This can be solved with non-linear curve fitting, but it will not be a trivial solution. If you are unfamiliar with non-linear curve fitting techniques, find a copy of "Numerical Recipes in C" by Press et. al. and read the chapter on Modeling of Data. LabVIEW contains the building blocks you need to perform any of the methods mentioned in the book (e.g. Levenberg-Marquardt and downhill simplex). You may want to modify them for speed.
The general method is to model the two peaks as two gaussians (or whatever function they are) with an unknown distance between them. The curve fit will find the distance between the peaks and the peak parameters so you can subtract the small peak.
WARNING: This method (and most others) gets fuzzy when the peaks merge. Be sure
to use the techniques to find out how good your fit is, or you can run into many problems.
You can probably solve this problem using convolution and/or fourier transform techniques, but I have never done that, so cannot comment on it.
Good luck!