A DCT and SVM based Odia Handwritten Character Recognition

Abstract

One machine can do the work what fifty ordinary men can. In today’s world advanced scientific techniques are used beyond the strength and capacity of human beings in various fields of technology. Character recognition basically known as OCR (Optical Character Recognition) is one such field of technology. Character recognition is an interesting and
challenging field in machine learning and pattern recognition area and used in numerous real life application. There are different works done in machine printed character recognition yet at the same time there is some hole display in even today in perceiving handwritten character.Consistent change is going ahead to fulfill the hole. OCR has taken an imperative part in today’s current world. It can be utilized to take care of different complex issues and makes human effort less and work easier. OCR is the technique of conversing scanned images of handwritten, typewritten text into machine readable text. Success of optical character recognition relies upon various factors, out of which two are feature extraction and classification algorithms used for recognition. In this project, I attempt to recognize Odia character using Discrete Cosine Transform(DCT).
The features extracted by the DCT are stored in an array. These coefficients are then used for classification purpose and classified using SVM classifier and compare with the other classifier. Now a days OCR (Optical Character Recognition) have a huge demand in every field of society such as industries, in postal department for sorting the letters and recognition purpose in other government office. Much more research has done on Various largely spoken
language like English, Latin alphabets but still there is some void in Indian languages. So now a day’s Indian script is an active area of interest for the researchers. Odia is one of them. There are various researches is going on for the recognition of alphabets in printed form but less con- centration is given to handwritten character. In my thesis a genuine endeavor has been made by reviewing the current works on Odia literature and to give the most recent
pattern on the recognition of Odia character. A top to bottom investigation of Odia character recognition has been carried out in my thesis by keeping an eye with the existing strategies and their constraints has been carried out and future research course in this field has been proposed.