Semi-Automatic Captioning of Films and Movies

Silvia Pfeiffer

CSIRO/MIS, Digital Media Information Systems

Tuesday 14 March at 11am

Abstract

Within the DMIS (Digital Media Information Systems) group, we look at using video and audio content analysis for industry applications. One of our projects was carried through with the ACC (Australian Captioning Centre) on automating the creation of captions (=subtitles for the deaf). The project was called CACTUS (Computerised Alignment for Captioned Television Using Speech-processing). This title describes what we concentrated on during the first phase of CACTUS: using automatic speech recognition (ASR) for captioning. For several reasons which will be discussed in the talk, current ASR is not able to perform this task. Therefore, in the second phase, we stopped trying to automatically determine the content of the captions, but rather concentrated on calculating the time boundaries of captions by using video and audio content clues. This has proven very successful and we are now on the verge of commercialisation.

To top

last updated December 01, 2003 10:37 AM
Sandrine.Balbo@cmis.csiro.au