CNN Based Handwritten Odia Character Recognition

Abstract

Odia is a 44 million-strong Indo-Aryan language spoken mostly in Odisha, India. Some characters are made up of many symbols that are linked together. Because of its vast variety of applications, Handwritten Character Recognition (HCR) is critical in Optical Character Recognition (OCR) and Pattern Recognition (PR). HCR is employed in fields such as bank checks, medical prescriptions, and tax returns, and contributes greatly to the spread of automation. Handwritten characters are substantially more difficult to discern than printed characters since different people have different writing styles. Character recognition technology may be used to scan a wide range of documents, including handwritten and printed materials. This study investigates and analyses the application of deep learning techniques such as Convolutional Neural Networks for the recognition of Odia characters. CNNs categorise characters by using neurons coupled in several layers to achieve optimal efficiency, which is inspired by the structure of the brain. The goal of this work was to categorise Odia characters using 6 layers of neurons. This method was able to obtain a 95.6 percent accuracy rate. The handwritten Odia characters may be simply translated into English or any other language once they have been identified.