Classification of characters and grading writers in offline handwritten Gurmukhi script

Abstract

Grading of writers based on their handwriting is a complex task mainly because of various writing styles of different individuals. In this paper, we have attempted grading of writers based on offline Gurmukhi characters written by them. Grading has been accomplished based on statistical measures of distribution of points on the bitmap image of characters. The gradation features used for classification are based on zoning, which can uniquely grade the characters. In this work, one hundred different Gurmukhi handwritten data sets have been used for grading the handwriting. We have used zoning; diagonal; directional; intersection and open end points; and Zernike moments feature extraction techniques in order to find the feature sets and k-NN, HMM and Bayesian decision making classifiers for classification.