Finite state representation of hand gesture using key video object plane

The use of hand gestures has become an important part of human computer interaction (HCI) in recent years. Vision-based hand gesture recognition involves visual analysis of hand shape, position and/or movement. Due to co-articulation that occurs during transition from one gesture to the next, problem is encountered in continuous hand gesture recognition. This may be tackled by identifying the key frames in the gesture video sequence. Key frames are the frames that can represent the salient content of a video shot in an abstracted manner. In this paper, we present an object-based scheme for key frame extraction using Hausdorff distance and subsequent local motion analysis by angular circular local motion descriptor (ACLM) for gesture representation. We propose a finite state machine (FSM) in which gestures are represented by the sequence of key frames and the corresponding key frame duration. Experimental results obtained demonstrate the effectiveness of our proposed scheme for key frame extraction and subsequent gesture representation.