Identification of Javanese Alphabet Handwritting by using Histogram Chain Code

Abstract

One of the wealth of the Indonesian nation is the number of tribes with their own language and script. One of the scripts that has existed for a long time before the independence of the Indonesian state is the Javanese script, with the use of Latin script which is used by almost every aspect of life, whether official activities or daily use, the use of traditional scripts, especially Javanese script, is increasingly rare. making it easier to learn Javanese script requires learning media with the ability to recognize Javanese characters. In this study, pre-processing is used, especially feature extraction using the Histogram Chain Code (HCC) method and classification using an artificial neural network using the Multi Layers Perceptron method. This study compares four research models by setting the number of HCC feature extraction parameters obtained from one intact image and 3 divided images of 4, 9 and 16 parts respectively so that the number of parameters of each HCC model is 8, 32, 72 and 128 parameters. characteristic. The training and testing process with the Multi Layers Perceptron method uses 2000 Javanese handwritten image data which is divided into 80%, namely 1600 images for the training process and 400 images for the testing process. This study resulted in different accuracy, namely 57%, 78%, 83% and 76%. The best accuracy is obtained from the HCC model with 72 parameters and the image is divided into 9 parts.