Recognition of variations using automatic Schenkerian reduction.

Marsden, Alan (2010) Recognition of variations using automatic Schenkerian reduction. In: International Society for Music Information Retrieval (ISMIR 2010), 2010-08-01.

[img]
Preview
PDF (ISMIR2010AMRev.pdf)
ISMIR2010AMRev.pdf

Download (265kB)

Abstract

Experiments on techniques to automatically recognise whether or not an extract of music is a variation of a given theme are reported, using a test corpus derived from ten of Mozart's sets of variations for piano. Methods which examine the notes of the 'surface' are compared with methods which make use of an automatically derived quasi-Schenkerian reduction of the theme and the extract in question. The maximum average F-measure achieved was 0.87. Unexpectedly, this was for a method of matching based on the surface alone, and in general the results for matches based on the surface were marginally better than those based on reduction, though the small number of possible test queries means that this result cannot be regarded as conclusive. Other inferences on which factors seem to be important in recognising variations are discussed. Possibilities for improved recognition of matching using reduction are outlined.

Item Type:
Contribution to Conference (Poster)
Journal or Publication Title:
International Society for Music Information Retrieval (ISMIR 2010)
Uncontrolled Keywords:
/dk/atira/pure/researchoutput/libraryofcongress/m1
Subjects:
ID Code:
34348
Deposited By:
Deposited On:
05 Oct 2010 09:52
Refereed?:
Yes
Published?:
Published
Last Modified:
13 Aug 2020 23:59