Bimanual movements are a subset ot human movements in which the two hands move together in order to do a task or imply a meaning A bimanual movement appearing in a sequence of images must be understood in order to enable computers to interact with humans in a natural way This problem includes two main phases, hand tracking and movement recognition. We approach the problem of hand tracking from a neuroscience point ot view First the hands are extracted and labelled by colour detection and blob analysis algorithms In the presence of the two hands one hand may occlude the other occasionally Therefore, hand occlusions must be detected in an image sequence A dynamic model is proposed to model the movement of each hand separately Using this model in a Kalman filtering proccss the exact starting and end points of hand occlusions are detected We exploit neuroscience phenomena to understand the beha\ tour of the hands during occlusion periods Based on this, we propose a general hand tracking algorithm to track and reacquire the hands over a movement including hand occlusion The advantages of the algorithm and its generality are demonstrated in the experiments. In order to recognise the movements first we recognise the movement of a hand Using statistical pattern recognition methods (such as Principal Component Analysis and Nearest Neighbour) the static shape of each hand appearing in an image is recognised A Graph- Matching algorithm and Discrete Midden Markov Models (DHMM) as two spatio-temporal pattern recognition techniques are imestigated tor recognising a dynamic hand gesture For recognising bimanual movements we consider two general forms ot these movements, single and concatenated periodic We introduce three Bayesian networks for recognising die movements The networks are designed to recognise and combinc the gestures of the hands in order to understand the whole movement Experiments on different types ot movement demonstrate the advantages and disadvantages of each network.