radams2000 at aol.com (RAdams2000) writes:
>Here is a basic question from someone who knows a lot
>about signal processing but only a little about the human
>sensory organs. Given that there are about 4000 nerve fibers that
>connect the basilar membrane in the ear to the barin, and that
>each fiber only fires at a 500 Hz or so rate, why dont we
>hear the collection of firing events as noise?? The standard argument
>seems to be that if you average (filter) some number of these fibers,
>the "noise" would be eleiminated, but a few basic calculations show
>that even a few thousand fibers averaged with a time constant on the
>order of a millisecond would only have a signal-to-noise ratio of
>40 dB or so; not exactly hi-fi! And the signal-to-noise ratio only
>increases as the square root of the number of fibers, so you dont
>win very fast by increasing the number of fibers included in the average.
>The same question applies to the eye, why dont we see snow??.
>You can argue that the brain does all sorts of eleborate processing,
>but noise is noise; its not predictable and even the most advanced
>signal processing could not eliminate it after the fact.
It's got a lot to do with pattern recognition - if an incoming pattern
resembles a familiar one, then it will probably be recognised as such,
and the patterns that fire a few neuronal levels later will be identical
to the ones that would fire if there had been no noise.
Lateral inhibition would also eliminate the amount of noise as the
pulses go from one level to the next. Of course, if the signal/noise
ratio is too low, the noise could inhibit the signal, as can happens
in instances of tinnitus.
It's also probable that selective attention, in the form of feedback
reinforcement of the signal, would contribute to the strength of the
signal, and lateral inhibition would then damp the noise. This is the
sort of effect that occurs when someone listens to a single voice
among a babble of voices.
Hope this helps. A more thorough account of how the brain works can
be found at http://www.maths.tcd.ie/~rof/neuro.
--
Ruadhan O'Flanagan
"Quidquid Latine dictum, sit altum viditur."
