3D Hand Motion Tracking Using Improved Hidden Markov Model of Behavior

We present a novel approach for 3D hand motion tracking of stage characteristics as observed in cognitive psychology. A Hidden Markov Model acquired in humancomputer interaction is used to represent the high-level structure of gestures reduce the dimensionality of the search space. The estimation of hand gestures is handled by combing the model and particle filter. The sampling is introduced PERM instead of standard re-sampling to compensate the error caused by the observation model. The simulated experiment demonstrates significant improvements in tracking speed and robustness over comparison methods.