Psychological and Computational Perspectives on Recognition of Moving Faces
Natalie Butcher,Darren Cosker,Nicholas Costen,Hui Fang,Harry J. Griffin,Alan M. Johnson,Eva G. Krumhuber,Karen Lander,David Marshall,Paul L. Rosin
2009-01-01
Abstract:Psychological and computational perspectives on recognition of moving faces Symposium Moderator: Nicholas Costen Symposium Presenters: Alan Johnston and Harry Griffin Nicholas Costen and Hui Fang Darren Cosker University College London Manchester Metropolitan University The University of Bath Department of Psychology, Department of Computer Science, Department of Computing and Mathematics, London, WC1E 6BT UK Manchester M1 5GD UK Bath BA2 7AY UK (a.johnston@ucl.ac.uk) (n.costen@mmu.ac.uk) (D.P.Cosker@cs.bath.ac.uk) Eva Krumhuber Swiss Center for Affective Sciences, University of Geneva 1205 Geneva, CH (eva.Krumhuber@unige.ch) Karen Lander and Natalie Butcher Paul Rosin and David Marshall The University of Manchester School of Psychological Sciences, Manchester M13 9PL UK (karen.lander@manchester.ac.uk) Cardiff University School of Computer Science, Cardiff, CF24 3AA UK (Paul.Rosin@cs.cf.ac.uk) Overview of presentations Keywords: Perception, computation, face recognition, move- ment. Darren Cosker and Eva Krumhuber The modelling of faces is an area of intense research, and facial models - developed by computer graphics and vision researchers – have a wide range of applications. Arguably the most publicly visible result of facial research appears in movies and video games. These often astounding feats are the result of intensive labour and artistic license. Facial mod- elling also has major application areas in psychology and neu- roscience. A powerful discovery from the application of such models in these fields is how the perception of a facial ani- mation can change given even a small subtle variation in ex- pression timing and duration (Krumhuber et al., 2007). An interesting relationship then exists between psychology re- search and the development of human-realistic facial mod- els for the entertainment industry, i.e. if we are to achieve human-realism we also need to understand how faces are per- ceived. Under this theme, we will discuss the modelling of faces from 2D images and dynamic 3D facial data (i.e. 3D captured at 60 frames per second), parameterisation of such data for creation of facial animations, and their application (Cosker, Borkett, Marshall, & Rosin, 2008). We will also re- port on some example recent collaborative studies that high- light the power of manipulating photorealistic facial expres- sions to create different overall impressions of a person and influence decision-making. People • Nicholas Costen is a Senior Lecturer in Computing at Manchester Metropolitan University. He holds degrees in both psychology and computer science, and has published extensively on the computational modeling of the percep- tion of static and dynamic faces. • Darren Cosker is a Royal Academy of Engineering / EP- SRC Research Fellow at the University of Bath. His inter- ests centre on human motion analysis and synthesis. • Alan Johnston is Professor of Psychology at University College London. His group studies the perceptual mech- anisms that underlie our capacity to see movement, depth and form. • Karen Lander is a Senior Lecturer in Psychology at the University of Manchester. Her interests focus on the im- portance of face animation in the recognition of familiar faces. • Paul Rosin is a Reader in Computing at Cardiff Univer- sity. His interests include low-level computer vision, talk- ing heads and 3D mesh processing. Nicholas Costen and Hui Fang Introduction Computational statistical models, derived from moving im- ages, may be used in the face recognition process. As a coun- terpart to psychological experimental results showing a sig- nificant beneficial effect of facial non-rigid movement, two features obtained from face sequences, the central tendency and type of movement variation, are associated to improve face verification compared with single static images. By us- ing a progressive optic-flow based correspondence algorithm (Fang & Costen, 2008), the correspondences across the se- Our knowledge of object, scene and face recognition is largely based on studies that have used static stimuli. How- ever, in reality, motion and change are fundamental aspects of our world. Do our perceptual/cognitive systems make use of these real world dynamics? The purpose of this sympo- sium is to showcase a number of research projects from both Psychology and Computer Science that explore the impact of motion on our perception and representation of faces.