Location and extraction of broadcast in news video based on QGMM and BIC

Author

Guo, Ling ; Shi, Ying-Chun ; Zhou, Xian-Zhong ; Zhang, Feng

Author_Institution

Dept. of Autom., Nanjing Univ. of Sci. & Technol., China

fYear

2005

fDate

21-23 Sept. 2005

Firstpage

662

Lastpage

666

Abstract

An algorithm on location and extraction of broadcast in news video is proposed in this paper. Firstly, input audio stream is divided into speech and non-speech segments by VQ (vector quantification) after a set of new features representing audio segments´ time-variant characteristics are extracted, including HZCRR (high zero-crossing rate ratio), LSTER (low short-time energy ratio) and HBFERR (high basic-frequency-energy rate ratio), etc. Then a QGMM (quasi Gaussian mixture model) is presented to describe the speaker´s identity and BIC (Bayesian information criterion) is used to detect speaker change. Finally speaker clustering is carried out with BIC, and location and extraction of broadcast is realized based on rules. Satisfactory results from experiments prove the effectiveness of this algorithm.

Keywords

Bayes methods; Gaussian processes; broadcasting; speaker recognition; speech processing; vector quantisation; video retrieval; Bayesian information criterion; audio stream; high basic-frequency-energy rate ratio; high zero-crossing rate ratio; low short-time energy ratio; news video; quasiGaussian mixture model; speaker clustering; speaker identity; time-variant characteristics; vector quantification; Automation; Bayesian methods; Broadcast technology; Broadcasting; Data mining; Multimedia communication; Multiple signal classification; Signal analysis; Speech; Streaming media;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer and Information Technology, 2005. CIT 2005. The Fifth International Conference on

Print_ISBN

0-7695-2432-X

Type

conf

DOI

10.1109/CIT.2005.137

Filename

1562730