Another interesting fact is that the beginning of a note (the "attack") is a
more important defining characteristic than its sustained quality. You can
splice a saxophone attack onto a sustained oboe note, and most people will
be fooled into thinking its a saxophone the whole time. The attack is also
the most difficult aspect of tone production for the performer, at least on
my instrument, the clarinet. After about 12 years of playing, I feel like I
got it more-or-less perfected this January.

Anyway, automated recognition might try to make use of this information, but
occasionally its obscured and unavailable. Those are the times when I have
the hardest times separating parts in a recording.

Eddie

----- Original Message -----
From: "Jecel Assumpcao Jr" <jecel at merlintec.com>
To: "The general-purpose Squeak developers list"
<squeak-dev at lists.squeakfoundation.org>
Sent: Thursday, September 18, 2003 9:07 PM
Subject: Re: music editing / decomposition


> Besides the complications already mentioned, it is important to remember
> that human sound source separation is greatly enhanced by the shape of
> the ears combined with slight movements of the head.
>
> If you sit at a party situation, you can easily focus on different
> conversations. But if you do a high quality stereo recording from your
> position and then listen to that, you will find it far more difficult
> to do the same thing. So a FFT/AI/whatever working from an MP3 starts
> out with a huge disadvantage since so much information has already been
> lost.
>
> -- Jecel
>
>