Gross Motor Skills Development in Children with and Without Disabilities: A Therapist’s Support System Based on Deep Learning and Adaboost Classifiers

Fundamental or Gross Motor Skills (GMS) are a set of essential skills both for basic movement activities and physical activities. Properly developing them is vital for children to develop a healthy lifestyle and prevent serious illnesses at an older stage of life, like obesity and cardio-respiratory problems. This is a problem for therapists because they must attend to many children lacking this skill set, and it’s even more time-consuming with children with disabilities. Therefore, this work presents a system that can assist therapists in giving therapy to more children with and without disabilities. To reach this goal, the system is divided into 3 phases: first, the data preprocessing phase, where images from 3 postures are collected: sitting, static crawling, and bound angle. Then all the images are resized. Model construction is the second phase. It consists of implementing the MoveNet algorithm that helps detect human posture through 17 key points of the body. Then, this algorithm is applied to the dataset created to obtain the coordinates from the postures collected. After that, an Adaboost model is created and trained, and tested. Next, the MoveNet algorithm is assembled with the Adaboost model to predict the three postures in live action. Then comes the third phase: model evaluation. This step includes evaluating the model assembled at Instituto de Parálisis Cerebral del Azuay (IPCA). Finally, the results of this evaluation are presented.