Bima script handwriting pattern recognition with gray level co-occurrence matrix feature extraction and zoning & classification of probabilistic neural networks

Abstract

The Bima's Script, natively known as Aksara Bima is one of Indonesia's local heritages that needs to be preserved. Based on an online questionnaire of 81 respondents from Bima, 66.7% of people were not familiar with Bima's Script and 45.7% of people did not even know the existence of Bima's Script. There are various ways of preserving the local script in Indonesia, such as the creation of learning media like games, transliteration, and Script pattern recognition. This research aims to build a machine learning model that is able to recognize the Bima Script's handwriting pattern by using the Gray Level Co-occurrence Matrix (GLCM) feature extraction combined with Zoning and Probabilistic Neural Network (PNN) classification. The best model obtained accuracy up to 81.35% by applying 4x4 zoning size and the 75% of the dataset as training data.