ABSTRACT

A mixer of printed and handwritten numerals may appear in a single document such as application forms, postal mail, and official documents. The process of identifying of such mixed numerals and sending it to respective OCRs is a complex task. In this paper, we present a novel method for recognition of printed and hand written isolated Kannada numerals using single OCR system. Printed/hand written Kannada numeral is scan converted to binary image and normalized to a size of 40 x 40 pixels. The boundary of the numeral is traced and chain code of the image is determined. These codes are represented in a complex plane and 10 dimensional Fourier descriptors are computed that form the feature vector. The 10 dimensional Fourier descriptors are input to multi-class SVM classifier to recognize the numeral class. The proposed algorithm is experimented on 5000 numeral images consisting of handwritten and printed numerals, each of size 2500. The experiment is carried using five-fold cross validation method and yielded recognition accuracy of 97.76%.