Han-Wen Nienhuys <[EMAIL PROTECTED]> writes:
> [EMAIL PROTECTED] writes:
> > > The trick is that it is very easy to do if you have the MIDI file.
> > > You grab the list of pitches from a MIDI file, and transpose it twelve
> > > times.
> > Just transpose once so the first note is `c', or `a', or whatever you like,
> > but consistently. The search engine merely does the same with the request.
> > No need to store twelve times. And of course, you do not even store the
> > first note, as it is implied.
> That doesn't work, because we don't have software that automatically
> selects the themes from a work. If I enter
> dis e dis e b d c a
> (leaving the first note off Fur Elise), I still want to find the same
> melody.
And besides, many works have an introduction which people do not necessarily
remember, and which is not part of the melody they would give to a search
engine.
A long time ago, I knew a musician for which the doctoral thesis was the
automatic (computerized) recognition of themes in Josquin des Pres' music.
I became familiar with his works after having been hired by an ethologist
that was hypothesising that in a particular species, fishes wer communicating
to each other through body, tail and wings (I do not know the real English
word) movements. The idea was to associate a "note" to each kind of movement
and then automatically search for "themes" in that fish music. (We also
tried some linguistic methods, after the initial usual statistics :-).
I wonder if it would not be a nice thing to aim for some automatic indexation
of themes, not necessarily bound to the beginning of a piece, accompanied
by extraction devices sophisticated enough to search either for whole teams
or sub-themes, or part thereof. Surely an interesting overall challenge! :-)
--
François Pinard http://www.iro.umontreal.ca/~pinard
