Sistem Pengenalan Pola Karakter Huruf Korea Menggunakan Metode Principal Component Analysis Dan Jaringan Syaraf Tiruan - Back Propagation

Abstract

This paper describes about a recognition sistem for Korean characters into Latin form by using Principle Component Analysis Method and Artificial Intelligence - Back Propagation Approach. This sistem uses image of Korean characters as input data with 65x65 pixel of original image size and is processed by image preprocessing in the form of pixel size conversion into 15x15 pixel binary image. Every image is then extracted to produce an image feature. The features are processed first using Principle Component Analysis to reduce image feature before they enter classification stage by using Artificial Intelligence - Back Propagation Approach. This study uses 10 sample data of Korean vowel letters, obtained from 25 different font types and each font consists of normal and bold sample. Total data reaches 500 samples divided into 70 training data and 30 testing data. Architecture of this artificial intelligence uses 3 hidden layers . Each hidden layer consists of 20, 20 and 5 neurons, one output neuron. The result of this sistem research is obtained an accuracy of 95%.
Keyword : Korean Letters a.k.a Hangul, Principle Components Analysis, Artificial Intelligence - Back Propagation