Multi-Stroke Handwriting Character Recognition and Enhancing Proficiency with CNN: A Touch-Based Writing Approach

Handwriting is an essential skill, and difficulties in this area can significantly impact learning and communication. In recent years, Convolutional Neural Networks (CNNs) have evolved as a powerful tool for image recognition and analysis, making them well-suited for handwriting recognition and improvement tasks. The goal of this study is to create and build a touch-based therapy application to improve hand dexterity for handwriting preparation using the CNN model. The system leveraged a large dataset of handwriting samples and employed CNN to learn and recognize patterns in handwritten characters. The CNN model was trained using supervised learning techniques, optimizing its ability to accurately classify and analyze different handwriting styles. To assess the effectiveness of the proposed system, a sequence of tests was conducted. The trained CNN model showed promising accuracy in recognizing and analyzing multi-user free-style multi-stroke handwriting characters and providing valuable feedback to the users. Participants who utilized the system reported improvements in letter formation, alignment, and overall handwriting legibility.