From: malcolm@interval.com (Malcolm Slaney)
Date: Sun, 9 Oct 1994 17:14:11 -0700
Subject: Pure Vision/Pure Audition at CCRMA
Message-Id: <9410100011.AA18636@interval.interval.com>

At this Thursday's CCRMA Hearing Seminar, I'll be presenting an
audiologist's review of "A Critique of Pure Vision."

  "A Critique of Pure Vision", by P. Churchland, V.S. Ramachandran and
  T. Sejnowski. In: Large-Scale Neuronal Theories of the Brain, by
  C. Koch and J. Davis, MIT Press, 1994.

Why, you might ask, is a vision paper being presented at a Hearing Seminar?
There is no direct answer, but I hope it will give us some insight on how
to talk about and organize our models of audition.

Every perceptual sound separation system to date as started with an
architecture proposed by David Marr (from his book Vision).  The systems
are purely hierarchical, with information and decisions flowing smoothly
from low-level detectors, to high-level objects.  But this chapter argues
that the neurophysiology is not so straightforward.  It's not a trickle of
information that flows top-down, but a torrent.

Quoting from the introduction:

 "We shall  begin by  briefly  presenting, in its  most extreme version,
  the  conventional wisdom [called Pure  Vision]. ... We then outline an
  alternative approach, which,  having lurked on the  scientific fringes
  as   a theoretical possibility, is  now  acquiring robust experimental
  infrastructure. ...  The  idea of   "pure vision"  is   a fiction,  we
  suggest,  that obscures  some  of  the  most important   computational
  approaches  used  by the  brain. Unlike  some  idealizations,  such as
  "frictionless  plane" that can    be    useful in achieving  a    core
  explanation,  "pure vision" is a notion  that impedes progress, rather
  like  the notion of   "indivisible   atom".  Take individually,    our
  criticisms of   "pure vision" are neither   new nor  convincing; taken
  collectively  in a computational  context, they make a rather forceful
  case."

I'll be reviewing the chapter on Thursday and relating it to the audio
perception world.  I've got computer animations of the examples from the
paper, and I'm working on audio analogues.  This should be interesting.

        Who:    Malcolm Slaney (Interval)
        Why:    Because the audio and visual worlds are more alike...
        What:   Pure Vision/Pure Audio
        When:   11AM, Thursday October 13, 1994
        Where:  CCRMA Library (Top floor of the Knoll at Stanford)

See you at CCRMA!

-- Malcolm