Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: [Phys-l] De-chording polyphonic music with software.

Wow! This is amazing. Note that the software is evidently doing something much more difficult than a Fourier spectral analysis. The implication is that it is determining which partials belong to each note being played simultaneously by a multiphonic instrument like a piano or a guitar and bundling them together so that each note can be transposed individually preserving its relative timbre without affecting the timbre of any other notes that are being played at the same time. That does seem to me to be an impossible task in general because the problem is clearly underdetermined. Consider, just for instance, playing a pair of notes separated by an octave and trying to determine which partials belong to which note. It seems to me that this could only be accomplished by making some arbitrary--if, possibly, intelligent--decisions along the way.

I notice that the demonstrations are apparently limited to recordings of single instruments and I suspect that's no accident. I'd be seriously blown away to hear piano notes altered within the context of a concerto!

John Mallinckrodt
Cal Poly Pomona

On Jul 18, 2008, at 9:12 AM, Bob Sciamanda wrote:

Suppose you record from a single microphone, onto a single track, a
polyphony of music from several instruments. Here is software which will
disect the resulting single sound track into its "component pieces" and
allow you to manipulate the individual pieces. You could, e.g., change a
single note within a chord, even though your original recording gives you
only the single track waveform of the composite chord; you could
individually manipulate the sound from each guitar string given only the
composite chord waveform, etc. (You can tune your guitar in
post-processing!) The results appear to be quite striking . . . limitations
are to be expected. Take a listen.

For a demo go to this URL and click on the play button of the video shown in
center screen:

Bob Sciamanda
Physics, Edinboro Univ of PA (Em)

Forum for Physics Educators