An Approach for Handwritten Alphanumeric Character Recognition: Leveraging CNN for Accurate Recognition

The classification of alphanumeric English hand-written characters is a challenging task. This paper presents a novel approach for the recognition of alphanumeric English handwritten characters using Convolutional Neural Network (CNN). To create a comprehensive and diverse dataset, three separate data sets namely MNIST, KAGGLE, and EMNIST, were combined, so that the proposed model can able to recognize numerals, capital and small letters of English language. The resulting dataset contained a total of 555,249 data samples, encompassing a wide range of handwritten characters. Prior to model training, preprocessing techniques were applied to normalize the intensity values of the pixel data, ensuring consistency across all samples. Additionally, one-hot encoding was employed to represent the class labels, enabling effective handling of the multi-class classification problem. The proposed model was trained on 70% of the data samples, from remaining 15% allocated for validation and an additional 15 % for testing to evaluate its performance. The experimental results demonstrate the effectiveness of the CNN model in classifying alphanumeric English handwritten characters with an accuracy of more than 92%.