Real-time gesture recognition using eigenspace from multi-input image sequences

✍ Scribed by Takahiro Watanabe; Masahiko Yachida

• Publisher: John Wiley and Sons
• Year: 1999
• Tongue: English
• Weight: 556 KB
• Volume: 30
• Category: Article
• ISSN: 0882-1666
• DOI: 10.1002/(sici)1520-684x(19991130)30:13<61::aid-scj7>3.0.co;2-u

✦ Synopsis

This paper discusses real-time gesture recognition from moving images that is efficient for human-friendly manmachine interaction. With the proposed method, the eigenspace is built for multi-input images taken from different directions; in so doing, 3D data can be acquired without such problems of stereo methods as those of correspondence or calibration. Besides, the new method supports real-time recognition of such gestures that cannot be processed using single input images because of self-occlusion or other complications. In addition, 3D interpretation ensures more accurate recognition of gesture parameters (speed, size). The proposed method was applied to a realtime interactive system (virtual conducting system), and proved to be efficient.