Gyroscope-Based Game Revealing Progress of Children with Autism

The paper concerns the automation of measuring progress of children with autism spectrum disorder. The proposed approach combines diverse approaches: e-technologies and mobile applications for autism, behavioral metrics derived from gyroscope and game state with machine learning methods to find interconnections between the metrics and the progress of a child. The paper presents a gyroscope-based game, specifically designed as an investigation tool for therapy progress monitoring. The game enables registration of behavioral patterns of use of the applications and tablet. The paper presents how the game was used in a study of behavioral metrics. 31 children with autism took part in the study. Each of them played the game several times during a 6-months period. The data gathered during the gameplay are used to calculate a set of metrics, that might be used in evaluation of a child's progress. Results in terms of classification accuracy reach 80%, however they depend on the particular skill category. The best accuracies are obtained for evaluation of stereotypic behaviors and gross motor skills of a child. The approach presented in the study is novel and was not applied before, therefore it might be interesting for other researchers working on supporting technologies for autism. The results might be also interesting for practitioners applying e-technologies in autistics therapy.