Information Technology Reference
In-Depth Information
(Roccetti et al., 2005) supporting the distribution
of MP3-based songs to 3G UMTS devices. These
applications rely on the existence of a central
server, which receives requests from and delivers
audio files to the mobile clients. Sadly enough,
CBMIR research in this type of infrastructures
has not yet received attention and is currently con-
sidering only metadata MIR. Though, aside from
these single-hop infrastructure wireless networks,
music delivery can also unfold over the emerging
wireless ad-hoc networks. The wireless ad-hoc
networks are peer-to-peer, multi-hop, mobile
wireless networks, where information packets are
transmitted in a store-and-forward fashion from
source to destination, via intermediate nodes. The
salient characteristics of these networks, that is,
dynamic topology, bandwidth-constrained com-
munication links and energy-constraint operation,
introduce significant design challenges.
This section of the chapter focuses on CBMIR
in wireless ad-hoc networks. Consider a number
of mobile hosts that participate in a wireless ad-
hoc network, where each host may store several
audio musical pieces. Assume a user that wants
to search in the wireless network, to find audio
pieces that are similar to a given one. For instance,
the user can provide an audio snippet (e.g., a song
excerpt) and query the network to find the peers
that store similar pieces. As will be described in
the following, the definition of similarity can be
based on several features that have been developed
for Content-Based Music Information Retrieval.
It is important to notice that the querying host is
assumed to have no prior knowledge of both the
qualifying music pieces and the hosts' locations
that contain them. This is the key differentiation
from existing researches that are interested in just
identifying the hosts, in a wireless ad-hoc network,
that contain a known datum. Moreover, the issues
discussed in this section are complementary to the
problem of delivering streaming media, such as
audio and video, as considered by Baochun and
Wang (2003) in wireless ad-hoc networks, since
the latter does not involve any searching for similar
audio pieces, and just focuses on transferring data
from one host to another.
requirements set by the Wireless
medium
This section focuses on methods for searching
audio music by content in wireless ad-hoc net-
works, where the querier receives music excerpts
matching to a posed query. The actual searching
procedure can benefit from the latest approaches
for CBMIR in wired P2P networks (see Section
“CBMIR in Wired Networks”). Nevertheless, the
combination of the characteristics of the wireless
medium and the audio-music data pose challeng-
ing requirements:
1.
CBMIR methods for wired P2P networks
do not consider the continuous alteration
of the network topology, which is inherent
in wireless ad-hoc networks, since Mobile
Hosts (MHs, the terms MH and peer are
similar in this context and thus interchange-
able) are moving and become in and out of
range of the others continuously. One impact
of this mobility is that selective propagation
of the query among MHs, e.g., by using data
indexing like DHT as proposed by Tzane-
takis et al. (2004) or caching past queries
(Kalogeraki et al. (2002) for text documents
and Karydis, Nanopoulos, Papadopoulos,
and Manolopoulos (2005b) for music), is
not feasible. Additionally, the recall of the
searching procedure is affected by the pos-
sibility of unsuccessful routing of the query,
as well as the answers, over the changing
network topology.
2
The need to reduce traffic, which results
from the size of audio-music data (approx.
8 MBytes for a 3 minute query). This is can
be achieved by replacing the original query
with a representation that utilizes appropri-
ate transcoding schemes. Although traffic
concerns CBMIR in wired P2P networks
Search WWH ::
Custom Search