Recognition of Space-Time Hand-Gestures using Hidden Markov Model

The rapidly growing interest in interactive threedimensional(3D) computer environments highly recommend the hand gesture as one of their interaction modalities. Among several factors constituting a hand gesture, hand movement pattern is spatiotemporally variable and informative, but its automatic recognition is not trivial. In this paper, we describe a hidden Markov(HMM)-based method for recognizing the space-time hand movement pattern. HMM models the spatial variance and the time-scale variance in the hand movement. As for the recognition of the continuous, connected hand movement patterns, HMMbased segmentation method is introduced. To deal with the dimensional complexity caused by the 3D problem space, the plane fitting method is employed and the 3D data is reduced into 2D. These 2D data are then encoded as the input to HMMs. In addition to the hand movement, which is regarded as the primary attribute of the hand gesture, we also consider the hand configuration(posture) and the palm orientation. These three major attributes are processed in parallel and rather independently, followed by the inter-attribute communication for finding the proper interpretation.