Detecting Children’s Fine Motor Skill Development using Machine Learning

Children’s fine motor skills are linked not only to drawing ability but also to cognitive, social-emotional, self-regulatory, and academic development Suggate et al. Journal of Research in Reading , 41(1) , 1–19 ( 2018 ), Benedetti et al. ( 2014 ), Liew et al. Early Education & Development , 22(4) , 549–573 ( 2011 ), Liew ( 2012 ) and Xie et al. ( 2014 ). Current educators are assessing children’s fine motor skills by either determining their shape drawing correctness Meisels et al. ( 1997 ) or measuring their drawing time duration Kochanska et al. ( 1997 ) and Liew et al. ( 2011 ) through paper-based assessments. However, these methods involve human experts manually analyzing children’s fine motor skills, which can be time consuming and prone to human error or bias Kim et al. ( 2013 ) and Lotz et al. ( 2005 ). With many children using sketch-based applications on mobile devices like smartphones or tablets Anthony et al. ( 2012 ), computer-based fine motor skill assessment has the potential to address limitations of paper-based assessment by using automated measurements. In this work, we introduce a machine learning approach for analyzing aspects of children’s fine motor skill development. We performed a study with 60 young children (aged 3 to 8 years old), and we implemented classifiers that determine children’s age category based on features related to fine motor skill, predominantly for curvature- and corner-based drawing skills, surpassing the performance of our previous work Kim et al. ( 2013 ) and of human evaluators. We also present dedicated discussion and statistical testing of sketch recognition features which will further enhance automated fine motor assessment.