Key Developments in Human Pose Estimation for Kinect

The last few years have seen a surge in the development of natural user interfaces. These interfaces do not require devices such as keyboards and mice that have been the dominant modes of interaction over the last few decades. An important milestone in the progress of natural user interfaces was the recent launch of Kinect with its unique ability to reliably estimate the pose of the human user in real time. Human pose estimation has been the subject of much research in Computer Vision, but only recently with the introduction of depth cameras and algorithmic advances has pose estimation made it out of the lab and into the living room. In this chapter we briefly summarize the work on human pose estimation for Kinect that has been undertaken at Microsoft Research Cambridge, and discuss some of the remaining open challenges. Due to the summary nature of this article, we limit our description to the key insights and refer the reader to the original publications for the technical details.