Imitation Lab

  
ImitationLab will develop a computer vision based motion capture system for modelling, analyzing, and imitating human motion (HUMIM).
  

Human Motion Imitation

  
HUMIM consists of a theoretical core, which will be used in the development of prototype systems for two different application areas: Rehabilitation training of arm and hand impairment and robot grasping instruction. Both applications include the core aspects of modelling and imitating human motion. The theoretical developments in HUMIM include visual tracking combined with a generative 3D human motion model describing the statistical variation of human motion. This motion model will be learnt from observations and adapted to the specific person / robot performing the motion. Hence, learning takes place both of the teacher¡¯s motion as well as of the student performing imitation. We consider imitation as the problem of mapping motion models between configuration spaces, e.g.\ mapping from instructor to robot or physiotherapist to patient. After mapping, motion may be imitated by synthesis from the generative motion model. The prototypes developed as part of HUMIM will be evaluated in end-user environments.