Taipei: Asia-Pacific Education & Research Association
Sign language is the primary means of communication between deaf people and hearing/speaking challengers. There are many varieties of sign language in different challenger community, just like an ethnic community within society. Unfortunately, few people have knowledge of sign language in our daily life. In general, interpreters can help us to communicate with these challengers, but they only can be found in Government Agencies, Hospital, and etc. Moreover, it is expensive to employ interpreter on personal behalf and inconvenient when privacy is required. It is very important to develop a robust Human Machine Interface (HMI) system that can support challengers to enter our society. A novel sign language recognition system is proposed. This system is composed of three parts. First, initial coordinate locations of hands are obtained by using joint skeleton information of Kinect. Next, we extract features from joints of hands that have depth information and translate handshapes. Then we train Hidden Markov Model-based Threshold Model by three feature sets. Finally, we use Hidden Markov Model-based Threshold Model to segment and recognize sign language. Experimental results show, average recognition rate for signer-dependent and signer-independent are 95% and 92%, respectively. We also find that feature sets including handshape can achieve better recognition result.
Proceedings of International Research Conference on Information Technology and Computer Sciences, pp.24-31