Pengenalan Aksara Jawa Tulisan Tangan dengan Menggunakan Ekstraksi Fitur Zoning dan Klasifikasi K-Nearest Neighbour

Abstract

Various studies on traditional script recognition continued to be developed using various methods. One of them is handwritten Javanese script recognition. This research aims to determine the accuracy of the Zoning Feature Extraction and K-Nearest Neighbour Classification method. The data used in this this research are handwritten Javanese script from 20 different peoples. Each of character images will be transformed into 120 x 120 pixels dimension and will undergo the thinning method. The feature extraction method used is the combination of the zoning method ICZ-ZCZ with the number of zones are 4, 6, 8, 9, 10, 12, 15, 16, 18, 20 and 24. K-Nearest Neighbour is used as the classifier with k values are 1, 3, 5 and 7. The highest accuracy was obtained on 12 zones with k = 1 on K-Nearest Neighbour with a value of 71.5%.