Expression-invariant 3D face recognition