Morphological gesture recognition algorithm

Abstract

Previously, new methods have been proposed in field of gesture recognition, however, a most important problem remains. That is the increment of processing time by reason of complexity of the algorithms. Mathematical morphology based on geometrical set theory is the best because real time image processing is possible. This paper presents a posture recognition algorithm using the orientation of primitive elements that are extracted by the morphological shape decomposition algorithm as a new parameter-based method. In the proposed algorithm, 4 postures are devised for navigating the video content editing system. We also propose a fast algorithm for gesture recognition of a human using chain code edge vectors measured from an image sequence. The key idea of the proposed algorithm is to track a trajectory of the center points in primitive elements extracted by morphological shape decomposition. The trajectory of the morphological center points includes information on shape orientation.