Could you explain further what you mean by this?
Yes - if you know a signal is band-limited you can use the fact that sampling aliases at multiples of the sampling rate to
allow a low rate of sampling - so long as the sampling rate is more than twice the bandwidth of the signal you can
accurately reconstruct the waveform (Nyquist's theorem only talks about baseband, but its more general than that).
So imagine a signal with components only in the 36kHz to 44kHz range. That's 8k of bandwidth, so sampling at 24kSPS
would capture all the information (assuming the samples are at points of time, not averages over a period).
In fact it would map 36kHz--44kHz down to 12kHz--20kHz as seen in the samples, which you could then process
digitally by upsampling and heterodyning with 24kHz. However aliasing does the mirror image too so you get
a reversed spectrum from 12k down to 4k as well (twice the sampling rate minus the signal frequencies).
Its not the simplest approach, but can be useful when you want to use cheaper ADCs and handle less data. Care
is required to ensure your assumptions are correct and no spurious aliased signals get in the way.
In effect its a cheap way of mixing down in frequency using the sample/hold circuit of the ADC as the mixer.
Thinking about my numbers above I think I've got it wrong as you need to keep the signal and its reflection
both within the nyquist limit, so I'd only be able to handle upto 6k of signal bandwidth with a 24k sample rate,
but that's the basic idea.