Segmenting and Recognizing Human Action from Long Video Sequences

Object detection and tracking in video is a challenging problem and recognizing human activities from video are one of the most promising applications of computer vision. In this paper, we present a general framework to jointly segmenting and recognizing videos of human action sequences. Hence, we use two approaches: (i) Intensity Range Based Background Subtraction (ii) Shape-Motion Prototype-Based approach. Here first one defines an intensity range for each pixel location in the background to accommodate illumination variation as well as motion in the background and second one is introduced for action recognition. It performs recognition efficiently via tree-based prototype matching and look-up table indexing. It captures correlations between different shape and motion by learning action prototypes in a joint feature space. It also ensures global temporal consistency by dynamic sequence alignment.