Automated robotic system for autism treatment

Abstract

This article is about possible implementation of automated robotic system for children's autism treatment. There are many well established approaches to autism treatment, however all of them are done manually by therapists and parents. In order to reduce manual work, we try to partially automate ABA method -, the one which has practically proved to be useful and scientifically well researched, - applied behaviour analysis based treatment. Simply put ABA, among others, includes time-consuming and laborious procedures which can be transferred into computational tasks. That allows us to apply computational algorithms from data mining and computer vision domain. Therefore, we propose that automation and further data intellectual analysis can increase productivity and effectiveness of applied method. In this work we first look at socio-economic and social aspects of issue, analyse frequently used methods and then based on specifics of autism and applied treatment we suggest basic workflow for automation of ABA method and data intellectual analysis. Workflow includes video and audio data acquisition and preprocessing for further emotion and behaviour recognition in correlation to external events. Similarly, suggested system allows monitoring and possible alteration of child's performance throughout the course of treatment. System implementation is heavily dependent on data mining and computer vision technologies, such as neural networks, clustering algorithms, video segmentation, feature extraction etc. Desired performance of overall system and its units is made possible by the function of feedback, on hardware and software level.