Multi Human Trajectory Estimation Using Stochastic Sampling and Its Application to Meeting Recognition

In this paper we present a stochastic sampling approach to estimate multiple human trajectory in the meeting. The algorithm is formalized as a energy minimization problem based on stochastic sampling of deterministic trajectory, and has some effectiveness to the low frame data with jumps and switchings and it can estimate a near optimal result in 9 times faster then the real-time by using Gibbs sampling. Also experiment is shown using meeting data of real environment.

• Type: Conference
• Year: 2007
• Where: MVA
• Authors: Yosuke Matsusaka, Hideki Asoh, Futoshi Asano
• Keywords: Computer Vision, Multiple Human Trajectory, MVA 2007, Stochastic Sampling, Stochastic Sampling Approach