The output of histfit is a vector of handles, one of which is a line object containing the x/y values for the curve. You could fit those line values to a gaussian function to get the fit parameters. Update: if you're using "kernel", that's a nonparametric kernel-smoothing distribution where the density is evaluated at 100 equally spaced points that cover the range of the data so it has no parameters to fit.
Alternatively, you could use fitdist instead of histfit to fit the distribution and return the fit parameters. These two methods are compared in this answer.
y = randn(1,1000)+6.2*9.8;
h = histfit(y);
lineobj = findobj(h,'type','line');
gx = lineobj.XData;
gy = lineobj.YData;
f = fit(lineobj.XData',lineobj.YData','gauss1')
% Now plot the fitted line again see that it's the same
hold on
plot(f,'g-')
