A recognition system of sign-language sentences integrating hand gestures and head movements

摘要

Japanese sign language (JSL) is one of communication methods for hearing-impaired people in Japan. We are developing a method for recognizing JSL gestures. This method recognizes each signed word represented in the inputted JSL gestures. To translate JSL into Japanese, the relationships between signed words have to be analyzed based on JSL grammar. Most of grammatical information in JSL is conveyed by non-manual signs (NMSs) such as facial expressions, glances and nods. However, the functions of NMSs are not clear in JSL. We focused on head movements among the NMSs in JSL and analyzed their grammatical functions. In this paper, we will describe the result of the analysis. And we propose a method for recognizing JSL sentences that analyzes the relationships between signed words based on the functions of head movements.