Sound Samples

Hi

I'm looking to use neural networks to analyse sound samples in *.wav format,
and to be able to differentiate between them.  What would be the best
approach to take?  What aspects of the sound files should I be trying to
extract to differentiate between them?

As a human, I can't generalise the files into type A and type B, and I
wondered how I would feed them into a neural network for analysis.  I can't
feed every bit of data through an ANN, can I?

Sorry for being so vague: if there is any other information needed to answer
my question, I will happily provide it.

MadJock