It might be possible, however it would likely require a lot of manual experimentation.
I experimented with an approach that involved subtracting the first identified Gaussian from the summed curve and then identifying the second Gaussian, however the results were not reliable. (Attempting to identify both gaussians from the original summed values without modifying them first yielded the same result even given different initial estimates.)
x = linspace(0, 10, 250);
g1 = 0.7*exp(-(x-3.5).^2);
g2 = exp(-(x-5).^2);
g12 = g1+g2;
figure
plot(x, g1)
hold on
plot(x, g2)
plot(x, g12)
hold off
grid
dgdx = gradient(g12);
[pks,locs] = findpeaks(dgdx)
xvals = x(locs)
figure
plot(x, dgdx)
grid
gmdl = @(b,x) b(1)*exp(-(x-b(2)).^2);
[B11,fv1] = fminsearch(@(b)norm(g12-gmdl(b,x)), [max(g12)*0.5; x(locs(1))*0.5])
[B12,fv2] = fminsearch(@(b)norm(g12-gmdl(B11,x)-gmdl(b,x)), [max(g12); x(locs(2))*1.5])
figure
plot(x, gmdl(B11,x))
hold on
plot(x, gmdl(B12,x))
hold off
grid
[B21,rn1] = lsqcurvefit(gmdl, [max(g12)*0.5; x(locs(1))*0.5], x, g12)
[B22,rn2] = lsqcurvefit(gmdl, [max(g12); x(locs(2))*1.5], x, g12-gmdl(B21,x))
figure
plot(x, gmdl(B21,x))
hold on
plot(x, gmdl(B22,x))
hold off
grid
I doubt that there is a way to autimatically identify both curves, however manualloy experimenting and then subtracting the first identified curve could work..
.
