Dyslexia deep clustering using webcam-based eye tracking

Abstract

Dyslexia is a neurodevelopmental impairment that causes difficulties in reading and can have significant academic, social, and economic impacts. In Morocco, Dyslexia accounts for 37% of children's school failures. Early detection of dyslexia is crucial to help children reach their academic potential and prevent low self-esteem. To address this issue, a dyslexia screening tool using webcam-based eye tracking was developed for the Arabic language. The tool was tested on a dataset of 61 students from three primary schools in southern Morocco, and the results showed that using Arabic dyslexic-friendly typefaces improved reading performance, particularly for those with low reading performance. Deep clustering was used to reduce the dimensionality of the dataset, and the subjects were gathered using unsupervised k-means based on AutoEncoder output. The three clusters produced showed a significant difference in many dyslexia traits, such as the number and duration of fixations, as well as the saccade period. These findings suggest that webcam-based eye-tracking techniques have the potential to be used as an initial dyslexia diagnosis tool to assess if a child exhibits some of the typical symptoms of dyslexia and whether they should seek a professional full dyslexia diagnosis.