Recognition of hand action in virtual maintenance training system

First the essential hand action elements through out the maintenance process are obtained by the modelling study of the equipment maintenance action, and according to them the necessary gestures incarnating the action intention are obtained; Then gestures recognition are realized by fuzzy neural network after their definition. Emphasis is introduction to the structure and learning algorithm of network, and based on the exist algorithm of relevance grade, a new algorithm named average derivation to relevance grade is given. Finally the convergence and high efficiency of the network is verified by a gesture recognition instance in the virtual maintenance training system established by us.