This paper describes a way of automatically detecting the timing for the superimposition of closed captions on TV programs that use electronic manuscripts. Speech is discriminated from music to decrease the number of false alarms and the amount of speech data within which a search must be carried out. Timing is then detected serially from the first sentence to the last sentence of a program by using a phonetically HMM-based word spotter. After all timing has been determined, it is checked for errors and revised. Detection rates of 96.9% and 98.5% were obtained for allowable timing errors of one and three seconds on the aired documentary programs.
Cite as: Maruyama, I., Abe, Y., Ehara, T., Shirai, K. (2000) An automatic timing detection method for superimposing closed captions of TV programs. Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000), vol. 3, 1109-1112, doi: 10.21437/ICSLP.2000-730
@inproceedings{maruyama00_icslp,
author={Ichiro Maruyama and Yoshiharu Abe and Terumasa Ehara and Katsuhiko Shirai},
title={{An automatic timing detection method for superimposing closed captions of TV programs}},
year=2000,
booktitle={Proc. 6th International Conference on Spoken Language Processing (ICSLP 2000)},
pages={vol. 3, 1109-1112},
doi={10.21437/ICSLP.2000-730}
}