You can first get an estimate of the zeros (if any) in your interval-of-interest by calculating it in that interval, then multiplying the function by circshift of the function to detect any zero-crossings that might be present. After that, use those estimates as your initial guesses for fzero
To illustrate:
x = linspace(0,50,200);
y = @(x) sin(x);
zx = x(y(x).*circshift(y(x),[0 -1]) <= 0); % Estimate zero crossings
zx = zx(1:end-1); % Eliminate any due to ‘wrap-around’ effect
for k1 = 1:length(zx)
fz(k1) = fzero(y, zx(k1));
end
