|
   
   
A simplified diagram of the auditory pathway
One of my main research interest is the development of analog electronic models of the
auditory pathway. I have 2 main reasons for building models of neural processing in the
auditory pathway:
 |
Firstly, a
faithful neural model will allow the study of neural processing in the brain without the
need for living neural tissue and without the need for simultaneously probing hundreds or
thousands of neurons in the brain. Once we have neuron models with the same transfer
function as biological neurons, and knowledge of their input and output connections, we
can build a hardware model that allows us to study the collective behaviour of such a
neural network and the individual behaviour of the single neurons in the network. |
|
Secondly, we
can hope that by using similar processing strategies as the brain, it will be possible to
perform perceptive tasks with the hardware model hereby approaching the excellent
performance of the brain on such tasks. In this case, the neuron model might not need to
replicate the biological neuron as faithful as in the first case. However, since the
actual amount of detail that has to be included in order for the model to function is
unknown, the only safe way to build such a model is to reproduce the biological neuron as
faithful as possible. Once a functional model is built it will allow us to distil the
amount of detail needed for a certain function and thus to create a simpler model. Because
of natural evolution it is very well possible though that most of the details of the
processing performed by a biological neuron are important. |
In any case it is not only a faithful neuron model which will allow us to build
hardware implementations of neural processing. The total amount of neurons that we can use
in the model will be as important. This thus leads to a trade-off between the size of the
single neuron circuit and the amount of detail incorporated in it.
I've already developed building blocks for the 3 neccesary ingredients, i.e., a Silicon Cochlea, an Inner Hair Cell model,
and a Spiking Neuron model.
For the moment I've used all these elements together to build models of Amplitude Modulation (Pitch) detection and a model of low frequency
periodicity detection in the auditory brainstem. |
