A Hierarchical Bayesian Hidden Markov Model for Multi-Dimensional Discrete Data
Shigeru Motoi, Yohei Nakada, Toshie Misu, Tomohiro Yazaki, Takashi Matsumoto, Nobuyuki Yagi · InTech eBooks · 2008
In this chapter, we have described an extended Bayesian HMM for multidimensional discrete data sequences including redundant components. For the extended model, we also described an implementation of Bayesian learning based on a Markov chain Monte Carlo scheme. We evaluated the performance of the extended model with this implementation using two example datasets. We also demonstrated its application to an event detection problem with 40-dimensional data sequences extracted from videos of actual soccer games.