A framework for dysgraphia detection in children using convolutional neural network

Abstract

Dysgraphia, a writing disorder in which any human may have difficulty in his writing at any level such as unrecognised letters/numbers or slow writing. This handwriting disorder is mainly observed among 10%-40% of school children. In present scenario, dysgraphia is diagnosed by the medical practitioners by analysing the person's written document and staff's impressions. Such diagnosis mechanism is very time consuming and may result in the undiagnosed dysgraphia when a child is having mild symptoms. Many researches have been conducted for the early diagnosis of the dysgraphia using various machine learning algorithms such as decision tree, random forest and support vector machine, etc. In this work, a novel framework using the concept of convolutional neural network is proposed for the accurate detection of dysgraphia. Further, the proposed model is tested on a self-created dataset including hundreds of handwriting images and performs well in terms of accuracy, recall, precision and F1-score.