2011/10/10 Kassen <signal.automatique_at_gmail.com>:
> Hey Dave,
>
>> If anyone has any pointers to where to look for ways to analyse simple
>> discrete sequences of notes it would be helpful for my fitness
>> functions!
>>
>
> I think your strategy of seeing at what number of steps the pattern is
> similar to itself is quite good. I think it could be extended by saying that
> notes that are a octave apart are also "almost the same" while notes -for
> example- 7 semitones apart might be "a little the same". A similar analysis
> could be applied to timing/rests.
>
> Furthermore phrases that are (mostly) within one scale (this will require a
> little library of scales) could also be seen as more fit.
>
> From there I would say that you would be well on track to filtering out
> phrases that sound "musical" to us, but not yet that sound like "good
> music". For "good music" I think you need a mixture of "expected" and
> "unexpected" events. Expected might be straight repetition in the timing and
> each note being at one of a small set of intervals from the last. I think
> that if you keep "breeding" your phrases with just the above rules and
> others like it you might be breeding towards more "bland" sounding phrases
> if you don't also include some encouragement towards "the unexpected". Now
> that I think of it a balance like that could be more appealing final results
> than just favouring lots of different notes. That last option might
> discourage riffs that include a lot of repetition of a single note, even
> though those could be musically nice.
>
> That's what I can come up with now, there are probably texts on the subject
> that are more well researched than me thinking out loud over post-dinner
> coffee :-)
For musical interestingness, the idea of the Wundt curve works pretty
well - you need to be unpredictable but not too unpredictable since
unpredictable is just noise. So one thing you could do for discrete
melodies, if they're not too short you could use something like a
markov model to predict the later parts of the melody based on the
early part. If the trained markov model improves prediction relative
to a uniformly-distributed markov model, but not to perfect
prediction, maybe that's useful...
Dan
Received on Tue Oct 11 2011 - 07:23:52 BST