Action recognition based on spatio-temporal interest points

Abstract

Action recognition has already become one of the important research directions in the computer vision domain. Action recognition techniques have made remarkable progress in recent years. This interest is motivated by a wide spectrum of applications in many fields, such as smart surveillance, virtual reality, human computer interaction, and motion analysis. In this paper, we divide action recognition based on spatio-temporal interest points into three fundamental processes: spatio-temporal interest point detection, feature classification and action representation and recognition. We focus on the three areas to take the further study on action recognition based on spatiotemporal interest points. Also we make a comparative studies of the approaches used systematically. We propose the existing problems and possible future development trends of action recognition.