I asked how to programmatically judge spectrum bands and @detly suggested using FWHM (full width at half maximum) to determine the widths of the peaks. I searched around and found that FWHM can be used for fitting models (I am practically a layman to this!), especially Guassian models. Specifically, 2.354 * sigma is the width for a Guassian model.
I am looking at a Guassian fit because of the presence of bad peaks. There are 4 well-formed peaks in this picture. Then there is the "double" peak (while both may be important) and two spread-out peaks. They will prove an impossible challenge to a naive FWHM.
Can you help with generating a guassian fitting (for purpose of FWHM calculation) of a peak in Scip/Numpy, given it's approximate location of its x coordinate? If Guassian is a bad choice, then some other scheme.