Highlights´ recognition and learning in soccer video by using Hidden Markov Models and the Bayesian theorem

Author

El Ouazzani, R. ; Thami, Rachid Oulad Haj

Author_Institution

Dep.: Syst. d´´Inf. Metiers, Ecole Nat. Super. d´´Inf. et d´´Analyse des Syst. (ENSIAS), Rabat, Morocco

fYear

2009

fDate

2-4 April 2009

Firstpage

304

Lastpage

308

Abstract

Our paper presents a new approach for the recognition of highlights in soccer video. Our contribution consists of the combination of Bayesian theorem inferences and Hidden Markov Models (HMMs). We build HMMs to calculate probabilities that a test video segment belongs to highlight and non highlight classes. Then, we apply the Bayesian theorem on the two previous probabilities. Our system has achieved an accuracy of 95.6% which is a good result of highlights detection in comparison with other methods.

Keywords

Bayes methods; hidden Markov models; image classification; image segmentation; learning (artificial intelligence); probability; sport; video signal processing; video streaming; Bayesian theorem; hidden Markov model; probability; soccer video highlight recognition; soccer video learning; soccer video shot classification; video segment test; video stream analysis; Algorithm design and analysis; Bayesian methods; Color; Gunshot detection systems; Hidden Markov models; Image segmentation; Multimedia systems; Probability; Testing; Viterbi algorithm; Bayesian theorem; Forward-Backward algorithm; Hidden Markov Models; Soccer video analysis; Viterbi algorithm; shots classification;

fLanguage

English

Publisher

ieee

Conference_Titel

Multimedia Computing and Systems, 2009. ICMCS '09. International Conference on

Conference_Location

Ouarzazate

Print_ISBN

978-1-4244-3756-6

Electronic_ISBN

978-1-4244-3757-3

Type

conf

DOI

10.1109/MMCS.2009.5256682

Filename

5256682