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.
| Original language | English |
|---|---|
| Title of host publication | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
| Publisher | Springer Verlag |
| Pages | 471-478 |
| Number of pages | 8 |
| Volume | 1024 |
| ISBN (Print) | 9783540606970 |
| DOIs | |
| Publication status | Published - 1995 |
| Externally published | Yes |
| Event | 3rd International Computer Science Conference, ICSC 1995 - Hong Kong, Hong Kong Duration: 1995 Dec 11 → 1995 Dec 13 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 1024 |
| ISSN (Print) | 03029743 |
| ISSN (Electronic) | 16113349 |
Other
| Other | 3rd International Computer Science Conference, ICSC 1995 |
|---|---|
| Country | Hong Kong |
| City | Hong Kong |
| Period | 95/12/11 → 95/12/13 |
Fingerprint
ASJC Scopus subject areas
- Computer Science(all)
- Theoretical Computer Science
Cite this
3-D emotion space for interactive communication. / Kawakam, Fumio; Ohkura, Motohiro; Yamada, Hiroshi; Harashima, Hiroshi; Morishima, Shigeo.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1024 Springer Verlag, 1995. p. 471-478 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1024).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - 3-D emotion space for interactive communication
AU - Kawakam, Fumio
AU - Ohkura, Motohiro
AU - Yamada, Hiroshi
AU - Harashima, Hiroshi
AU - Morishima, Shigeo
PY - 1995
Y1 - 1995
N2 - 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.
AB - 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.
UR - http://www.scopus.com/inward/record.url?scp=84957654250&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84957654250&partnerID=8YFLogxK
U2 - 10.1007/3-540-60697-1_136
DO - 10.1007/3-540-60697-1_136
M3 - Conference contribution
AN - SCOPUS:84957654250
SN - 9783540606970
VL - 1024
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 471
EP - 478
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PB - Springer Verlag
ER -