Comparative Result on Bangla, Arabic and Gurumukhi Numerals Recognition

Gita Sinha, Ashif Habibi, Dr. Pankaj Kumar Chaudhary

Abstract: In this paper we present the recognition accuracy of Bangla, Arabic and Gurumukhi numerals and compare their result using different parameters like gamma and alpha. We describe two feature extraction techniques and two classifier for recognition of offline numerals. We have got more accurate result on these experiments using SVM classifier. Images are divided in 8*8, 16*16, 32*32 according image size feature vector is calculated. Highest recognition on Bangla numerals produces 98.45 %, Arabic 97.21 % and 99.73 % on Gurumukhi numerals.

Keywords: SVM-support vector machine, AN-Arabic numerals, BN-Bangla Numerals and Gurumukhi Numerals, ICZ-image centroid zone, ZCZ- zone centroid zone