A survey of vision-based methods for action representation, segmentation and recognition

Action recognition has become a very important topic in computer vision, with many fundamental applications, in robotics, video surveillance, human–computer interaction, and multimedia retrieval among others and a large variety of approaches have been described. The purpose of this survey is to give an overview and categorization of the approaches used. We concentrate on approaches that aim on classification of full-body motions, such as kicking, punching, and waving, and we categorize them according to how they represent the spatial and temporal structure of actions; how they segment actions from an input stream of visual data; and how they learn a view-invariant representation of actions.