Recognition of Handwritten Hangeul Characters Using Convolutional Neural Network

Abstract

The rapid increase in Indonesian tourists visiting South Korea has highlighted a growing interest in Korean culture, largely fueled by the Korean Wave. However, the inability to read the Korean alphabet (Hangeul) leaves many tourists vulnerable to scams. This paper proposes a novel solution to address this issue by developing a system for recognizing handwritten Hangeul characters, aimed at assisting Indonesian tourists in navigating South Korea safely. The research introduces a hybrid algorithm that integrates Vision Transformers (ViTs) with Convolutional Neural Network (CNN), aiming to overcome the limitations of CNN in capturing global features. The dataset utilized comprises 2,400 images of handwritten Hangeul characters, categorized into consonants and vowels. The study involved pre-processing, training, validation, and testing with three data split ratios (60:20:20, 70:15:15, 80:10:10) and two learning rates (0.001 and 0.0001) over 10 epochs. This hybrid model approach is designed to enhance recognition accuracy and improve the system's adaptability to diverse inputs.