Skip to Main content Skip to Navigation
Theses

Reconnaissance de séquences d'états par le Modèle des Croyances Transférables. Application à l'analyse de vidéos d'athlétisme.

Abstract : This thesis focuses on automatic recognition of dynamical systems. A methodology based on state sequence models has been chosen: states describe the system at particular instants while transitions allow it to evolve along time. In this thesis, two new methods based on Transferable Belief Model, a non-probabilistic model for uncertain reasoning based on belief functions, are proposed: the first one is deterministic and inspired from works in Artificial Intelligence and the second one is stochastic generalizing Hidden Markov Models, initially developed in probability theory, to belief functions. These algorithms, which are generic, have been integrated in a system aiming at recognizing human motions in athletics videos. The system has been set in collaboration with University of Crete within SIMILAR European Network of Excellence. The state sequence recognition method proposed in this thesis have been assessed on a database made of 74 videos and compared to probabilistic Hidden Markov Models.
Complete list of metadatas

Cited literature [266 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00260770
Contributor : Emmanuel Ramasso <>
Submitted on : Wednesday, March 5, 2008 - 10:14:12 AM
Last modification on : Thursday, July 9, 2020 - 5:02:02 PM
Long-term archiving on: : Thursday, May 20, 2010 - 11:56:31 PM

Identifiers

  • HAL Id : tel-00260770, version 1

Citation

Emmanuel Ramasso. Reconnaissance de séquences d'états par le Modèle des Croyances Transférables. Application à l'analyse de vidéos d'athlétisme.. domain_stic.inge. Université Joseph-Fourier - Grenoble I, 2007. Français. ⟨tel-00260770⟩

Share

Metrics

Record views

850

Files downloads

1262