VORTRAG
*******
Oesterreichisches Forschungsinstitut fuer Artificial Intelligence(OFAI)
Freyung 6/6, A-1010 Wien
Tel.: +43-1-53361120, Fax: +43-1-5336112-77, Email: sec@oefai.at
-------------------------------------------------------------------------
Dr. Douglas Eck
www.iro.umontreal.ca/~eckdoug
University of Montreal Department of Computer Science
Montreal Center for Brain, Music, and Sound (BRAMS)
USING AUTOCORRELATION TO FIND TEMPORAL STRUCTURE IN MUSIC
Autocorrelation is a simple, fast-to-compute method that has long
been used as a tool for analyzing metrical structure in music (e.g.
Judy Brown, 1993). Because autocorrelation can be performed online
and works on a variety of inputs including filtered and rectified
digital audio, it is an interesting method for exploring
non-stationary effects such as acceleration and deceleration in
performed music. However autocorrelation has a severe limitation:
while it provides information about the magnitude of signal energy
at different periods, it discards all information about phase. I will
address this issue by presenting a naive way to compute a
phase-preserving autocorrelation for music. I will go on to discuss
a faster method that limits the number of lags where phase is
preserved. Meter induction results will be presented from the Essen
Database and the Finnish Folksong Database. I will conclude by
observing that this model performs online dimensionality reduction
and can be applied outside the areas of beat induction and meter
detection. I will discuss ongoing research into using the results of
the model as input to a gradient-based automatic music composition
learner.
*********
Zeit: Mittwoch, 16. Februar 2005, 18:30 Uhr pktl.
Ort: Oesterreichisches Forschungsinstitut
fuer Artificial Intelligence, OFAI
Freyung 6, Stiege 6, 1010 Wien.
OESTERREICHISCHES FORSCHUNGSINSTITUT
FUER ARTIFICIAL INTELLIGENCE
o.Univ.-Prof. Ing. Dr. Robert Trappl