OCR HURUF JAWA DENGAN FITUR KODE RANTAI DAN LEVENSHTEIN DISTANCE

Abstract

A simple application for the introduction of Javanese characters is needed when the massive use of national characters is needed. Applications are needed to recognize and read characters from Javanese characters in received messages. Simple features and methods to recognize Javanese characters are needed so that the application can do training at any time not too much training data. One of the simpler features available is chain code. The problem is that the chain code is a string of characters, it requires an appropriate distance measurement, not the distance measurement commonly used in OCR. Levenshtein Distance is a measurement of the distance between two strings. In accordance with the selected feature, the levenshtein distance will be used to measure the distance of two chain codes. In this study, it will be proven that the chain code feature with distance measurement using levenshtein distance is able to recognize Javanese script objects based on stored training data. The classification method used is the nearest-neighbor classification. The choice of this method is intended to be able to provide training data information that is closest to the object you want to recognize. The results showed that the chain code feature with distance measurement with levenshtein distance was able to recognize the training data feature that was closest to the object feature.