The Dynamics of Handwriting Improves the Automated Diagnosis of Dysgraphia
Konrad Zolna, Thibault Asselborn, Caroline Jolly, Laurence Casteran, Marie-Ange Nguyen-Morel, Wafa Johal, Pierre Dillenbourg
The handwriting disorder (called dysgraphia) is far from a singular problem as nearly 8.6% of the population in France is considered dysgraphic. In addition, research emphasises the fundamental importance of detecting and treating these writing difficulties as early as possible because they can affect a child's entire life, with deteriorating performance and self-confidence in a wide variety of school activities.
Currently, a standard test called BHK detects writing difficulties. This detection, carried out by therapists, is difficult because of its high cost and subjectivity. We present a digital approach to identify and characterise handwriting difficulties using a Recurrent Neural Network (RNN) model. Each child is asked to write all the letters of the alphabet as well as the ten numbers on a tablet. Once the exercise is complete, the RNN provides a diagnosis within seconds. It is remarkably effective in correctly identifying over 90% of children who have previously been diagnosed with dysgraphia using the BHK test. The main advantage of our tablet-based system is that it captures the dynamic characteristics of writing, which an expert, such as a teacher, is unable to do.
The following shows that the integration of dynamic characteristics, accessible through the use of a tablet, are beneficial for our digital test because they help to determine the children with dysgraphia versus those without.
