EEG Based Analysis of the Perception of Computer-Generated Faces

We investigate the perception of different categories of human faces with varying degrees of realism. Computer generated humans are an integral part of video games, movies and immersive environments. However, generating perceptually believable human faces remains a significant challenge in the field of Computer Graphics. Part of the problem lies in understanding exactly how computer-generated human faces are perceived by the human visual system. Currently, there exists no agreed upon standard for defining and measuring facial realism within the graphics community. We employ techniques from cognitive neuroscience to explore the perception of computer-generated faces in an attempt to measure facial realism. An Electroencephalograph (EEG) is used to investigate the difference in perception of still images of real, altered real, computer-generated and line drawings of human faces. Our results indicate distinct differences in the neural response to real, altered real and line drawings of faces. Interestingly, we also observe a difference between real and computer-generated faces. Our results suggest the possibility of using neural correlates for defining and quantifying facial realism within a graphics context.