VORTRAG
*******
Oesterreichisches Forschungsinstitut fuer Artificial Intelligence(OeFAI)
Freyung 6/6, A-1010 Wien
Tel.: +43-1-53361120, Fax: +43-1-5336112-77, Email: sec@oefai.at
-------------------------------------------------------------------------
Fabien Gouyon
Universitat Pompeu Fabra, IUA, Music Technology Group, Barcelona und
Oesterreichisches Forschungsinstitut fuer Artificial Intelligence,
Music Processing Group
FROM LOW-LEVEL SOUND PROCESSING TO CONTENT-BASED MUSIC PROCESSING:
ACTIVITIES OF THE MUSIC TECHNOLOGY GROUP (MTG) IN BARCELONA
Since its creation, the MTG focused on building signal models for
the analysis and synthesis of musical sounds, extending the doctoral
work of its director, Xavier Serra, on Spectral Modeling Synthesis
in Stanford University. From this initial research direction, the
focus has been widened to embrace related topics. Among others are
singing voice synthesis and transformation, time-stretching of audio,
interactive composition systems, fingerprinting techniques for song
and music recording identification and audio content analysis,
description and transformation.
In this lecture, we will provide a short overview of current research
efforts in the MTG and we will put a special focus on the work being
done on music content-based processing. One aspect of processing
musical content is the automatic description of music in terms of
highly abstract representations. Applications to content-based
description are manifold: browsing musical databases, performance
analysis, etc. We will concentrate on musical expressiveness
transformations; that is, the editing and transformation of musical
audio signals triggered by musically-meaningful representational
elements, in contrast to low-level signal descriptors. We will
demonstrate a computer software for modifying rhythmic performances
of polyphonic musical audio signals. The rhythmic dimension it
triggers is the swing. It first describes offline the rhythmic
content of an audio signal: determination of tempi and beat indexes
at the quarter-note and eighth-note levels, as well as estimation of
the swing ratio. Then, the signal is transformed in real-time using
a time-stretch algorithm. We will present basic techniques provided
by commercial products for swing modification and compare these to
our system.
Zeit: Dienstag, 16. Dezember 2003, 18:00 Uhr pktl.
Ort: Institut fuer Medizinische Kybernetik und Artificial
Intelligence der Universitaet Wien (IMKAI)
Wien 1, Freyung 6, Stg.2 (Schottenhof), Tel. 4277-63101
www.ai.univie.ac.at/imkai
OESTERREICHISCHES FORSCHUNGSINSTITUT
FUER ARTIFICIAL INTELLIGENCE
o.Univ.-Prof. Ing. Dr. Robert Trappl