A sound data base ... et al

Subject: A sound data base ... et al
From: Kevin Austin (kevin.austin@videotron.ca)
Date: Thu May 20 2004 - 13:47:32 EDT

A major problem for ea in the creation of 'databases' (and
transitional tables etc) is having some form of descriptive
'language' (notation ?) that is even crudely functional.

I'll provide 2 examples.

(1) The Haydn Symphony Minuet database. This would take the minuets
of (say) the first 60 Haydn Symphonies and (through MIDI <ug!>)
encode the notes used. This information could be parsed and
manipulated by a number of existing tools. Because they will largely
be 8, 12 and 16 bar phrases, mid-level structures can be pre-encoded.
Pitch content and background information (major / minor; cadential
identification; basic circle of fifths motion etc). It will be
possible to create crude frequency / probability tables. This is not
to say that these tables represent the music, but compared to case
(2) below, these are light years away.

(2) Two recently performed ea pieces in the <aibohPhobia> (fear of
palindromes) concerts. One was live (direct digital synthesis), one
was a Reaktor
(http://www.native-instruments.com/index.php?reaktor4_us) piece. At
the pitch level, both were not complex, exploring instead aspects of
(for example) extremes of pulse-width modulation and non-linear
distortion and formant structures. They were both 'continuous', that
is lacking in internal points of (structural) articulation.

And now the difficulty begins, as there is not a widely agreed upon
(non-gestural) language to start to describe what happened during the
pieces. To attempt to create a database it seems would require a
method of streaming and segmenting the incoming sounds.

The Haydn assumes a restricted (coherent) basic set of sonic elements
(notes); the sources of the cibohPhobic pieces would be described by
the logical limitations of the instrument(s) (software) in use.

A parallel could be that Haydn and Stockhausen share the same
database for the Sonatas and Klavierstuck XI

The descriptive system for 'sound' may have to be significantly wider
than that for note-based music, for while the Haydn symphony will
maintain its identity when played by orchestra, piano, organ or trio,
in the 'sonic database', this identity could be lost. But all hope is
not lost as the area of voice recognition has pointed to a number of
ways of extracting important information based, even under variant
conditions (formant and spectral peaks, pacing, articulation of
consonants etc etc).



This archive was generated by hypermail 2b27 : Sat Dec 22 2007 - 01:46:01 EST