Machine Learning Methods for Dysgraphia Screening with Online Handwriting Features

Abstract

Dysgraphia, a major learning disorder that primarily interferes with writing skills can hinder the academic track of children unless recognized in the early stage. The diversity in the symptoms, as well as the emergence in different ages, makes the diagnosis quite an intricate task. This work proposes automated methods that can be used for the diagnosis of dysgraphia by analyzing handwriting. Particularly this work examined the effectiveness of kinematics and dynamics of handwriting for discriminating abnormal writing. Furthermore, this work focused on developing methods that utilize fewer features for classifying dysgraphic and non-dysgraphic subjects. The proposed methods are evaluated in a publicly available online handwritten dataset. Obtained results indicate that the proposed method can diagnose the existence of dysgraphia with an accuracy of 77% with a limited number of features.