The following matrix is not hermitian.
A = [0 2; -2 0]
It has an eigenvalue decomposition with a unitary matrix of eigenvectors, though, so it is normal.
[V,D] = eig(A);
norm( V'*V - eye(2) )
The eigenvalues are pure imaginary.
lambda = diag(D)
The singular values are the complex magnitudes of the eigenvalues.
svd(A)