Tools for Music Information Retrieval and Playing - Intelligent Music Information Systems: Tools and Methodologies - page 132

Figure 19. General architecture The processing phase is characterized also by

other feature extraction techniques and parameters
typical of audio processing. All those techniques
will be analyzed in the next sections.

musIcal feature extractIon

A musical feature extraction system is a part
contained in a MIR system that works requiring
brute force and sophisticated signal-processing
technology, which provides objective information
about music content (Pachet, 2003).
This section will describe the common tasks
performed by this system.

pitch tracking

The conventional fashion of organization of
music collection using singer's names, album's
name, or any other text-based manner is becom-
ing inadequate for effective and efficient usage
of the music collection for average users. People
sometimes prefer to access the music database by
its musical content rather than textual keywords.
Content-based music retrieval has thus become
an active research area in recent years.
Pitch extraction or estimation, more often
called pitch tracking, is a simple form of automatic
music transcription which converts musical sound
into a symbolic representation (Martin, 1996;
Scheirer, 1997).
The basic idea of this approach is quite simple.
Each note of music (including the query) is repre-
sented by its pitch. So a musical piece or segment
is represented as a sequence or string of pitches.
The retrieval decision is based on the similarity
between the query and candidate strings.
Pitch is normally defined as the fundamental
frequency of a sound. To find the pitch for each
note, the input music must first be segmented into
individual notes. Segmentation of continuous
music, especially humming and singing, is very

conversion uses different pitch-tracking
algorithms. If the input is entered from a
keyboard or it is a score fragment, conver-
sion is not necessary and the sequence can
be directly built.

• Musical Query Environment: The Feature

Extractor converts acoustic input first into
an audio feature sequence and then into its
related symbolic representation. Then the
similarity function is computed.

The system we described allows to organize,
manage, and utilize information of a heterogene-
ous set of music source material.
This work is an improvement of the informa-
tion system described in the context of the “Teatro
Alla Scala” project. The improvements regard the
use of the graph model of musical data at different
levels, the XML format for the representation of
musical work and the new user interfaces for the
navigation of musical source material.
Such a system takes advantage of organizing,
managing, and utilizing information of a hetero-
geneous set of music source material through the
XML multilayered structure.