Bayesian fusion of hidden Markov models for understanding bimanual movements

Understanding hand and body gestures is a part of a wide spectrum of current research in computer vision and human-computer interaction. A part of this can be the recognition of movements in which the two hands move simultaneously to do something or imply a meaning. We present a Bayesian network for fusing hidden Markov models in order to recognise a bimanual movement. A bimanual movement is tracked and segmented by a tracking algorithm. Hidden Markov models are assigned to the segments in order to learn and recognize the partial movement within each segment. A Bayesian network fuses the HMMs in order to perceive the movement of the two hands as a single entity.