International Journal of Science, Engineering and TechnologyAuthors: Miss Pratiksha S. Kotkar, Ms. Hemlata Dakhore
Abstract: This paper presents a web-based system for handwritten digit and alphabet recognition designed to support interactive learning. The proposed approach combines Convolutional Neural Networks (CNNs) with real-time computer vision techniques to recognize user input directly within a browser environment. Handwritten input is captured through a digital canvas and processed using OpenCV.js for noise reduction, scaling, and centering, ensuring consistency with the training data. The processed input is then classified using a CNN model implemented in TensorFlow.js, enabling fast and efficient prediction without reliance on external servers. The system is developed using the EMNIST dataset, which includes both digits and alphabets, allowing it to handle a wide range of inputs. Experimental results show that the model achieves high accuracy while maintaining low latency, providing immediate feedback to users. In addition, the application includes a performance tracking mechanism that records user progress over time, supporting continuous learning. The proposed system demonstrates how browser-based artificial intelligence can be used to create accessible and responsive educational tools. By integrating handwriting practice with instant feedback, it offers a practical solution for improving basic literacy and numeracy skills in a digital environment.
DOI:
