GEOMETRIC FACIAL MOTION SYNTHESIS - 3D Face Processing: Modeling, Analysis and Synthesis - page 56

the fact that the subjects tend to trust real face more than synthetic face when
confusion happens. Therefore‚ when the visual information conflicts with the
audio information‚ the synthetic face is less capable of conveying fake emotion
information.
Overall‚ the experimental results show that our real-time speech-driven syn-
thetic talking face successfully affects human emotion perception. The effec-
tiveness of the synthetic face is comparable with that of the real face though it
is slightly weaker.
6. Summary
In this chapter‚ we present face animation methods based on geometric mo-
tion model. We describe three types of animation: text-driven‚ offline speech-
driven and real-time speech-driven face animation. In all of the scenarios‚ our
proposed geometric facial motion model serves as a compact and effective rep-
resentation of visual information. A human perception study is conducted to
demonstrate that the effectiveness of our synthetic face animation in conveying
emotional cues.