Anon,
you could either write some code that converts .wav to hdf5 according to the examples in the "fileimport" module (look at the source), or you could subclass the TimeSamples or MaskedTimeSamples class. Depending on the hardware and recording process it could also be a good idea to directly store the data as hdf5.
I think the problem that would be to solve here is the import of a multichannel .wav file into Python. Probably the scipy.io.wavfile module could help.
Regards,
Ennes