Hidden Markov models based shot detection in sports video

Wei Gang Zhang; Jian Qiu Lin; Yang Liu; Xiao Peng Chen; Xiao Peng Wang

Hidden Markov models based shot detection in sports video

Wei Gang Zhang^*
, Jian Qiu Lin
, Yang Liu
, Xiao Peng Chen
, Xiao Peng Wang

^*Corresponding author for this work

Faculty of Computing

School of Computer Science and Technology (School of Software), Harbin Institute of Technology Weihai
Harbin Institute of Technology Weihai

Research output: Contribution to journal › Article › peer-review

Abstract

Shot detection is the first stage in video structure analysis. Although existing algorithms have acquired good performances in most generic videos such as news video, movie, etc., their applications in sports video encounter new problems, since most sports videos share two inherent characteristics: relative fixed scene and large motions. This paper presents a new method based on machine learning for shot detection in sports video. In this method, the characteristics of sports video are fully considered and Hidden Markov Models (HMMs) are employed to recognize shot transition patterns (cuts and gradual shot transitions) automatically. Therefore, this method does not suffer from any trouble from threshold selection problem. A series of experiments on various types of sports videos are performed and the results are encouraging.

Original language	English
Pages (from-to)	81-85
Number of pages	5
Journal	Harbin Gongye Daxue Xuebao/Journal of Harbin Institute of Technology
Volume	37
Issue number	SUPPL. 4
State	Published - Dec 2005

Keywords

Hidden Markov model
Shot detection
Sports video

Cite this

@article{989cc65c1e834504932869077f61249a,

title = "Hidden Markov models based shot detection in sports video",

abstract = "Shot detection is the first stage in video structure analysis. Although existing algorithms have acquired good performances in most generic videos such as news video, movie, etc., their applications in sports video encounter new problems, since most sports videos share two inherent characteristics: relative fixed scene and large motions. This paper presents a new method based on machine learning for shot detection in sports video. In this method, the characteristics of sports video are fully considered and Hidden Markov Models (HMMs) are employed to recognize shot transition patterns (cuts and gradual shot transitions) automatically. Therefore, this method does not suffer from any trouble from threshold selection problem. A series of experiments on various types of sports videos are performed and the results are encouraging.",

keywords = "Hidden Markov model, Shot detection, Sports video",

author = "Zhang, \{Wei Gang\} and Lin, \{Jian Qiu\} and Yang Liu and Chen, \{Xiao Peng\} and Wang, \{Xiao Peng\}",

year = "2005",

month = dec,

language = "英语",

volume = "37",

pages = "81--85",

journal = "Harbin Gongye Daxue Xuebao/Journal of Harbin Institute of Technology",

issn = "0367-6234",

publisher = "Harbin Institute of Technology",

number = "SUPPL. 4",

}

TY - JOUR

T1 - Hidden Markov models based shot detection in sports video

AU - Zhang, Wei Gang

AU - Lin, Jian Qiu

AU - Liu, Yang

AU - Chen, Xiao Peng

AU - Wang, Xiao Peng

PY - 2005/12

Y1 - 2005/12

N2 - Shot detection is the first stage in video structure analysis. Although existing algorithms have acquired good performances in most generic videos such as news video, movie, etc., their applications in sports video encounter new problems, since most sports videos share two inherent characteristics: relative fixed scene and large motions. This paper presents a new method based on machine learning for shot detection in sports video. In this method, the characteristics of sports video are fully considered and Hidden Markov Models (HMMs) are employed to recognize shot transition patterns (cuts and gradual shot transitions) automatically. Therefore, this method does not suffer from any trouble from threshold selection problem. A series of experiments on various types of sports videos are performed and the results are encouraging.

AB - Shot detection is the first stage in video structure analysis. Although existing algorithms have acquired good performances in most generic videos such as news video, movie, etc., their applications in sports video encounter new problems, since most sports videos share two inherent characteristics: relative fixed scene and large motions. This paper presents a new method based on machine learning for shot detection in sports video. In this method, the characteristics of sports video are fully considered and Hidden Markov Models (HMMs) are employed to recognize shot transition patterns (cuts and gradual shot transitions) automatically. Therefore, this method does not suffer from any trouble from threshold selection problem. A series of experiments on various types of sports videos are performed and the results are encouraging.

KW - Hidden Markov model

KW - Shot detection

KW - Sports video

UR - https://www.scopus.com/pages/publications/33645636064

M3 - 文章

AN - SCOPUS:33645636064

SN - 0367-6234

VL - 37

SP - 81

EP - 85

JO - Harbin Gongye Daxue Xuebao/Journal of Harbin Institute of Technology

JF - Harbin Gongye Daxue Xuebao/Journal of Harbin Institute of Technology

IS - SUPPL. 4

ER -

Hidden Markov models based shot detection in sports video

Abstract

Keywords

Other files and links

Fingerprint

Cite this