Automated System to Measure Static Balancing in Children to Assess Executive Function

Abstract

We present multiple methods based on computer vision and deep learning to automate the task of "Balancing on one foot". This is one of the Activate Test of Embodied Cognition (ATEC) tasks used to measure cognitive skills in children through physical activity. A dataset of 27 children performing the ATEC task is used to train and validate the deep learning models used to automate the task. As opposed to most balance identification systems that use sensors, our proposed approach relies only on computer vision which can be easily deployed at home or classroom environment, is portable, and cheap. Our proposed system automatically identifies the task and assigns an ATEC and an ergonomics score for the "Balancing on one foot"task. Our proposed system achieves an accuracy of 97% when calculating the raw score for the ATEC task and 86.5% for assigning the ergonomic score.