You could capture the spectrum of the whole file with Audacity, which could be interesting, but you'll have no way to know what is signal and what is noise. You'll have to include a time variant to do that.
One crude way of doing it with Audacity would be to chop the file of in short chunks, and analyze each chunk by itself. Then you can compare the average spectrums of each chunk and see which parts of it varies over time, and which do not (that would be noise).
I've done much the same in the past using a program called Baudline. If you're on a PC or a Linux machine, I'm sure you can use RMAA to do the same in a more automated manner.