Skip to Main Content
This paper proposes a finite state and fuzzy logic based approach to hand gesture learning and recognition. The location of 2D image positions of the hands of the user is obtained by Edge detection and vector extraction. These are used to identify the hand posture as well as the center of the hand of the user. We first learn the spatial information without data segmentation and alignment. Then the data is grouped into clusters that are associated with information for temporal alignment. The points thus obtained are clustered using Fuzzy c-mean clustering algorithm. These clusters of hand posture further determine the states of the Finite State Machine(s) through which the succeeding gesture has to be matched. To build a Gesture Recognizer (GR) the temporal information is integrated. Each hand gesture is defined to be an ordered sequence of states in spatial-temporal space i.e. FSM corresponding to it. The number of states/clusters in an FSM represents a trade-off between the accuracy of gesture recognizer and the amount of spatial-temporal data it stores.