How to fit a Gaussian curve by code?

Hi,

I want to fit a Gaussian curve by coding because I want to fit many type of 1-D data. I'm not using the "curve fitting" app.

However, when I write the code below, the result has something wrong.

I hope that the result should be the same as using "curve fitting" APP.

load('Gaussian.mat');
x0 = [0 0 0]; 
fitfunc = fittype('a.*exp(-((x-b)/c).^2)');
[fitted_curve,gof] = fit(x,y,fitfunc,'StartPoint',x0);
% Save the coeffiecient values for a,b,c and d in a vector
coeffvals = coeffvalues(fitted_curve);


% Plot results
figure(2)
plot(x,y,'r');
hold on
plot(x,fitted_curve(x),'k','LineWidth',1.5);
hold off
xlabel('Time');
ylabel('Voltage');
title('sinusoidal drive waveform & fitting curve');
set(gca,'fontsize',14);

[Torsten]

Since your data look in no way like a Gaussian, I think the fit is the best MATLAB could do out of it.

Answer 1

[Matt J]

If we throw away the data values with y=0, then the remaining data fits a Gaussian quite well. I recommend downloading gaussfitn for the fit.

https://www.mathworks.com/matlabcentral/fileexchange/69116-gaussfitn?s_tid=srchtitle

load Gaussian

keep=(y>0);
x=x(keep)/1e6;
y=y(keep);

[params,resid]=gaussfitn(x,y,{0,1,20,[]},{0,max(y),max(x)},{0,[],[]});

Local minimum possible.

lsqcurvefit stopped because the final change in the sum of squares relative to 
its initial value is less than the value of the function tolerance.

[A,mu,sig2]=deal(params{2:4})%parameters

A = 2.9709e+04

mu = 19.9730

sig2 = 6.9382

fun=@(x)  A*exp( -0.5 * (x-mu).^2./sig2);


%Plot the fit
h=plot(x,fun(x),'-',x(1:20:end),y(1:20:end),'o');

h(1).LineWidth=2;
xlabel x; ylabel y; legend('Fit','Data Samples')

[Torsten]

:-)