Re: OT - Enough to Gag a Maggot

[jporter@xxxxxxxxxxxx (Julie Porter)]

>On 11 Sep 2000, at 23:52, Julie Porter wrote:
>> Another
>> problem I recently ran into is in music to take a analog recording
>> (*.wav) and convert it to a MIDI file, recording the keys and timings
>> of the instrument played. A person with a trained ear can do this
>> without thinking.
>
>Hi Julie,
>It is  not easy to transform a sound to MIDI unless it is a
>monophonic source. I know that there are a couple of programs
>claiming to do that, but I was told that none works very well.
>The problem is that while the wav is the sound data, the MIDI format
>is just the "note" data. The transformation programmes usually work
>by Fourier Transforming the data and looking at the peaks in the
>spectrum (those should be the fundamental) but as you can
>imagine, different instruments have different spectral
>characteristics so more than 1 note is difficult to discern.
>I think that a programme called Cakewalk (version 9?) has a facility
>to do that. I do not know how good it is.
>
>Cheers,
>
>Gabriel
Actually I was aware of this. The point I was making is that our aproach is
simular to MIDI, in that FSG et all represents the positions and timing,
but not the harmonics.
Much of the discussion on the muisc box list results in the same as the
above. Interesting that you can pick up in one post what took 20 or 30 in a
diffrent context.
In the end it all comes down to brute force. Try as many combinations as
fits the model. In our case what have we done to define the model?
-jP