I have a radar that can scan its surrounding. If there is a new object in a previously empty environment, it will cause a bell curve in the amplitude of the light waves coming back. If there are two objects, they will produce two bell curves and if there are n objects, they will produce n bell curves, and because the objects are all roughly the same size, they should produce a corresponding bell curve with the same standard deviation, just a different mean.
If I have the data from a radar (which theoretically looks like the sum of several bell curves), is there a way to algorithmically decompose the data and produce the points at which I think an object can be?
Illustration: If I have the green curve (which is the sum of the yellow and blue curve), is it possible to solve for the component yellow and blue curves?
This is a typical problem in signal representation. Each of the gaussian function has its own parameters; these are unknown but, in a nice case as the one you gave in your post, they can be estimated visually. Now, you have the data for the overall signal. You just need to fit your data (nonlinear least squares method) using as parameters those of the two gaussian curves.
I suggest you look at http://en.wikipedia.org/wiki/Spectral_line_shape