Statistical synthesis of facial expressions for the portrayal of emotion

This paper presents a novel technique for the generation of ‘video textures’ to display human emotion. This is achieved by a method which uses existing video footage to synthesise new sequences of coherent facial expression and head motions. An ‘expression space’ which is defined by sets of emotion models is constructed using principal components analysis (PCA) and the application of an auto-regressive process (ARP). Using this expression space a tool has been developed which enables this space to be navigated, and to view new video sequences generated from it. This tool provides an intuitive interface, which allows a nonspecialist user easy traversal of the highly dimensional expression space and enables them to select an emotion and automatically generate new facial expression sequences. CR Categories: I.3.m [Computer Graphics]: Miscellaneous— Emotion Modelling; I.3.4 [Image Processing and Computer Vision]: Image Representation—Statistical.

• Type: Conference
• Year: 2004
• Where: GRAPHITE
• Authors: Lisa Gralewski, Neill W. Campbell, Barry T. Thomas, Colin J. Dalton, David P. Gibson
• Keywords: Coherent Facial Expression, Expression Space, Facial Expression, GRAPHITE 2004