Re: Information loss (Re: Analytical approach to temporal coding...)

[Ramdas Kumaresan <kumar@xxxxxxxxxxx>]

(I tried posting this before, but somehow it did not get posted)

Dan Ellis, John Hershey and others,

Clearly the signal representation at the periphery can not afford to
throw
away  too much information or must  retain as much info as possible. Any
throwing
away  has to be done at higher stages. We have just proposed an analog
to
time converter. This type of conversion at the periphery  is deemed
necessary because the higher  nuclei seem to depend on spike
rate/coincidence detection (i.e. 'time' processing)  to further refine
the
representation.

We agree with John Hershey that an analysis of how the original signal
is
altered if the temporal code is perturbed needs to be done. But the
preliminary
evidence is that it is quite stable, unlike the traditional
zero-crossing
representations.

Ramdas Kumaresan