Spatial Measures between Human Poses for Classification and Understanding
Statistical analysis of humans, their motion and their behaviour is a very well-studied problem. With the availability of accurate motion capture systems, it has become possible to use such analysis for animation, understanding, compression and tracking of human motion. At the core of the analysis lies a measure for determining the distance between two human poses; practically always, this measure is the Euclidean distance between joint angle vectors. Recent work  has shown that articulated tracking systems can be vastly improved by replacing the Euclidean distance in joint angle space with the geodesic distance in the space of joint positions. However, due to the focus on tracking, no algorithms have, so far, been presented for measuring these distances between human poses.
In this paper, we present an algorithm for computing geodesics in the Riemannian space of joint positions, as well as a fast approximation that allows for large-scale analysis. In the experiments we show that this measure significantly outperforms the traditional measure in classification, clustering and dimensionality reduction tasks.
KeywordsJoint Angle Geodesic Distance Joint Position Latent Variable Model Geodesic Curve
Unable to display preview. Download preview PDF.
- 2.Carmo, M.P.D.: Differential Geometry of Curves and Surfaces. Prentice Hall (1976)Google Scholar
- 3.Erleben, K., Sporring, J., Henriksen, K., Dohlmann, H.: Physics Based Animation. Charles River Media (August 2005)Google Scholar
- 7.Hauberg, S., Sommer, S., Pedersen, K.S.: Natural metrics and least-committed priors for articulated tracking. In: Image and Vision Computing (2011)Google Scholar
- 8.Kjellström, H., Kragić, D., Black, M.J.: Tracking people interacting with objects. In: IEEE CVPR (2010)Google Scholar
- 9.Lu, Z., Carreira-Perpinan, M., Sminchisescu, C.: People Tracking with the Laplacian Eigenmaps Latent Variable Model. In: NIPS, vol. 20, pp. 1705–1712. MIT Press (2008)Google Scholar
- 10.Murray, R.M., Li, Z., Sastry, S.S.: A Mathematical Introduction to Robotic Manipulation. CRC Press (March 1994)Google Scholar
- 11.Poon, E., Fleet, D.J.: Hybrid monte carlo filtering: Edge-based people tracking. In: IEEE Workshop on Motion and Video Computing, p. 151 (2002)Google Scholar
- 13.Priyamvada, K.K., Kahol, K., Tripathi, P., Panchanathan, S.: Automated gesture segmentation from dance sequences. In: Int. Conf. on Automatic Face and Gesture Recognition (2004)Google Scholar
- 14.Ramanan, D., Baker, S.: Local distance functions: a taxonomy, new algorithms, and an evaluation. TPAMI (4) (2011)Google Scholar
- 15.Ripley, B.D.: Pattern recognition and neural networks. Cambridge University Press (1996)Google Scholar
- 16.Sheikh, Y.A., Khan, E.A., Kanade, T.: Mode-seeking by medoidshifts. In: IEEE 11th International Conference on Computer Vision, pp. 1–8 (2007)Google Scholar
- 17.Shon, A.P., Grochow, K., Rao, R.P.: Robotic imitation from human motion capture using Gaussian processes (2005)Google Scholar
- 19.Sminchisescu, C., Jepson, A.: Generative modeling for continuous non-linearly embedded visual inference. In: ICML 2004, pp. 759–766. ACM (2004)Google Scholar
- 22.Urtasun, R., Fleet, D.J., Fua, P.: 3D People Tracking with Gaussian Process Dynamical Models. In: IEEE CVPR, pp. 238–245 (2006)Google Scholar
- 23.Urtasun, R., Fleet, D.J., Hertzmann, A., Fua, P.: Priors for people tracking from small training sets. In: ICCV, vol. 1, pp. 403–410 (2005)Google Scholar
- 25.Yao, A., Gal, J., Fanelli, G., van Gool, L.: Does human action recognition benefit from pose estimation? In: BMVC (2011)Google Scholar