Optimizing Writing Skills in Children Using a Real-Time Feedback System Based on Machine Learning

Writing is a fundamental skill in children's educational development, but many face significant difficulties in learning it, which can affect their academic progress and self-esteem. Traditional handwriting assessment methods are based on manual observations and are often subjective, providing delayed and limited feedback. This work proposes a real-time feedback system based on Convolutional Neural Networks (CNNs) designed to detect and correct abnormal patterns in children's handwriting as they write. The system preprocesses handwriting data using techniques such as Discrete Wavelet Transform (DWT) and Kalman filtering to refine signal quality. Then, CNNs are trained to analyze writing strokes in real-time, providing immediate visual and verbal corrections that guide students in improving their writing. The results demonstrate a significant improvement in several key metrics: a 20% increase in letter precision, 18% in stroke consistency, and 11% in writing speed compared to a control group. In addition, a reduction in applied pressure of 0.8 Newtons was observed, indicating better pencil control. These findings underline the system's effectiveness in improving children's writing skills, providing a robust and accessible tool for primary education. The study highlights the importance of real-time feedback and its potential to transform educational practices. It offers an innovative approach that combines technical precision with ease of use in an academic environment.