Abstract
In this paper, the methods for modeling facial expression and emotion arc proposed. This Emotion Model, called 3-D Emotion Space can represent both human and computer emotion conditions appearing on the face as a coordinate in the 3-D Space.
For the construction of this 3-D Emotion Space, 5-laycr neural network which is superior in non-linear mapping performance is applied. After the network training with backpropagalion to realize Identity Mapping, both mapping from facial expression parameters to the 3-D Emotion Space and inverse mapping from the Emotion Space to the expression parameters were realized.
As a result a system which can analyze and synthesize the facial expression were constructed simultaneously.
Moreover, this inverse mapping to the facial expression is evaluated by the subjective evaluation using the synthesized expressions as test images. This evaluation result proved the high performance to describe natural facial expression and emotion condition with this model.
This is a preview of subscription content, log in via an institution.
Preview
Unable to display preview. Download preview PDF.
References
P.Ekman and W.Friesen:“Facial Action Coding System”, Consulting Psychologists Press, 1977.
Y.Katayama and K.Ohyama:“Some Characteristics of Self Organizing Back Propagation Neural Networks”, Spring National Convention Record, IEICE-J, SD-1-14, p.7–309, 1989.
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 1995 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Kawakami, F., Ohkura, M., Yamada, H., Harashima, H., Morishima, S. (1995). 3-D emotion space for interactive communication. In: Chin, R.T., Ip, H.H.S., Naiman, A.C., Pong, TC. (eds) Image Analysis Applications and Computer Graphics. ICSC 1995. Lecture Notes in Computer Science, vol 1024. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-60697-1_136
Download citation
DOI: https://doi.org/10.1007/3-540-60697-1_136
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-60697-0
Online ISBN: 978-3-540-49298-6
eBook Packages: Springer Book Archive