Synthesizing speech animation by learning compact speech co-articulation models | IEEE Conference Publication | IEEE Xplore