A LITERATURE STUDY ON HUMAN MOTION ANALYSIS USING DEPTH IMAGERY

Analysis of human behavior through visual information is a highly active research topic in the computer vision community. This analysis is achieved in the literature via images from the conventional cameras; however recently depth sensors are used to obtain new type of images known as depth images. This human motion analysis can be widely applied to various domains, such as security surveillance in public spaces, shopping centers and airports. Home care for elderly people and children can use live video streaming from an integrated home monitoring system to prompt timely assistance. Moreover, automatic human motion analysis can be used in Human–Computer/Robot Interaction (HCI/HRI), video retrieval, virtual reality, computer gaming and many other fields. Human motion analysis using a depth sensor is still a new research area. Most work is focused on motion capture of articulated body skeletons. However, the research community is showing interest in higher level action related research. This report explains the advantages of depth imagery and then describes the new categories of depth sensors such as Microsoft Kinect that are available to obtain depth images. High-resolution real-time depth images are cheaply available because of tools like Microsoft Kinect. The main published research on the use of depth imagery for analyzing human activity is reviewed. A growing research area is the recognition of human actions and hence the existing work focuses mainly on body part detection and pose estimation. The publicly available datasets that include depth imagery are listed in this report, and also the software libraries that are available for the depth sensors are explained. With the development of depth sensors, an increasing number of algorithms have employed depth data in vision-based human action recognition. The increasing availability of depth sensors is broadening the scope for future research. This reports provides an overview of this emerging field followed by various vision based algorithms used for human motion analysis