Deep Learning-Based Monitoring System for Early Childhood Motor Skill Development

Early childhood education plays a vital role in supporting the cognitive and motor development of children aged zero to six years. However, due to limited verbal communication skills at this stage, young children are often unable to accurately report their school activities, presenting challenges for parental monitoring. To address this issue, this research proposes the development of a Deep Learning-based system capable of automatically analyzing and recording children's daily activities within children education institutions. The system aims to facilitate real-time access to developmental records by both parents and educators. This system received input in the form of images of early childhood activities from video recordings and videos in real time at school. The methodology employs BlazePose for human pose estimation and a modified Convolutional Neural Network (CNN) for classifying activities based on image datasets. The expected output is a software product in the form of an intelligent system to automatically monitor the development of early childhood motor skills. This study focused on activities classified as gross motor skills, which engage the body’s large muscle groups and facilitate movements. Examples of such activities include sitting, standing, and sleeping. Experimental results showed that the proposed system demonstrated effective performance, achieving the highest accuracy of 97.77%. Some errors occurred due to dependence on camera angles and the similarity of poses across different viewpoints.