Boosted learning in dynamic Bayesian networks for multimodal detection

Tanzeem Chaodhury, James M. Rehg, Vladimir Pavlović, Alex Pentland

Research output: Contribution to conferencePaper

6 Citations (Scopus)

Abstract

Bayesian networks are an attractive modeling tool for human sensing, as they combine an intuitive graphical representation with efficient algorithms for inference and learning. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs) which allow the power of statistical inference and learning to be combined with contextual knowledge of the problem. Unfortunately, simple learning methods can cause such appealing models to fail when the data exhibits complex behavior We first demonstrate how boosted parameter learning could be used to improve the performance of Bayesian network classifiers for complex multimodal inference problems. As an example we apply the framework to the problem of audiovisual speaker detection in an interactive environment using "off-the-shelf" visual and audio sensors (face, skin, texture, mouth motion, and silence detectors). We then introduce a boosted structure learning algorithm. Given labeled data, our algorithm modifies both the network structure and parameters so as to improve classification accuracy. We compare its performance to both standard structure learning and boosted parameter learning. We present results for speaker detection and for datasets from the UCI repository.

Original languageEnglish (US)
Pages550-556
Number of pages7
DOIs
StatePublished - Jan 1 2002
Externally publishedYes
Event5th International Conference on Information Fusion, FUSION 2002 - Annapolis, MD, United States
Duration: Jul 8 2002Jul 11 2002

Other

Other5th International Conference on Information Fusion, FUSION 2002
CountryUnited States
CityAnnapolis, MD
Period7/8/027/11/02

Fingerprint

Bayesian networks
Sensors
Learning algorithms
Skin
Classifiers
Fusion reactions
Textures
Detectors

All Science Journal Classification (ASJC) codes

  • Information Systems

Keywords

  • Bayesian networks
  • boosting
  • mutlimodal fusion
  • speaker detection

Cite this

Chaodhury, T., Rehg, J. M., Pavlović, V., & Pentland, A. (2002). Boosted learning in dynamic Bayesian networks for multimodal detection. 550-556. Paper presented at 5th International Conference on Information Fusion, FUSION 2002, Annapolis, MD, United States. https://doi.org/10.1109/ICIF.2002.1021202
Chaodhury, Tanzeem ; Rehg, James M. ; Pavlović, Vladimir ; Pentland, Alex. / Boosted learning in dynamic Bayesian networks for multimodal detection. Paper presented at 5th International Conference on Information Fusion, FUSION 2002, Annapolis, MD, United States.7 p.
@conference{5d74b0e943fe407899893e9bc1b23a0b,
title = "Boosted learning in dynamic Bayesian networks for multimodal detection",
abstract = "Bayesian networks are an attractive modeling tool for human sensing, as they combine an intuitive graphical representation with efficient algorithms for inference and learning. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs) which allow the power of statistical inference and learning to be combined with contextual knowledge of the problem. Unfortunately, simple learning methods can cause such appealing models to fail when the data exhibits complex behavior We first demonstrate how boosted parameter learning could be used to improve the performance of Bayesian network classifiers for complex multimodal inference problems. As an example we apply the framework to the problem of audiovisual speaker detection in an interactive environment using {"}off-the-shelf{"} visual and audio sensors (face, skin, texture, mouth motion, and silence detectors). We then introduce a boosted structure learning algorithm. Given labeled data, our algorithm modifies both the network structure and parameters so as to improve classification accuracy. We compare its performance to both standard structure learning and boosted parameter learning. We present results for speaker detection and for datasets from the UCI repository.",
keywords = "Bayesian networks, boosting, mutlimodal fusion, speaker detection",
author = "Tanzeem Chaodhury and Rehg, {James M.} and Vladimir Pavlović and Alex Pentland",
year = "2002",
month = "1",
day = "1",
doi = "10.1109/ICIF.2002.1021202",
language = "English (US)",
pages = "550--556",
note = "5th International Conference on Information Fusion, FUSION 2002 ; Conference date: 08-07-2002 Through 11-07-2002",

}

Chaodhury, T, Rehg, JM, Pavlović, V & Pentland, A 2002, 'Boosted learning in dynamic Bayesian networks for multimodal detection', Paper presented at 5th International Conference on Information Fusion, FUSION 2002, Annapolis, MD, United States, 7/8/02 - 7/11/02 pp. 550-556. https://doi.org/10.1109/ICIF.2002.1021202

Boosted learning in dynamic Bayesian networks for multimodal detection. / Chaodhury, Tanzeem; Rehg, James M.; Pavlović, Vladimir; Pentland, Alex.

2002. 550-556 Paper presented at 5th International Conference on Information Fusion, FUSION 2002, Annapolis, MD, United States.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Boosted learning in dynamic Bayesian networks for multimodal detection

AU - Chaodhury, Tanzeem

AU - Rehg, James M.

AU - Pavlović, Vladimir

AU - Pentland, Alex

PY - 2002/1/1

Y1 - 2002/1/1

N2 - Bayesian networks are an attractive modeling tool for human sensing, as they combine an intuitive graphical representation with efficient algorithms for inference and learning. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs) which allow the power of statistical inference and learning to be combined with contextual knowledge of the problem. Unfortunately, simple learning methods can cause such appealing models to fail when the data exhibits complex behavior We first demonstrate how boosted parameter learning could be used to improve the performance of Bayesian network classifiers for complex multimodal inference problems. As an example we apply the framework to the problem of audiovisual speaker detection in an interactive environment using "off-the-shelf" visual and audio sensors (face, skin, texture, mouth motion, and silence detectors). We then introduce a boosted structure learning algorithm. Given labeled data, our algorithm modifies both the network structure and parameters so as to improve classification accuracy. We compare its performance to both standard structure learning and boosted parameter learning. We present results for speaker detection and for datasets from the UCI repository.

AB - Bayesian networks are an attractive modeling tool for human sensing, as they combine an intuitive graphical representation with efficient algorithms for inference and learning. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs) which allow the power of statistical inference and learning to be combined with contextual knowledge of the problem. Unfortunately, simple learning methods can cause such appealing models to fail when the data exhibits complex behavior We first demonstrate how boosted parameter learning could be used to improve the performance of Bayesian network classifiers for complex multimodal inference problems. As an example we apply the framework to the problem of audiovisual speaker detection in an interactive environment using "off-the-shelf" visual and audio sensors (face, skin, texture, mouth motion, and silence detectors). We then introduce a boosted structure learning algorithm. Given labeled data, our algorithm modifies both the network structure and parameters so as to improve classification accuracy. We compare its performance to both standard structure learning and boosted parameter learning. We present results for speaker detection and for datasets from the UCI repository.

KW - Bayesian networks

KW - boosting

KW - mutlimodal fusion

KW - speaker detection

UR - http://www.scopus.com/inward/record.url?scp=10244224789&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=10244224789&partnerID=8YFLogxK

U2 - 10.1109/ICIF.2002.1021202

DO - 10.1109/ICIF.2002.1021202

M3 - Paper

AN - SCOPUS:10244224789

SP - 550

EP - 556

ER -

Chaodhury T, Rehg JM, Pavlović V, Pentland A. Boosted learning in dynamic Bayesian networks for multimodal detection. 2002. Paper presented at 5th International Conference on Information Fusion, FUSION 2002, Annapolis, MD, United States. https://doi.org/10.1109/ICIF.2002.1021202