Recognition of Bisindo Alphabets Based on Chain Code Contour and Similarity of Euclidean Distance

Abstract

In Indonesia, there are two forms of sign language practiced in the community, i.e., Indonesian sign language or known as BISINDO, and Indonesian sign language system or known as SIBI. In this study, we conduct research about recognition of Bisindo alphabets using contour chain code for the method of feature extraction and similarity of Euclidean distance for the method of recognition. The features used are the probability of chain code generated from contour following and the formation of chain code. The proposed method in this study consisted of five section, i.e., input test image, segmentation, edge detection, feature extraction and matching process of the alphabet. In the testing of the proposed method, we used 52 images of hand gestures used as test images. The images are in the form of static images and 26 images of hand gestures used as reference images which represent 26 alphabets BISINDO from A to Z, where the images stored in the database. The test images of different shapes and sizes with image references. For recognition, we do the matching between the probability of the test image chain code with the probability of the reference image chain code using Euclidean distance. The measurement result of Euclidean distance in this study was generated average accuracy rate of similarity above 94%. This indicates that the method proposed in this study was effective and produce the level of similarity of BISINDO alphabets was accurate.