Find a positive peak in a data set, using a robust parabolic fitter. This looks for the highest channel in data and then fits a parabola to the half-height points around it. This tends to be a very stable way to find a peak, even on a somewhat sloping background, and with little knowledge of the real shape. It does require that the peak be wide enough that at least 3 points fit between its half-height points.
Definition at line 835 of file fitting_toolkit.py. 00835 : 00836 """find a positive peak in data, assuming the background is 0. 00837 Data can either be a list of (x,y) or just a list of y, in which case integer x is assumed. 00838 find_peak returns xcenter, hwhm, height. 00839 Algorithm is parabolic fit to data between half-height points around max of data. 00840 This makes sense if hwhm > 1 channel, so at least 3 points 00841 get included in the fit. It breaks for peaks narrower than this, 00842 but in that case, using the highest point is about the best one can do, anyway. 00843 """ 00844 da=array(data, numeric_float) #put it in a well-known format 00845 if type(data[0]) is type(1.0): 00846 x=range(len(data)) 00847 y=data 00848 else: 00849 x=data[:,0] 00850 y=data[:,1] 00851 00852 topchan=Numeric.argmax(y) 00853 topy=y[topchan] 00854 #find half-height points 00855 startx=topchan 00856 while(y[startx] >= 0.5*topy): startx -= 1 00857 endx=topchan 00858 while(y[endx] >= 0.5*topy): endx += 1 00859 00860 #f=polynomial_fit(2, xcenter=x[topchan]) 00861 f=find_peak_fitter 00862 f.fit_data(x[startx:endx+1], y[startx:endx+1], xcenter=x[topchan]) #clears the fitter and does fit 00863 c,b,a=f.funcparams #a*(x-x0)^2+b*(x-x0)+c 00864 x0=-b/(2.0*a)+x[topchan] 00865 hwhm=math.sqrt(-c/(2.0*a)) 00866 return x0, hwhm, c #returns center, hwhm, height 00867 ##
|
1.5.4