PENGENALAN POLA SUKU KATA AKSARA BIMA DENGAN BARIS TANDA BUNYI MENGGUNAKAN EKSTRAKSI CIRI MOMENT INVARIANT DENGAN METODE ANN

The Recognition of Bimanese Script with Diacritics Syllable Pattern using Artificial Neural Network from Feature Extraction Based on Invariant Moment

Abstract

The Bimanese script is one of the archipelago's cultural heritage that needs to be preserved. The problem arose when some Bimanese people doubted the existence of the Bimanese script. Therefore, it is essential to safeguard the Bimanese script and learn the Bimanese script starting from reading and then understanding the letters. After that, add a line of sound marks to entirely understand the Bimanese script's meaning. This study aims to build an Artificial Neural Network (ANN) model to recognize the Bimanese Script Syllable Pattern with Sound Sign Lines by using Moment Invariant feature extraction. Before doing the training, first, determine the parameters on the ANN using the Tuning Hyperparameter, in the test, using a dataset of 2250 images of the Bimanese script. Based on the results of the tests carried out based on the optimal parameters, the accuracy is 77.59%, precision is 78.44%, recall is 77.61%, and F1-Score is 77.33%. Then for testing using K-Fold cross-validation, the best results were obtained using K = 9 with a ratio of 8:1 where the resulting accuracy was 79.74%. Overall the results of this study are expected to preserve the Bimanese script and are developed more widely.