Simultaneous Synchronization of Text and Speech for Broadcast News Subtitling

Authors:
Jie Gao;Qingwei Zhao;Ta Li;Yonghong Yan
Affiliations:
ThinkIT Speech Lab, Institute of Acoustics, Chinese Academy of Sciences, Beijing, P.R. China 100190;ThinkIT Speech Lab, Institute of Acoustics, Chinese Academy of Sciences, Beijing, P.R. China 100190;ThinkIT Speech Lab, Institute of Acoustics, Chinese Academy of Sciences, Beijing, P.R. China 100190;ThinkIT Speech Lab, Institute of Acoustics, Chinese Academy of Sciences, Beijing, P.R. China 100190
Venue:
ISNN 2009 Proceedings of the 6th International Symposium on Neural Networks: Advances in Neural Networks - Part III
Year:
2009

Citing 5
Cited 1

Fundamentals of speech recognition

Fundamentals of speech recognition
Automatic time alignment of phonemes using acoustic-phonetic information

Automatic time alignment of phonemes using acoustic-phonetic information
A One-Pass Real-Time Decoder Using Memory-Efficient State Network

IEICE - Transactions on Information and Systems
LyricAlly: Automatic Synchronization of Textual Lyrics to Acoustic Music Signals

IEEE Transactions on Audio, Speech, and Language Processing
Speechbot: an experimental speech-based search engine formultimedia content on the web

IEEE Transactions on Multimedia

Towards precise and robust automatic synchronization of live speech and its transcripts

Speech Communication

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we present our initial effort in automatic generation of subtitle for live broadcast news programs, utilizing the fact that nearly perfect transcriptions are available. Instead of using the former error-prone automatic-speech-recognition (ASR)-based method, we propose to formulate the subtitling problem as synchronization of text and speech, which is further simplified into an anchor points estimation problem. The Viterbi algorithm for hidden Markov model (HMM) is augmented with new criterions for the online anchor points estimation. Experiments indicate that our proposed methods show satisfying performance for the simultaneous subtitling application. We also present a brief introduction into our whole subtitling system under further development.