[Pd] OT- FFT and human auditory cortex

[padawan12]

Very interesting. You two Chuck(k)s certainly got me thinking
over some old stuff, but this left right thing is really 
fascinating. For one thing it makes me think maybe there's
more to localisation than we know about right now.

> There has been a long standing debate over functional specialization
of the different hemispheres.
  
Can only comment from a cog sci perspective, but I remember a general
theme in Fodor and some other texts that we often see complementary
behaviour with one hemispherical faculty taking up the loose ends or
blindspots left by the other. In that regard all our senses are
"stereoscopic", or at least not as discrete per organ as we assume,
we like to get two points of view from which perception emerges.
(eg taste is linked to smell) 

Since time and frequency acuity are mutually exclusive at the limit
wouldn't it make sense that the brain would evolve to have one side
process things using a complementary principle than the other?
  
>What's really going on is that subcortical structures in
>the auditory pathway differentially project to the right and left
>hemispheres.

Precisely. 


 
> The cochlea is actually a dynamic organ, in and of itself

And can exhibit very localised and very fast adaptive behaviour almost
within itself as a protection mechanism. This is how it achieves such 
an awesome dynamic range,  after lying in a soundproof isolation
tank for many hours subjects report hearing a hissing noise which we 
assume is brownian motion of air molecules. On the other hand an impulse
at 130dBU can effectively shut down the auditory system and dilate the
tympanic membrane with the same kind of reflex to an eyelid closing.
I've always found this amazing. That our senses are capable of picking
up the smallest practically measurable "quanta" of information (eg
the eye retina can detect a single photon of light) 

> tonotopically

What do you mean by that term please? In simple language.

> I'm searching my brain for a better term than timing
> encoding, but not sure
 
PPM (pulse position modulation/encoding)? Well, that's what I'd call
it with my electronic engineers hat on.


> As we proceed through the auditory pathway, the ability to
> encode timing decreases,
 
See below vis Shannon, timing is increasingly less relevant higher up
the tree.

> so that at the cortical level, timing
> encoding only pertains to the low <100-200 Hz frequencies.
  
The magic number at ten to twenty milliseconds has fascinated me
for ages. It comes up time and again, in Garbors work, in Warren,
Jones and Lee (all cognitive scientists - check McAdams and Bigand
compilation "Thinking in Sound"). Something very important is 
happening here. It is the essence of granular synthesis and marks
the important point where discontinuity becomes continuity 
(eg grains "fuse" at this point and we have to move to a wavelet
like model where position and frequency become combined) Can you
throw any hypothesis at this speaking as neurobiologist?
Do you think maybe one side of the brain is taking over from 
the other?

>okay, it goes cochlea->auditory meatus (midbrain ~
>pons/mesencephalon)->lateral lemniscus (of pons)->superior olivary
>complex (of pons)->inferior colliculus->medial geniculate nucleus (of
>thalamus)->primary auditory cortex. 

<Vader voice> Impressive :) I bet you didn't have to look that one up
either.


> Ultimately, there are several
> different ways that sound is encoding up to and including the primary
> auditory cortex.  It is "like" and FFT, but it is also "like" a
> wavelet transform, and it is also "like" a bank of hair-trigger
> resonators.

It's not defeatist to assume that maybe the brains internal encoding 
is quite unlike, and doesn't map elegantly onto, our traditional
mathematics at all. Not even on the level of fuzzy ANNs. That would
make these mathematical "utilities" no less powerful, rather like wave
particle duality doesn't really weaken either of the models. In fact
it makes perfect sense, in the terms of Shannon a function of time 
is being compressed in semantics, it is reduced to an ever smaller
set of more salient features. The crazy thing about this adaptive 
system is that it's the brain itself which is deciding (in a dynamic
fashion) what these salient features are according to context.
I remember from somewhere that in 60ms a human or monkey can tell the
difference between a bolt of lightning, a twig snapping and a raindrop
on a leaf. All this information must be in the very small attack of
a signal. It happens long before the frontal brain can classify
and tag the sound with a word. While it's obvious from a survival
POV in evolutionary biology it also indicates that adaptive feedback
must be occurring at a very low level. In that sense the ear (cochlea/
transducer) can "focus" rather like the eye. I wish I could remember
the experiment, but I thought at the time it seemed a bit wooly,
involving lots of retrospective timing assumptions.
It would be interesting to know if any advances have been made with
invasive potential measurement since. My synthetic sound design work
has frequently confirmed this in practice, where very tiny changes 
to the spectrum or shape of the attack portion from 10-100ms completely
changes the emotional response to a sound.




