Loading...

Hand Posture Recognition through Stereo Vision System

Sangi, Mehrdad | 2010

469 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 40135 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Jahed, Mehran
  7. Abstract:
  8. The mankind has always utilized vision as the most essential tool in evaluating his environment. Currently, one of the most important research areas in computer vision is to characterize a three dimensional object through its features in a natural scene. This task may be conducted through a stereo vision system. The objective of this study is to introduce a quantitative measure in order to evaluate the operation of an automatic recognition system for selected hand postures. A key application of our proposed system is to train a prosthetic hand through surface electromyogram (sEMG) signals. At the onset of this process, the patient conducts a certain posture through application of sEMG signal. The posture is next evaluated through the stereo vision system which is appropriately associated with a given class of movement or posture of the hand. The advantage of this method stems from its ability to provide a quantitative error measure to ultimately eliminate therapist errors. In this work, we introduce a 3D hand model reconstruction method which offers flexible and elaborate representation of hand gestures. We used 20 landmarked points on tips and joints of the fingers and calculated the 3D coordinates of these points through a stereo vision system. Our results show that such reconstruction provides a precise 3D hand model only to be influenced by intrinsic and extrinsic camera parameter estimation errors. As our proposed 3D reconstruction method requires only 20 points, it is rated among very fast algorithms suitable for real-time hand gesture recognition applications. Finally we introduce a measure for comparing two 3D models for hand postures using Iterative Closest Point (ICP) registration that determines the mapping between these models by computing the translation and rotation matrices.

  9. Keywords:
  10. Camera Calibration ; Three Dimensional Reconstruction ; Stereo Vision ; Hand ; Matching ; Iterative Closest Points

 Digital Object List

 Bookmark

No TOC