Letter Tracing: a Serious Game to Teach Handwriting and Assess Proficiency through Machine Learning.

Abstract

This work introduces the "Letter Tracing", a serious game designed to teach correct letter formation. Co-designed through collaboration with clinicians and technicians, the game underwent testing with 9 first- and 26 second-grade children to assess its ability to capture the evolution of handwriting abilities over time and handwriting proficiency. In the game, children were asked to trace the initial two letters of 12 words following a demonstration of correct movements, in a gamified environment. A set of indicators derived from raw executions was computed and utilized to train diverse machine learning models, predicting both the grade and the risk of a handwriting delay as evaluated by a handwriting fluency test. The classification yielded an accuracy of 71% in predicting the grade and 71% in predicting the risk of handwriting delay. These results hold promise for the game's potential as a training tool, as it effectively models the maturation of children's handwriting and their proficiency.