Compressible flow simulations of voiced speech using rigid vocal tract geometries acquired by MRI.