Model-free anchor speaker turn detection for automatic chapter generation in broadcast news

Author

Delphine, Charlet

Author_Institution

France Telecom Orange Labs., Lannion, France

fYear

2010

fDate

14-19 March 2010

Firstpage

4966

Lastpage

4969

Abstract

This paper shows how speaker clustering can be used to automatically generate chapters in broadcast news programs, through the detection of anchor speaker turns, without any pre-calculated model of the anchor speaker. Speaker clustering is applied on each broadcast news program, and thanks to the specific structure of broadcast news, a cluster is chosen and assigned to the anchor speaker. That enables to automatically extract reports (those portions of broadcast news which are not from anchor speaker). Experiments performed on a specifically collected database of 38 French broadcast news programs show a F-measure of 89% for the anchor speaker speech detection and prove the interest of the approach.

Keywords

broadcasting; pattern clustering; speaker recognition; speech synthesis; French broadcast; anchor speaker turn; automatic chapter generation; model free speaker detection; news program broadcast; speaker clustering; Clustering algorithms; Indexing; Multimedia communication; Multimedia databases; Parameter extraction; Scheduling; Smoothing methods; Speech processing; TV broadcasting; Telecommunications; automatic chapter generation; multimedia indexing; speaker clustering;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics Speech and Signal Processing (ICASSP), 2010 IEEE International Conference on

Conference_Location

Dallas, TX

ISSN

1520-6149

Print_ISBN

978-1-4244-4295-9

Electronic_ISBN

1520-6149

Type

conf

DOI

10.1109/ICASSP.2010.5495090

Filename

5495090