Title :
A general pose estimation algorithm in a multi-Kinect system
Author :
Saiyi Li ; Pathirana, Pubudu N. ; Bonacci, Jason
Abstract :
Microsoft Kinect which has been primarily aimed at the computer gaming industry has been used in bio-kinematic research related implementations. A multi-Kinect system can be useful in exploiting spatial diversity to increase measurement accuracy. One of the main problems in deploying multi-Kinect systems is to estimate the pose, including the position and orientation of each Kinect. In this paper, a singular value decomposition (SVD) least-squares algorithm is extended to a more generic time-series based approach to solve this pose estimation problem utilising 3D positions of one or more joints in skeletons obtained from a multi-Kinect system. Additionally, computer simulations are performed to demonstrate the use and to evaluate the efficiency of the proposed algorithm. The former is further validated with a commercial Vicon system.
Keywords :
computer games; digital simulation; interactive devices; least squares approximations; pose estimation; singular value decomposition; time series; Microsoft Kinect; SVD least-squares algorithm; bio-kinematic research; commercial Vicon system; computer gaming industry; computer simulations; general pose estimation algorithm; multiKinect system; singular value decomposition least-squares algorithm; spatial diversity; time-series based approach; Accuracy; Joints; Noise; Three-dimensional displays; Trajectory; Vectors; SVD; fusion; multi-Kinect; skeleton;
Conference_Titel :
Information and Automation for Sustainability (ICIAfS), 2014 7th International Conference on
DOI :
10.1109/ICIAFS.2014.7069576
