Pengenalan Aksara Sunda Menggunakan Ekstraksi Ciri Zoning dan Klasifikasi Support Vector Machine

Abstract

This research aims to determine the most effective feature extraction method used in Sundanese script recognition. The data used in this research are generated in the form of image files, each of which contains a Sundanese character. Feature extraction method used is the variations of zoning method: Image Centroid and Zone (ICZ), Zone Centroid and Zone (ZCZ), and combination of ICZ and ZCZ. The number of zones used are 4, 6, 8, and 12 zones. Support Vector Machine is used as classifier, with linear, quadratic, polynomial, and RBF kernel. Among the feature extraction methods used, the hybrid feature extraction method ICZ and ZCZ obtains maximum accuracy for all the number of zones used. On the other side, the feature extraction method using 12 zones obtains the maximum accuracy for all the feature extraction methods used. From this result, it can be concluded that the most effective feature extraction method used in Sundanese script recognition is the hybrid method ICZ & ZCZ with 12 zones.