On-line dynamic hand gesture recognition system for the Korean sign language (KSL)

Abstract

Human-hand gestures have been used as a means of communication among people for a long time, being interpreted as streams of tokens for a language. They may vary from the stylized lexicon of a traffic cop to the highly developed syntax of a natural language such as the sign language. The sign language which include many hand gestures is a method of communication for Deaf persons born with profound hearing loss. It is understood by means of gestures of both hands and fingers. The Deaf persons dependent on the KSL is their inability to communicate with normal people in their society. Usually, they must rely on interpreters (normal people) who are fluent in the KSL. It is very difficult to learn the KSL and takes a long time to express the KSL fluently. And these interpreters are few and far between and not easily available. Successful rehabilitation of the Deaf persons in a hearing society must provide a mechanism by which meaningful communication between the Deaf persons and hearing people is possible. For this to happen, one must allow hearing people to use spoken language and the Deaf persons to use the KSL. This requires a speech and sign recognition system, that is, KSL interpretation system, a visual sign display to show signs to the Deaf persons, and a synthesized speech system to provide speech to the hearing person. Recent technological advances allow researchers to begin research on some of the problems associated with building an automatic communication system which can provide communication between the Deaf persons and hearing people. Considerable work is in progress in Korean to KSL sign generation, and in progress in automatic speech recognition. Recognition of the KSL, which is the topic of this thesis, has been attempted. This thesis deals with a system which recognizes the Korean Sign Language (KSL) and translates it into a normal Korean speech. The recognition of changing patterns of dynamic gestures in the time domain is essential to understand...