Hi, I would like to fit a signal pulse with a rectangular pulse function. this is my script, a rectangular pulse signal should be fitted with the same function:
clear all
xdata=[1:1000];
a=100;
b=30;
c=12;
ydata=squarepulse(xdata,a,b,c);
plot(xdata,ydata)
par0=[ 100
2
7];
lb=[xdata(1)
1
5];
ub=[xdata(end)
50
20];
options = optimoptions('lsqcurvefit','tolx',1e-10,'tolfun',1e-10,'maxfunevals',5000000)
parOut=lsqcurvefit(@squarefun,par0,xdata,ydata,lb,ub,options)
yfit=squarefun(parOut,xdata);
plot(xdata,ydata,xdata,yfit,'r')
and this is the script for build the rectangular pulse signal and used to fit:
function y = squarefun(par,xx)
y=xx*0;
for i=1:length(xx)
if xx(i)<=par(1)
y(i)=heaviside(xx(i)-par(1)+par(2))*par(3);
else
y(i)=heaviside(-xx(i)+par(1)+par(2))*par(3);
end
end
end
the problem is that the parameters par(1) and par(2) (x position of the signal and its width) do not change, only the par(3) is correctly found. If I substitute the rectangular function with a gaussian the minimisation works good. Thank you regards Andrea
