RECOGNITION OF HIRAGANA JAPANESE HANDWRITING CHARACTERS USING SUPPORT VECTOR MACHINE AND SCALE INVARIANT FEATURE TRANSFORM

Abstract

The abundance of characters in Japanese Hiragana, the similarity in character shapes, and the lack of familiarity among the public with Hiragana in daily life make it difficult to learn. People tend to be more accustomed to romanized writing (alphabet) than specific characters, leading to difficulties in understanding Hiragana with its various sizes and shapes. This research aims to develop an effective and systematic Japanese Hiragana handwritten recognition system using Support Vector Machine (SVM) and Scale Invariant Feature Transform (SIFT) methods. The research methodology includes problem identification, literature review, data collection, data preprocessing, system design, implementation, and evaluation. The obtained data undergo augmentation and image preprocessing processes to create a larger variety and amount of data. Furthermore, feature extraction is performed on the data using the SIFT method before training the model using SVM. The research results show that the SVM-SIFT model achieves an accuracy of 0.928261, which is superior to the SVM model without SIFT with an accuracy of 0.389130. The best CV score for the SVM model without SIFT is 0.7746709410609622. Testing proves that the use of SVM SIFT is effective for classifying handwriting that varies in shape and size.