INDONESIAN SIGN LANGUAGE (BISINDO) RECOGNITION USING ORB FOR BILINGUAL LANGUAGE

• Main Author: Arrizal P, Rahmatullah
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36877
• Institution: Institut Teknologi Bandung

Sign language refers to a language which is used by some society as a means for daily communication. Unfortunately, there are still many general society who still do not understand about Sign language so that it limits the communication process. In Indonesia, there are two sorts of sign languages which have been used. They are SIBI and BISINDO. BISINDO is more widely used because it is more naturally used by the deaf in Indonesia. This research proposed a sign language recognition method by using computer vision which can be translated to bilingual language. The data used is alphabet gesture 26 types from A to Z. For extracting features form gesture used ORB as detector-descriptor has strengths of computation process which is fast and also invariant with measurement and rotation, therefore it gives the possibility of the input process of image which was tested from video. The structure which was applied in building vocabulary in translation process is Binary Search Tree with the strengths of the fast computation process. The training data used in this system is 1x26 and the testing data 3x26 images. The data is tested on three types of images, there are grayscale, binary, and the result of edge detection. The results of the testing of data from 3x26 images have a test of 98.71% for grayscale images, 98.71% for binary images, and 93.58% for edge detection images. For testing on streaming camera results 16 letters can be identified in the grayscale image, 21 in the binary image, and 15 in the image resulting from edge detection.