The vOICe to AIM (at)

Subject: The vOICe to AIM
From:    at <meijerNATLAB.RESEARCH.PHILIPS.COM>
Date:    Mon, 16 Jun 1997 09:28:02 +0200

                                            June 16, 1997

I'd like to thank all of you who responded, publicly
or privately, to my inquiry on auditory models (subject
was "Re: Time and Space"). So far, it seems that consensus
about the validity of any model for the kinds of auditory
profile analysis that I'm after is rather unlikely.
Of course, that does not automatically imply that there
are no models that might be up to the job, but just that
there is no widespread agreement about it yet. I personally
do not take a position in this issue, because I simply do
not know how good or bad the available auditory models are.

As an auditory model testbench, I have now put a 1.05
second, 20 kHz sample rate, 16-bit mono, .wav sound file
on my site (see URL below). The actual frequencies present
in this sound are in the range [500Hz, 5kHz], and the
sound was synthesized by my Java application from the
arti2.gif test image at my site (again, see URL below).

In order to get at least some idea of what a contemporary
(and readily available) auditory model would make of my complex
sounds, I have now run a few experiments using Roy Patterson's
AIM model. First draft results can be found on my new page at
URL

   http://ourworld.compuserve.com/homepages/Peter_Meijer/aumodel.htm

where I used as much as possible the AIM default settings
in generating auditory spectrograms, basilar membrane motion
plots and neural activity patterns from the AIM model.

In other words, I did not search for a special corner in
parameter space to squeeze out seemingly best results: I only
adapted the AIM output gain to get proper plotting ranges.
I'm sure the results will be controversial, but it illustrates
how an auditory model may in principle be used to investigate
auditory processing of auditory images as generated via
The vOICe image-to-sound mapping. I hope this will in the
longer run help bridge the gap between the available large
body of "microscopic" knowledge of auditory processing and
perception and the "macroscopic" level of complex sound
processing as required for mental reconstruction of sonified
imagery. The vOICe mapping might serve as a research vehicle
in bringing many issues in auditory perception together
within a single, concrete and conceptually simple framework.

Results from other auditory models may be added at a
later stage when other models become available to me.

Best wishes,

Peter Meijer