An Improved Fast Bangla Character Recognition Method of Printed Text by Constructing Character Signature Database

摘要

This paper presents an offline character recognition system to recognize Bangla printed characters using previously extracted character signatures. In this recognition system, Bangla texts are taken as image files, noise is removed, tilting angle is corrected. The text is then segmented into lines, words and then into characters. And finally, using the previously calculated character signature, the system recognizes the characters with an average success of 80%. This main strength of this paper lies on that fact that the proposed method is very much computation efficient. But is produces quite satisfactory result. The proposed method is an improvement over the previously proposed approach that uses linear segmentation of isolated characters. But is our case instead of isolated characters; we have taken noisy, rotated printed text and applied proposed non-linear segmentation. Spatial rotating algorithm is applied for the selection of reference axes considering the rotation of the signatures. Also the algorithm for creating character signature database and character recognition system is presented.