Re: Cochlear AGC attack/releases times?

[David Mountain <dcm@xxxxxx>]

I'll start my answer with my engineer's hat on.  I usually
usually refer to nonlinearities that are essentially instantaneous as
"compression" and nonlinearities that respond to some time average of the
signal as "automatic gain control" or AGC.

The auditory system includes a lot more than the cochlea so the answer to
your question is not straight forward.  If you are asking about the
nonlinearity in cochlear mechanics, I would say that the answer is that it
is essentially instantaneous (comparable to the rise time of the cochlear
filters) and hence refer to it as compression.  Additional compression
takes place in the hair cells due to the nonlinear nature of the
transduction process.  The synapse between the inner hair cell and the
auditory nerve fibers is best described as an AGC with two or more time
constants. There is a fast component to the gain change (adaptation in the
physiological jargon) that is the major component and has a time constant
in the 2-4 ms range. There is a slower component with a time constant in
the 10's of ms and there are other components that operate on time scales
over over 1 s.  Additional adaptation takes place in the central nervous
system.

The cochlear efferent system also can alter the gain. The primary efferent
effect build up with a time constant in the 50-100 ms range but here again
there are longer term effects as well.

--------------------------------------------------------------------

David C. Mountain, Ph.D.
Professor of Biomedical Engineering

Boston University
44 Cummington St.
Boston, MA 02215

Email:   dcm@xxxxxx
Website: http://earlab.bu.edu/external/dcm/
Phone:   (617) 353-4343
FAX:     (617) 353-6766
Office:  ERB 413
On Mon, 23 May 2005, Richard H. wrote:

> Hi,
>
> Does anyone know how the cochlear AGC can be modelled?
>
> Suppose a short tone burst hits the cochlear, how long does it take for the "gain" to be reduced? And then to recover?
>
> Are we talking 1 millisec or 1000 millisecs? [Or 0 millisecs i.e. the transfer function is more like an amplitude compander, without
> any time aspect]
>
> And is this AGC very frequency specific - narrow nand, or perhaps critical band?
>
> Or does the cochear not really work like an electronic AGC [or compander] at all?
>
> kind regards,
>
> Richard
>
>