Action recognition can be defined as a problem to determine what kind of action is happening in a video. It is a process of matching the observation with the previously labelled samples and assigning label to that observation. In this paper, a framework of the action recognition system based on saliency detection and random subspace ensemble classifier, is introduced in order to increase the performance of the action recognition. The proposed action recognition framework can be partitioned into three main processing phases. The first processing phase is detecting salient foreground objects by considering pattern and color distinctness of a set of pixels in each video frame. In the second processing phase, changing gradient orientation features are used as a useful feature representation. The third processing phase is recognizing actions using random subspace ensemble classifier with discriminant learner. Experimental results are evaluated on the UIUC action dataset. The proposed action recognition framework achieved satisfying action recognition accuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.