Information Technology Reference
In-Depth Information
This would also be able to return further results
to the user, which would not be returned via
conventional methods— any samples that match
the users search criteria, will be returned. This
means that when ignoring criteria such as pitch and
tempo and investigating the timbre of the sound
file, new sounds that could be adapted to work
for the user, could be suggested. For example, a
sped up sample of a car collision may easily work
well in place of a snare sample!
Another example consists of a scenario that
requires speech recognition algorithms to detect a
certain caller from a large amount of recorded tele-
phone audio data. Given a sophisticated enough
algorithm (created in Triana as a workflow), the
DART framework could be used to scan terabytes
of recorded audio data to help trace calls from a
specific caller.
The flexibility of the DART system stems
from the ability to easily refine and change the
Triana workflow that dictates what the worker
nodes will be processing in their screensaver/
idle/nice time.
can spare some disk space for the greater good of
the system. Such approaches have been shown to
work elsewhere where incentives are given in order
to participate, for example, ability to download
files, point scoring for recognition, and so forth.
In DART, the motivation will be provided by the
usefulness of the applications that will make use
of this framework by allowing users to search a
global audio information space to, for example,
find other similar music they may like. Although,
the database is decentralized and therefore consists
of transient participants (who may disconnect for
example), the caching scheme employed ensures
replication across the various nodes in order to
retain the information. This mechanism is use-
ful because the distributed database is persisted
based on the users' demand and therefore will
help prune and optimise the process.
The change in use of the Internet over the past
five years clearly points to approaches such as this
for building next-generation MIR applications.
DART maybe a first step in this direction but we
hope it will lead to further frameworks or take-up
of this type of technology.
future trends
conclusIon
DART is a new research project that is attempt-
ing to provide a platform for addressing future-
looking MIR applications. DART attempts to
gain access to the participants of the Internet as
a whole rather than providing an access point or
a main database for storing global information.
We believe such a decentralized approach is
key to the future possibilities of MIR applica-
tions. DART assumes zero configurability and
administrations and aims to scale and reuse the
resources already available on the Internet. It does
this by involving the participation of end users in
all aspects of the system. The DART system has
no single point of control but rather it organically
lives across the network. The DART database, for
example, evolves across a decentralized layer of
super peers, which are simply Internet users that
In this chapter, we provided an overview of a
new framework called DART, which attempts to
extend the state-of the-art in the application of P2P
networks for MIR. At its essence, DART enables
users to provide input to a massively distributed
database through the analysis of their audio files
for application in MIR. DART provides decen-
tralized P2P mechanisms to allow the scalability
of such a database for dynamic construction and
manageability. DART is a research project that
aims to provide this framework as open source to
the community for simplifying the development of
Internet scale MIR applications. The framework
is based on Java and our interests lie in the ability
to create a global music information space that
allows us to investigate novel ways of searching
Search WWH ::
Custom Search