Title: Improved handwritten digit recognition using artificial neural networks

Authors: Debabrata Swain; Badal Parmar; Hansal Shah; Aditya Gandhi

Addresses: Department of Computer Engineering, Pandit Deendayal Energy University, Knowledge Corridor, Raisan Village, PDEU Rd., Gandhinagar, Gujarat, 382007, India ' Department of Computer Engineering, Pandit Deendayal Energy University, Knowledge Corridor, Raisan Village, PDEU Rd., Gandhinagar, Gujarat, 382007, India ' Department of Computer Engineering, Pandit Deendayal Energy University, Knowledge Corridor, Raisan Village, PDEU Rd., Gandhinagar, Gujarat, 382007, India ' Department of Computer Engineering, Pandit Deendayal Energy University, Knowledge Corridor, Raisan Village, PDEU Rd., Gandhinagar, Gujarat, 382007, India

Abstract: Handwritten digit recognition is one of the significant challenging problems, finding usage in fields like postal mail arranging and healthcare. Thus, it evokes the necessity for a framework that can apprehend the penmanship of all age groups with increased precision. Our proposed system uses neural networks to implement an acute number recognition system. It focuses on improving a neural network's recognition of handwritten digits by employing the MNIST digit dataset. This work examines how more appropriate optimisers can improve neural networks' general accuracy. An optimiser is an essential part that aids in tracking down the ideal arrangement of weights and their values for improving accuracy. After extensive experimentation, the model achieved recognition accuracy of 99.87% with an RMSProp optimiser.

Keywords: image classification; artificial neural networks; MNIST dataset; handwritten digit recognition; optimisers.

DOI: 10.1504/IJCSM.2023.131625

International Journal of Computing Science and Mathematics, 2023 Vol.17 No.4, pp.353 - 370

Received: 27 Oct 2021
Accepted: 07 Feb 2022
Published online: 21 Jun 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article