We propose a method for human head pose estimation based on images acquired by a depth camera. During an initialization phase, a reference depth image of a human subject is obtained. At run time, the method searches the 6-dimensional pose space to find a pose from which the head appears identical to the reference view. This search is formulated as an optimization problem whose objective function quantifies the discrepancy of the depth measurements between the hypothesized views to the reference view. The method is demonstrated in several data sets including ones with known ground truth and comparatively evaluated with respect to state of the art methods. The obtained experimental results show that the proposed method outperforms existing methods in accuracy and tolerance to occlusions. Additionally, compared to the state of the art, it handles head pose estimation in a wider range of head poses.
You may also be interested in an earlier work that performed head pose estimation based on multiple RGB cameras.
Head pose estimation based on depth data
- Paschalis Panteleris, Xenophon Zabulis, Antonis A. Argyros
- This work has been supported by the EU projects DALi, HOBBIT and the FORTH-ICS internal RTD Programme on Ambient Intelligence and Smart Environments (AmI).
- P. Padeleris, X. Zabulis and A.A. Argyros, "Head pose estimation on depth data based on particle swarm optimization", In IEEE Computer Vision and Pattern Recognition Workshops (CVPRW 2012), IEEE, pp. 42-49, Providence, Rhode Island, USA, June 2012.
The electronic versions of the above publications can be downloaded from my publications page.