Hand Skeleton Graph Feature for Indonesian Sign Language (BISINDO) Recognition Based on Computer Vision

Sign language is a means of communication for The Deaf. Indonesian Sign language or BISINDO is one of the sign languages that is used in Indonesia. For The Deaf with The Deaf sign language is a means of communicating effectively, but not for The Deaf with the hearing. This is partially due to insufficient basic knowledge of The Hearing about how to communicate with The Deaf. A sign language translator needed to help The Deaf communicate with The Hearing. Limited of sign language translator is the reason for this research to develop sign language recognition methods. This research is about the development of methods for recognizing basic sign language alphabet and numbers based on computer vision. Basic sign language alphabet and numbers are demonstrated by arms, so they can be the basis to recognize alphabet and number from them. In this research skeletons graphs are extracted. Features are obtained from angle as direction for each chosen vertex. These features are known as skeletal based. To calculate similarity of the alphabet and numbers based on features, this research uses K-Nearest Neighbor (KNN). The best result of recognize sign language alphabet is 99.70% and to recognize sign language numbers the accuracy is 99.81%.