Real-Time Hand Pose Estimation Using Depth Camera

In recent years, we have witnessed a steady growth of the research in real-time 3D hand pose estimation with depth cameras, since this technology plays an important role in various human–computer interaction applications. In this chapter, we first review existing techniques and systems for real-time 3D hand pose estimation. Then, we will discuss two point-set-based methods for 3D hand pose estimation from depth images: (1) point-set-based holistic regression method that directly regresses holistic 3D hand pose; (2) point-set-based point-wise regression method that generates dense outputs for robust 3D hand pose estimation. Extensive experiments are conducted to evaluate the effectiveness of these two methods. We will also discuss the limitations and advantages of the proposed methods.