Gujarati Character Recognition

Abstract

This project deals with an optical character recognition (OCR) system for Gujarati handwritten characters and handwritten numerals. One may find so much work for Indian languages like Hindi, Kannada, Tamil, Bangla, Telugu, Malayalam etc, but Gujarati is a language for which hardly any work is traceable. A large database of 1, 28,000 characters and 14,000 Numerals is generated for handwritten Gujarati characters and numerals respectively. Gujarati, an Indo-Aryan language is spoken by about 65 million people in the Indian states and across the world. These days recognition of Gujarati handwritten characters and numerals is quite challenging area for research, because of majority of Gujarati characters and numerals are very similar in appearance. In proposed work Preprocessing includes binarization, resizing and thinning. Feature extraction is crucial step for classification of an object. The features of Gujarati characters and numerals is abstracted by different feature extraction methods such as profile based feature, signature feature, boundary rectangular feature, primary stroke, chain code and zoning feature extraction methods. The objective is to identify Gujarati handwritten characters and numerals with the help of Naive Bayes Classifiers. The highest accuracy is achieved with chain code feature extraction method as compare to other features. The suggested method is simple and has promising discrimination accuracy. The outcome of this research work is development of software algorithm for efficient recognition of handwritten Gujarati characters and numerals.