Optical Character Recognition for Hangul Character using Artificial Neural Network

Abstract

Korean language is one of the languages that are becoming widely known in the world, along with the occupation of Korean music (K-POP). Hangul is a character used to write Korean, which is not like Latin, which is relatively more easily understood by the majority of people in this world. This research aims to analyze the performance of an Artificial Neural Network (ANN) in recognizing Hangul characters with a simplified optical character recognition (OCR) method. The OCR process is carried out by entering characters in a 15x13 tile area, then the characters that enter fully on the tile will be changed to value 1 while others become 0 values so that a binary image is generated. The next step is the character pattern crafting process towards the field. The results were recognized by ANN, in an experiment using four types of training data composition: testing, namely 50%: 50%, 60%: 40%, 70%: 30%, and 80%: 20%. The dataset used is 40 Hangul characters in which there are 10 sample data each, so in total there are 400 data. Based on testing, the highest accuracy is produced with a composition of 50%: 50% where the accuracy is 97%.