That's true; now you just need to determine the location of the relevant peaks (whether it be the spectral width of the peak, or using a simple thresholding algorithm--if it's above value X then it's a peak--or running the data through a low-pass filter to smooth out the noise prior to finding the peaks).
Or, if you know a priori where the peaks ought to be, you can use these peak locations to create an appropriate comb filter which filters out only the values at those wavelengths (and of a certain spectral width, once again). But again, you'd need to know exactly what you're looking for, and either compensate for spectral shifting or make the comb sufficiently wide that you don't end up cutting off the peak.