🚀 go-pugleaf

On Nov 30, 3:28 pm, "Arny Krueger" <ar...@hotpop.com> wrote:
> "Travis Garrison" <travisgarri...@gmail.com> wrote in
> messagenews:6fbc5497-574a-4238-8577-ec6a5e884b93@t47g2000hsc.googlegroups.com
>
>
>
>
>
> > At the International Computer Music conference in Miami a
> > few years back, I recall attending a paper session
> > discussing this exact topic. The researchers took on the
> > task of separating out the two instruments from a Mozart
> > piano/violin duet.  As this was a piece with a readily-
> > available score, the researchers knew in advance which
> > pitches and rhythms were present in each instrumental
> > part.  The extraction was a matter of synchronizing a
> > MIDI score with the audio recording, and using that as a
> > guide to perform selective FFT bin extraction.  In this
> > case, separating the violin from the piano met with
> > marginal success.  The difficulties lie when the two
> > instruments occupy the same frequency range.  An FFT
> > analysis will give you frequency content over time, but
> > cannot distinguish whether a certain frequency bin
> > originated with piano or violin.
>
> > The resulting extracted instruments were recognizable,
> > but both suffered from washy FFT resynthesis artifacts.
> > But still impressive. Give it a few more years and I'm
> > sure someone will crack this problem.
>
> YOu seem to be conversant with the SOTA in this area. Has anybody come up
> with software that would transcribe a recording into MIDI?- Hide quoted text -
>
> - Show quoted text -

Not sure if this is close to what you are referring to, but, there are
a number of WAV/MP3 to MIDI converters:

http://www.widisoft.com/
http://www.pluto.dti.ne.jp/~araki/amazingmidi/

etc...ad infinitum...obviously, the more isoloated the recording the
better the resultant transcription. A solo Guitar, piano or bass line
works well, Jethro Tull, Thick as a Brick...not so well...but still
recognizable with most of these...well, the two or three I've tried.

Cheers, CS