The significance of representation

["K.Baker" <PC1KLB@xxxxxxxxxxxxx>]

Dear Auditory List,

I have been recently pondering on the problem of how to represent
the acoustic signal which hits the ear as a first stage process for
a computational model of auditory scene analysis. However, as I have
pondered, the light has become faint to the point where I think I
may be getting bit lost.

It is clear that most animals have evolved to solve the problem of
identifying more than one (concurrent) sound event from the
aggregate sound-pressure waveform. This may be considered a basic or
fundamental task for animals to master in order to increase their
chances of survival in the environment. As I understand it, the
functioning of the peripheral auditory system in many animals
differs quite markedly in the frequency resolution, critical
bandwidth distribution along the basilar membrane, etc. One of my
ponderables is: Does this indicate that the initial transformation
of the signal is irrelevant (to a large degree) for the ability to
identify sound events? If this is so, does this imply that any old
representation of the signal will do for a basic model which is able
to accomplish auditory stream segregation?

I am aware that the degree of resolution and functioning which
different cochleas (cochleae?) produce will make some difference to
the auditory 'skills' of the animal. My point is that the skill (or
primitive process) of stream segregation is not primarily reliant upon
this initial representation of the signal.

I would be extremely grateful if anyone could help with this issue.
May be I am looking at this completely the wrong way? Or may be I
have started off with some mistaken assumptions?

Kevin L. Baker
University of Sheffield, UK

email:
>From UK - pc1klb@uk.ac.sheffield.primea
>From everywhere else - pc1klb@primea.sheffield.ac.uk