Toward the Automatic Generation of Cued Speech