College of Information Science and Technology, Qingdao University of Science & Technology, Qingdao China

---

Abstract—In this paper, we proposed an unsupervised posture modeling method based on Gaussian Mixture Model (GMM). Specifically, each learning posture is described based on its movement features by a set of spatial-temporal interest points (STIPs), salient postures are then clustered from these training samples by an unsupervised algorithm, here we give the comparison of four candidate classification methods and find the optimal one. Furthermore, each clustered posture type is modeled with GMM according to Expectation Maximization (EM) estimation. The experiment results proved that our method can effectively model postures and can be used for posture recognition in video.

Index Terms—NERF C-means; posture modeling; posture recognition; GMM

[PDF]

Cite: Xijun Zhu, Chuanxu Wang, "Unsupervised Posture Modeling and Recognition based on Gaussian Mixture Model and EM Estimation," Journal of Software vol. 6, no. 8, pp. 1445-1451, 2011.