"Content-Oriented Audiovisual Library Access" is a research
project granted by Swiss National Funds (SNF) and realized in MEDIA
research group of DI-
LITH in Swiss Federal Institute of Technology.
The aim of the project is to design and implement a digital audio-visual
library system for TV broadcasters and video archive owners with facilities
to provide effective content-oriented access to internal and external
end-users.
With the recent advances in the technologies of audio-visual data digitalisation,
compression and streaming, there is an increasing amount of audio-visual
data being produced, transferred and stored in digital format. With this
increase in the use of digital audio-visual material, there exists nowadays
an important need for computer systems, which allow an effective exploitation
(i.e. processing, management and retrieval) of these materials. The project
"COALA" was mainly initiated to satisfy this need: to provide
effective content-based access to audio-visual data.
Recently different research and commercial applications have been developing
query interfaces and browsing mechanisms to explore the contents of an audio-visual
database (refer to Research). However,
there are few works that have considered an effective integration of these
two mechanisms in a unique system.
Some important design features of our system are:
- Integration of both mechanisms of advanced querying and navigation of audio-visual data.
- Relevance feedback from browsing to querying.
- Personalisation of advanced querying and navigation interfaces based on different user profiles.
In order to validate the design features of the system (i.e. complex query and navigation model)
we needed to create a rich video corpus which contains different possible descriptions related
to the audio-visual data.
Moreover, as to maintain the interoperability of the system with various applications,
the document model on which the system operates should be confirm to a well-known standard.
Such a standard for audio-visual document description is currently being developed by MPEG
community, called as MPEG-7. Being a member of
this community, we aim at developing an MPEG-7 compatible system.
We are collaborating with the TSR: Television
Suisse Romande. We have studied the current archiving system they use, different users of this
system, their needs and their problems.
We are preparing a corpus containing 10 hours of TSR television news with all corresponding
descriptors and scripts. This corpus will be in near future extended to 20 hours containing
also the teletext subtitles of the news programs. The Data Definition Language used currently
to represent the corpus is XML schema. We aim at producing an MPEG-7 version of this corpus
too. More explanations on the corpus can be found in Research.
In order to facilitate the creation of the corpus two applications have been developed: LogCreator and
StlML. More details on these applications and also their demos can be found on Demos.