Re: Auditory objects

[Stephen McAdams <Steve.McAdams@xxxxxxxx>]

bert@EPX.CIS.UMN.EDU wrote:
>
> I am making a demo for my Hearing Science course on
> auditory object perception. I would like to create a
> demo that I heard S. McAdams present during his talk at
> berkeley a few years ago, but I don't know the details
> of the signals.
>

This demo was created by Roger Reynolds and Thierry Lancino at IRCAM
on the basis of work I was doing on FM in the early 80s.
It was taken from a piece called Archipelago by Reynolds. It was
created in the following way:

Take an oboe sound. Analyze it with a phase vocoder technique which
gives as output the functions of amplitude and frequency change over
time for each harmonic. Then resynthesize the sound with additive
synthesis, applying modifications to these functions. The modifications
applied by Reynolds was first of all to stretch the 2 sec original
sound out to about 30 seconds, by simply reading through the amp
and freq control functions more slowly (note that with additive
synthesis this will not affect the original pitch though one needs
to take care in the region of the attack). Even and odd harmonics
were sent to separate channels so that they could be individually
spatialized (in my demo version they simply come from separate
speakers). Then after the beginning of the sound during which the
oboe percept is heard slowly increase the modulation width of a
frequency modulation applied to odd partials, fade this modulation down
and then do the same to the even partials, then fade back in the
modulation on the odd partials but at a different modulation rate
than that applied to the even set. The odd and even partials segregate
into separate auditory images. The odd partials give the original
pitch and a hollow, clarinet-like timbre. The even partials give a
pitch an octave higher (surprise!) and due to the formant-like
spectrum of the oboe a soprano-like timbre. Curiously this sopranoness
depends on the presence of the vibrato.

Voila, amusez-vous bien!
Steve