Infinite hidden Markov models for unusual-event detection in video.

Published

Journal Article

We address the problem of unusual-event detection in a video sequence. Invariant subspace analysis (ISA) is used to extract features from the video, and the time-evolving properties of these features are modeled via an infinite hidden Markov model (iHMM), which is trained using "normal"/"typical" video. The iHMM retains a full posterior density function on all model parameters, including the number of underlying HMM states. Anomalies (unusual events) are detected subsequently if a low likelihood is observed when associated sequential features are submitted to the trained iHMM. A hierarchical Dirichlet process framework is employed in the formulation of the iHMM. The evaluation of posterior distributions for the iHMM is achieved in two ways: via Markov chain Monte Carlo and using a variational Bayes formulation. Comparisons are made to modeling based on conventional maximum-likelihood-based HMMs, as well as to Dirichlet-process-based Gaussian-mixture models.

Full Text

Duke Authors

Cited Authors

  • Pruteanu-Malinici, I; Carin, L

Published Date

  • May 2008

Published In

Volume / Issue

  • 17 / 5

Start / End Page

  • 811 - 822

PubMed ID

  • 18390385

Pubmed Central ID

  • 18390385

Electronic International Standard Serial Number (EISSN)

  • 1941-0042

International Standard Serial Number (ISSN)

  • 1057-7149

Digital Object Identifier (DOI)

  • 10.1109/tip.2008.919359

Language

  • eng