Figure 13
From: Lip-Synching Using Speaker-Specific Articulation, Shape and Appearance Models
Results of the MOS test. Three groups can be distinguished: (a) the trajectory formation systems PHMM and TDA are not distinguished from the resynthesis of original movements; (b) the audio-synchronous HMM trajectory formation system is then rated best, and (c) the concatenation system with multi-represented audiovisual diphones is rated significantly worse than all others.