Revolutionizing Dysgraphia Detection: Combining Feature Fusion with Non-Discriminatory Regularization

Dysgraphia, a learning disability that significantly impairs writing abilities, poses considerable diagnostic challenges due to its varied symptoms. Traditional manual assessments are prone to bias and inefficiency, necessitating advanced diagnostic methods. While recent AIbased models have been proposed to address this issue, many rely on traditional machine-learning techniques, limiting their effectiveness. This study introduces a novel approach combining feature fusion and Non-discriminatory regularization, inspired by leading methodologies in the field. By replacing the soft voting mechanism with a 5-layer neural network in the proposed ensemble model, our model achieves a remarkable accuracy of 99.68 %. This innovative framework integrates diverse handwriting characteristics for robust analysis, enhancing the accuracy and efficiency of dysgraphia recognition. Furthermore, the proposed model is suitable for deployment on web-based platforms, enabling real-time analysis and remote diagnostics for dysgraphia across educational and healthcare settings. Our results demonstrate significant improvements over existing models, offering a more reliable, scalable, and timely method for the identification, intervention, and personalized recommendations for dysgraphia in children. Future directions include incorporating longitudinal tracking of handwriting changes and providing customized interventions based on user data, ensuring more tailored support for individuals with dysgraphia.