How to fit multiple guassian peaks ?

[banyal]

Hi,

I have a  two column data set (x,y) that contain large number of Gaussian peaks ( several hundredes infact). You can think of it more like a wavelength spectrum. The number of spectral peaks is not known exactly. There is no problem in detecting the peaks as they are of same amplitude. Eventually I want to fit a Gaussian funtion to each of these spectral  peak and obtain the the parameters  such as  amplitude, line-center and sigma.  This seems a somewhat tricky in LabVIEW. There is an example ( Sum of 3 Gaussian with offset) that comes bundled with LabVIEW. This example VI however uses a static refrence to a model function which is sum of three gaussian curves.  This works well for few number of peaks in the data. If we have large number of peaks ( e.g spectroscopy data) then creating such a model function for a large number of peaks is nealy  impractical.  Is there a  better way to handle  such peak-fitting problems in LabVIEW? Would greatly appreciate if someone could point out any existing  LV code/library function.

Regards,

RK

[altenbach]

Well, all you need to do is make the model function scaleable so it generates an arbitrary number of peaks, based on the number of fitting paramters given. (See e.g. example #2 here)

You could get initial estimated from the Peak detector.

Can you show us a typical spectrum?

Are all widths similar?

[BowenM · Accepted solution]

I looked in to this problem a few years ago.  The problem I encountered was that for each peak in my signal, I was basically trying to fit 4 parameters... so 10 peaks = 40 individual parameters that had to be approximated for the fitting function to even be close.

I ended up doing exactly what altenback said:  I used the built in peak detect, "guessed" the edges of the peak, split the array, and then did a sum of 2 gaussian + background fit for that split array. 

[banyal]

Many Thanks Altenbach. That was something I was after. I've attached an example data file of the sep...

Regards,

RK

[starbob]

Greetings,

I was perusing the message boards, specifically the discussion on fitting multiple Gaussian peaks within a single spectrum, and I came across the response you posted. Just wondering if you still have the VI for such a functionality and if you would be willing to share it.

Best,

Stuart