Detection of hands-raising gestures using shape and edge features

This paper introduces a method of hand-raising gestures detection in indoor environments, using shape and edge features. Past approaches have detected the gestures through recognizing the action for isolated or seated persons. Here, to deal with movements, non-rigidity and partially occlusions of human bodies, the gestures are detected by searching for raised hands and arms rather than recognizing the action. First, background subtraction is employed to obtain body silhouette. And then, according to the particular shape edge features of raised hands and arms, CR (candidate region) search, SR-transform based shape and GLAC edge features extraction and classification, are applied to find raised hands. The classification is implemented by a hierarchical detector which consists of four SVM classifiers. Experiments show that this method can detect hand-raising gestures well, even for moving persons in crowd.