Handwritten English Alphabet Recognition Using Convolutional Neural Network

Abstract

This research paper presents an approach for English alphabet recognition using machine learning. The proposed system utilizes a convolutional neural network (CNN) to identify individual characters within an input image. The dataset used in this research consists of a large collection of handwritten alphabet images, sourced from Kaggle's A-Z Handwritten Alphabets dataset in CSV (comma-separated values) format, which were preprocessed and augmented to improve the model's accuracy. We trained and tested our model using this dataset and achieved a validation accuracy of 97.9% and a training accuracy of 95.8%. The validation loss was 0.0745, and the training loss was 0.1517. Our system outperformed previously published methods for English alphabet recognition. We also tested our model on a real-world data set containing handwritten letters, and the results were promising, with an accuracy of 90%. The proposed system has various applications, including optical character recognition, document digitization, and automated handwriting analysis.