toy to warn parents to pay more attention to their children and consult an expert in case they have a chance of autism.
Methods: To validate the findings, 24 children in which 14 are autistic and 10 are normal, were selected and tested. The children in the autistic group were between 3.5 to 6.5 year old (mean=4.5179, SD=0.7750) and the normal children were between ages 4 to 6 (mean=5.35, SD=0.5798) years.
Results: Support Vector Machine (SVM) classifier with soft margin and polynomial kernel is applied to the collected data and the maximum achieved accuracy is 80%.
Conclusion: This paper presents a novel approach in screening autistic children using an intelligent toy car. A toy car is selected since most of autistic children are male. The selected features, aimed at better distinction of the two groups are used in anSVM classifier with soft margin and polynomial kernel. The future work is focused on improving the classification process and choosing better features to reach higher classification rate.
Keywords: Autism, Repetitive behavior, Intelligent toy car, Classification, Support Vector Machine