A Novel Data Representation Method for 3D Model-Based Gesture Tracking

This paper introduces a new spline-based approach for the representation of object together with an associated search method using best-curve matching technique. The targeted object is acquired from the video stream using a stereo camera. Many image processing techniques (such as edge detection, frame differencing) are integrated and implemented to extract the edges of the moving object. This paper also presents a curve-matching technique for querying a training-database for updating (storing) and retrieving the shape data. In addition to the parameters required to represent the spline-curve, this approach uses weighting mechanism to provide probabilistic information for the variation of the knots.