The Robotics Institute

Carnegie Mellon Robotics Institute

When a robot recognizes human behaviour and tries to perform the same behaviour afterward, a human behaviour model which is highly applicable in different environment is required. In this paper, we limits the task field to hand operation and describe a novel method of constructing a behaviour model of hand operation using data-glove and HMM, which abstracts human hand action from four attributes ``which hands'', ``state of grasping'', ``hand motion'' and ``relative position of target objects and hands in 3D space'' along the time series. We also describe a method of performing the same human operation by robot with 9eye stereo vision system and 3D template matching, and finally we show the experimental results using a real robot.

Keywords

Gesture Recognition, Humanoid, Stereo Vision, Skill Modelling

Notes

Associated Center(s) / Consortia: Vision and Autonomous Systems Center

Text Reference

Koichi Ogawara, Soshi Iba, Hiroshi Kimura, and Katsushi Ikeuchi, "Recognition of Human Behavior with 9eye Stereo Vision and Data Glove," Computer Vision and Image Media, March, 2000.

BibTeX Reference

@inproceedings{Iba_2000_3280,
   author = "Koichi Ogawara and Soshi Iba and Hiroshi Kimura and Katsushi Ikeuchi",
   title = "Recognition of Human Behavior with 9eye Stereo Vision and Data Glove",
   booktitle = "Computer Vision and Image Media",
   month = "March",
   year = "2000",
}

The Robotics Institute

Menu