easy to understand, 2) Natural and easy to understand, 3) Unnatural, but can understand, 4) Very unnatural and difficult to understand, 5) Very unnatural, and unlike human motion. The last two columns in Table 1. show the average natural degrees of motion and hand shape of all the signs.
We described a new method to encode and synthesize JSL. We also performed an experiment to evaluate how correctly a hearing-impaired person is able to understand the sign-language animation. The obtained results from five hearingimpaired subjects and a translator show that the synthesized animation has a 90.2% average recognition rate. In addition to the recognition rate, the quality of animation is also an important factor for transmitting the information to the signers. Although the natural degree depends on the personal feeling of subject, the experimental results give a 4.07 and 4.08 score (maximum is 5.0) on average. These results demonstrate that this is an effective method that can be used to accurately synthesize JSL.
Our future work will focus on the synthesis of continuous sign sentences based on this method. Furthermore, we hope to make the sign-language animation more natural by integrating the facial expressions and the body gestures.
Kurokawa, T. ( 1992). Gesture dictionary and its application to Japanese-to-Sign language translation. Proc. the 3rd Advanced Symposium on Human Interface, 51-58.
Lu, S., Imagawa, K., & Igi, S. ( 1997a). An Active Gazing-Line Generation System for Improving Sign-Language Conversation. Proc. 7th Int. Conference on Human-Computer Interaction, 283-286.
Lu, S., Igi, S., Matsuo, H., & Nagashima, Y. ( 1997b). Towards a Dialogue System Based on Recognition and Synthesis of Japanese Sign Language. Proc. International Gesture Workshop, Bielefeld, Germany, 259-271.
Nagashima, Y., Kamei, S., & Sugiyama, Y. ( 1996). Sign Language Animation System of Morpheme Information Driven Type. Technical Report of IEICE, ET96-86, 73-78.
Ohki, M., Sagawa, H., et al. ( 1994). Pattern recognition and synthesis for sign language translation system. Proc. ASSETS '94.