[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: spatial separaton
I'm researching auditory spatial perception, especially with reference to
artificial environments. It seems to me that the problems are reciprocal
(and surprisingly similar) to artificial perception in real environments
Surely, when speaking of 'real' perceptions of 'real' sources, the role of
exploratory behaviour of ambulant organisms relies on the interdependance of
so many factors that themselves relate to holistic gestalts. Analysing those
gestalts by varying one factor at a time is potentially unrewarding. Another
approach to experiment design might be to look to experimental psychology
(where Gibson's 'ecological' ideas have thrown up some useful insights).
I'm personally optimistic about the possibility of convergence of
approaches such as Jackendoff's computational theory, and ecological theory,
to form a richer framework whereby a number of parallel computation-modules
form a problem-solving network rich in redundancy. this gets round
'bottom-up / top-down' questions which still imply sort of two
unidirectional information streams which somehow meet in the middle (sorry
for simplifying). The required feed-forward /feed-back /feed-sideways
information flows might better facilitate Gregory's
'perception-as-hypothesis-testing', which seems a much better bet than
behavourist 'stimulus-response' models.
In this way, differing 'higher order' (I've some reservations about the
term) classes of information-processing might well contribute in generating
perceptual hypotheses on an equal footing with 'lower-order' (same
Simply, if spatial separation is evidenced in inter-aural differences, a
perceiver will use that, but if not, not. But over time, and with the
capacity to move through an environment, one can deliberately generate
inter-aural differences if need be. This is especially so for comparitively
local sources (triangulation and all that), and increasingly difficult (ie -
takes more time and effort) for 'distant' sources. As it should be. so
"> Perhaps because of the integration of consitstant interaural cues over
time yeilding a clearer 'picture' of the scene?"
...- sounds like a description of a specific memory function, to me. And
that memory is especially useful in conjunction with movement!
"Yes, perhaps the human is better at Auditory SA than the machine,
> because people are especially good at integrating all the information"
That's basically all the human brain does!
I'm just working on the notion of "selective INattention" as a contributor
to the functioning of selective attention. In other words, the two
processing functions work in parallel.
In fact this discussion touches on the question I asked earlier of the list
for opinions on "sound fields" as against "sound environments"; what we
actually perceive are the latter, but what we use to conceptualise
perception are the former. Perceptual segregation in the former is
relatively poor, compared to that in the latter. In 'real' environments,
some classes of information are manifested only in richly varying conditions
through space and time; controlling variables in experiment design using too
strict parameters might well negate exactly those classes of information
which facilitate what we actually mean by "perception", as against the
simpler 'stimulus-response' type reflex which is manifestly not what we mean
by perception. Given that even the simply defined 'precedence effect' has a
substantial 'soft' component (Moore, Brian, AES16th .,1999), richer
hypotheses are needed. Simply enriching 'bottom-up models with a leavening
of 'top-down-ness' is not enough.
Are there any experimental psychologists out there?
----- Original Message -----
From: "Ward Drennan" <ward@IHR.GLA.AC.UK>
Sent: 01 February 2001 10:26
Subject: Re: spatial separation and ASA
> > In other words,
> > in cases where two signals were partially segregated by factor X,
> > adding a difference in location would strengthen the segregation.
> By separating spatially, you'll get a change in signal to noise ratio at
> the two ears that will inevitably help. I think the work of Plomp (1976,
> Acustica 34, 200-211) on speech intelligibility sheds some light on
> this. In an anechoic environoment, spatial separation of two speech
> sources gives a 5 dB advantage for speech intelligibility. If the
> environment gets very echoic though, the advantage drops to 1 to 2 dB.
> Note that head shadow plays a big role here. If you were to low-pass
> filter the speech (limiting mostly to interaural timing cues), I bet the
> advantage would drop to 2 or 3 dB in the anechoic and be non-existant in
> Perhaps there are some more studies showing this? Speech has a number of
> segregating cues, but can anyone point to more constrained studies that
> show this, maybe with only one factor X?
> >We know that spatial
> > differences do play an important role in SEQUENTIAL grouping.).
> Perhaps because of the integration of consitstant interaural cues over
> time yeilding a clearer 'picture' of the scene?
> > My point about spatial information was not that it was very weak
> > but that it was only one of a number of cues for sound
> > separation. To this should be added the idea that it may need
> > those other cues in order to be effective itself.
> Yes, perhaps the human is better at Auditory SA than the machine,
> because people are especially good at integrating all the information.
> Best Regards,
> Ward Drennan
tel: (0114) 2661509