Information Technology Reference
In-Depth Information
syntactic/semantic structures of musical data and
the metric relationships between musical objects.
In the context of structural similarity, we show that
it is possible to extract some structural feature from
the score that are invariant under a lot of standard
and nonstandard musical transformations.
This section ends with an implementation
of the model within the structural layer of MX,
an IEEE standard XML language specifically
designed to support interchange between musi-
cal notation, performance, analysis, and retrieval
applications.
The second part deals with state of the art
musical feature extraction. We make use of a
bottom-up strategy. First we focus on three dif-
ferent blind tasks: beat and tempo tracking, pitch
tracking, and automatic recognition of musical
instruments; the attribute blind refers to the fact
that these tasks deal with audio signals without
paying attention to the symbolic information
layer (score). Second we present the most useful
algorithms which have proven to be most effec-
tive in solving these problems in general purpose
situations, providing also an overview into specific
task applications. These algorithms work both on
compressed and uncompressed data; particular
attention will be given to MPEG audio formats
like AAC and MP3. We then introduce second
level tasks, such as automatic genre extraction
and score extraction, that make use of proprietary
algorithms too, which will be described in the
chapter. We analyze the relationships between
MIR and feature extraction presenting examples
of possible applications. Finally we focus on
automatic music synchronization, a non-blind
task on score and the corresponding audio perfor-
mance, pointing out both solving algorithms and
their applications in MIR, music playing, music
education, and musicology. We introduce a new
audio player that supports the MX logic layer and
allows users to play both the symbolic score and
the related audio file coherently, offering a new
experience in music listening.
musIcal metrIcs
Everyday experience tells us that the most out-
standing music search engines deal with music-
extraneous metadata like author's names, dates
of publication, and so forth.
Music information retrieval (MIR) research
has been producing numerous methods and tools,
but the design of complete, usable working systems
is still in its infancy (Hewlett & Selfridge-Field,
2005). The urgent need for such systems is strongly
felt, in particular by cultural heritage institutions
that possess large music holdings.
Up to date, different musical similarity metrics
or similarity functions have been applied to music
contents. As an example, this approach has been
implemented in a software module which has been
providing an effective environment for querying
music scores in the Musical Archive Information
System (MAIS) of the Italian theatre “Teatro Alla
Scala” (Figure 1).
This module supports both standard and con-
tent-based queries in order to allow both metadata
and semantic queries.
Musical Archive Information System is an
integration of many research efforts developed
at Laboratorio di Informatica Musicale (LIM).
It is a system that allows to organize, manage,
and utilize information of an heterogeneous set
of music source material.
The main functionalities are database man-
agement and content data navigation. Database
management allows the consultation of structured
and unstructured multimedia information. Struc-
tured information is retrieved through traditional
queries. Unstructured information is retrieved by
nontraditional queries such as humming or play-
ing of a melody (content queries). Content data
navigation allows the nonsequential synchronized
rendering of retrieved audio and score sources.
Three main modules compose the frame-
work of this information system: a multimedia
database, an archive of musical files, and a set
of interfaces.
Search WWH ::
Custom Search