Machine Learning for Recognition of Handwritten Devanagri Character

A framework known as optical character recognition is capable of converting handwritten or printed pictures into machine-editable structures. Several Indian dialects, including Hindi, Nepali, Marathi, Sindhi, and others, use the Devanagari script. In India, Hindi is the national language and the language that is spoken by most people. This script organises the development of the language. Nowadays, people are interested in gathering data in sophisticated configurations that are available in paper archives and then employing a search process to reuse this data later. In this study, we provide a novel approach to Devanagari script character recognition for printed Hindi characters. The principal objective of this research is to identify the consonants and the vowels, which can subsequently be used for recognizing the complicated inferred words. The primary focus of this project is on the identification of each individual consonant and vowel. Several pre-processing techniques, including feature extraction, segmentation, and classification, have been researched and used in this project to create an advanced Hindi OCR system. The K-NN classification technique was used in earlier research; however, in the work that is being suggested, we have employed a hybrid technique that combines neural networks with K-NN. The recognition rate of the suggested approach is 97.4<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, whereas that of the existing techniques is 94.5%. This suggests that the approach which proposed here is superior to techniques employed in the method which already exist.