Title :
Design of audio-visual TV broadcast news transcription system prototype
Author :
Chaloupka, Josef
Author_Institution :
Lab. of Comput. Speech Process., Tech. Univ. of Liberec, Liberec, Czech Republic
Abstract :
This contribution focuses on the design of our automatic audio-visual TV broadcast news transcription system, where we would like to extend our Czech transcription system to use information from the visual signal of TV news video recordings. The subsystems for visual signal segmentation, for visual speaker identification and for visual voice activity detection are described here. These subsystems should help to develop our automatic audiovisual transcription system.
Keywords :
audio-visual systems; image segmentation; speaker recognition; television broadcasting; video signal processing; Czech transcription system; TV news video recordings; automatic audio-visual TV broadcast news transcription system; visual signal segmentation; visual speaker identification; visual voice activity detection; Discrete cosine transforms; Hidden Markov models; Humans; Image color analysis; Image segmentation; Speech recognition; Visualization; audio-visual TV broadcast news transcription; visual signal segmentation; visual speaker idntification; visual voice activity detector;
Conference_Titel :
ELMAR, 2011 Proceedings
Conference_Location :
Zadar
Print_ISBN :
978-1-61284-949-2
Electronic_ISBN :
1334-2630