On Wed, 24 May 2006 11:25:53 -0500
"Charles Henry" <czhenry at gmail.com> wrote:

> Okay, just had this material in a class this semester...and I've done
> some research...
> There has been a long standing debate over functional specialization
> of the different hemispheres.  One hypothesis is that the left
> hemisphere deals with high frequency information, and the right deals
> with low frequency information.  This does not mean that the auditory
> information is actually divided disjointly to the hemispheres, but as
> a general theme in lateralization, the left hemisphere resolves higher
> frequency information...
> This hypothesis explains that the left hemisphere can represent fine
> timing information by having a more full set of frequencies, and that
> the right hemisphere resolves pitch contour by pertaining more to slow
> changes in frequency.
> 
> but....the neurological side doesn't really compare with this
> hypothesis.  It's not something that can be boiled down to a single
> hypothesis.  What's really going on is that subcortical structures in
> the auditory pathway differentially project to the right and left
> hemispheres.
> 
> in an article by Ligeois-Chavel (2001, New York Academy of Sciences),
> intracerebrally recorded evoked potentials (IEPs) were taking during a
> simple pitch experiment.  The IEPs showed that the right hemisphere
> tonotopically encodes pitch information->there were position different
> "signatures" like event-related potentials that varied with respect to
> frequency.  In the left hemisphere, there was no tonotopic
> organization, the areas of the brain under study responded equally to
> a wide band of frequencies.
> 
> While the coding in the auditory cortex is "like" an FFT, it is NOT an
> FFT.  there are many fine differences....the whole auditory pathway is
> not real clear cut where functions are localized, exactly.  At
> different stages in the auditory pathway, frequencies can be
> tonotopically or rate encoded.  The cochlea is actually a dynamic
> organ, in and of itself->it's not a passive organ with a system of
> resonators and transducers, it's like a bank of critically tuned
> resonators at unstable equilibrium (like a hair trigger).  The cochlea
> encodes a train of phase locked pulses, which are transmitted by the
> auditory nerve, which contains both tonotopically and timing encoded
> (fine timing) information.  The cochlea can represent frequencies with
> timing encoding (I'm searching my brain for a better term than timing
> encoding, but not sure) up to around 5 kHz (it seemed a little high to
> me, when I first read the number...I thought it would be around 1
> kHz).  As we proceed through the auditory pathway, the ability to
> encode timing decreases, so that at the cortical level, timing
> encoding only pertains to the low <100-200 Hz frequencies.  Hence, the
> primary auditory cortex accomplishes things by rate encoding (i.e. the
> strength of the tone is represented by the frequency of neuron action
> potentials, and not by timing between them)
> 
> (I can't remember these too well at the moment, I just graduated this
> spring...and my brains a little fried...there may be more.  I just
> feel like I've left some out)
> okay, it goes cochlea->auditory meatus (midbrain ~
> pons/mesencephalon)->lateral lemniscus (of pons)->superior olivary
> complex (of pons)->inferior colliculus->medial geniculate nucleus (of
> thalamus)->primary auditory cortex.  (the named areas occur on both
> sides of the brain.  There is also, of course, and auditory chiasm
> taking place somewhere around the superior olivary complex)
> 
> It's not real clear what each of these things does, and how they
> contribute to specific functions of the auditory system.  There are
> some differences in encoding at different areas, such as the inferior
> colliculus.  The IC is organized into isofrequency layers, each layer
> encodes a different band of frequencies (this is more "like" a wavelet
> transform), and fine timing information here works to accomplish sound
> localization (also has to do with eye position information from
> superior colliculus)
> There may be other information that is encoded at the level of the IC
> (I'm leaning strongly towards pitch in the IC).  Different kinds of
> information are projected from the subcortical structures to the
> cortical levels...it just depends upon what kind of processing is
> taking place, and how it's organized.
> 
> but the point (there IS a point) is that there are many different
> representations of frequency information along the auditory pathway.
> Each structure named does some kind of processing and passes the
> processing along to the next structure.  Ultimately, there are several
> different ways that sound is encoding up to and including the primary
> auditory cortex.  It is "like" and FFT, but it is also "like" a
> wavelet transform, and it is also "like" a bank of hair-trigger
> resonators.
> 
> Chuck
> 
> On 5/24/06, Chuckk Hubbard <badmuthahubbard at gmail.com> wrote:
> > http://cercor.oxfordjournals.org/cgi/content/full/11/10/946
> >
> > I came across this fact researching my final paper for Perception
> > class: the left auditory cortex is known to resolve temporal changes
> > in sound better, while the right auditory cortex resolves tonal and
> > harmonic information more finely.  As soon as I read this I thought of
> > FFT.  Could it be the difference between the brain hemispheres is
> > related to the auditory cortices having different block sizes?
> > -Chuckk
> >
> > --
> > "Far and away the best prize that life has to offer is the chance to
> > work hard at work worth doing."
> > -Theodore Roosevelt
> >
> > _______________________________________________
> > PD-list at iem.at mailing list
> > UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list
> >
> 
> _______________________________________________
> PD-list at iem.at mailing list
> UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list