Recognising of repetitive and stereotyped movements for children with autism spectrum disorder. (3rd February 2021)

• Record Type: Journal Article
• Authors: Jazouli, Maha; Ezghari, Soufiane; Majda, Aicha; Zahi, Azeddine; Aalouane, Rachid; Zarghili, Arsalane
• Is Part Of: International journal of bioinformatics research and applications. Volume 16:Number 4(2020)
• Page Start: 355
• Page End: 372
• Publication Date: 2021-02-03
• Subjects: autism spectrum disorder -- stereotypical motor movements -- stereotyped behaviours -- Kinect sensor -- gesture recognition -- machine learning; Bioinformatics -- Periodicals; 570.285
• Languages: English
• ISSNs: 1744-5485

Abstract

People with autism often engage in stereotyped and repetitive motor movements. Hence, our aim is to put out a smart video surveillance system that facilitates the diagnosis of autism for doctors. In this respect, we propose an automatic stereotypical motor movement detection system in real time. Firstly, we use the Kinect sensor to monitor the child's with autism movements. Secondly, we propose a data integration process to make the provided data from Kinect sensor more comprehensive and specific. Thirdly, we perform the gesture detection by using the well know machine learning algorithms such as decision tree, artificial neural network and nearest neighbour. The obtained result is very promising and shows that the data integration step enhances the gesture recognition. The experimental results show that our system can achieve above 99.8% recognition rate. Also, we evaluate the efficiency of different machine learning algorithms in recognition tasks.