Facial Expression Generation in 3D space

Abstract

Facial expression is one of the most powerful resources for people to coordinate conversation and communicate emotions and other mental, social, and physiological cues. Because humans have high sensitivity for facial expression, many areas such as game, movie, avatars and social robots tries to mimic these facial expressions in their artificial human models to increase realistic features. This research focuses on facial expression generation in 3D space and proposes a novel methodology to generate facial expressions for mesh models in 3D space. 3D point cloud of a neutral human face is taken as an input to the system and preprocessing techniques are applied on face mesh generated using a mesh construction algorithm. Finally, with relevant features extracted from these meshes, a facial expression generation techniques, ratio mapping technique and difference feature vector adding mechanism are applied to generate happy, fear, sad and disgust expressions. Finally, process and outcome evaluations have been conducted to gauge the realism and perceptibility of artificially expression generated faces. The results of these experiments prove that generated expressions are real enough to identify, however further improvements needs to be done.