Printed and Handwritten Kannada Numerals Recognitionusing Directional Stroke and Directional Density With KNN

Abstract

Abstract-In real life applications number of document contains printed as well as handwritten numerals in a single document. The process of recognition of such mixed numeral with respect to single OCR is complicated task. In this paper, we present a novel method for recognition of printed and handwritten Isolated Kannada numerals using single OCR system. We considered Directional Stroke and Directional density based feature for recognition system is proposed. The proposed system extracts Directional stroke on various angles and Directional density/profile on four side of the numeral image. Further, the stroke and density based extracted feature is feed for recognition systems. A Euclidian distance criteria and k-NN classifier is employed to classify the numeral class. A total 5000 numeral images, which includes 4000 for handwritten images and 1000 for printed images considered for experiments and overall accuracy found to 98.04%. The novelty of the proposed method is thinning free, and without size normalization.