Paper Abstract and Keywords |
Presentation |
2010-03-16 12:20
Improvement of Accuracy in Bayesian Hidden Markov Model Approach for Sports Event Detection Tomohiro Yazaki (Waseda Univ.), Toshie Misu (NHK), Yohei Nakada, Shigeru Motoi, Go Kobayashi, Takashi Matsumoto (Waseda Univ.), Nobuyuki Yagi (NHK) PRMU2009-301 HIP2009-186 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
The problem of detecting the occurrence of target events in a given data sequence can be found in many fields , such as signal processing, pattern recognition, and image processing. We propose an event prediction method by using the Bayesian hidden Markov Model (HMM) and apply it to a video dataset of an actual soccer game. In this method, time-series data on the players' position are estimated from the video dataset and modeled with the HMM. The accuracy of this event prediction method can be improved by modifying the duration-modeling capability of the HMM, since a geometric distribution is used as its duration probability. In this paper, we use a generalized HMM in the event prediction algorithm for improving the duration-modeling capability. In addition, we test the efficiency of the proposed method with a video dataset of J-league soccer matches. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Event detection / Sports video analysis / Generalized hidden Markov Model / Bayesian learning / Markov chain Monte Carlo / / / |
Reference Info. |
IEICE Tech. Rep., vol. 109, no. 470, PRMU2009-301, pp. 401-406, March 2010. |
Paper # |
PRMU2009-301 |
Date of Issue |
2010-03-08 (PRMU, HIP) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
PRMU2009-301 HIP2009-186 |
|