2006 Fiscal Year Final Research Report Summary
Affective postures in interactive systems : Application to e-learning
| Project/Area Number |
17500147
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Sensitivity informatics/Soft computing
|
|---|
| Research Institution | University of Aizu |
|---|
Principal Investigator |
GUPTA Renu University of Aizu, Dept. of Computer Software, Assistant Professor, コンピュータ理工学部, 講師 (90381417)
|
|---|
| Project Period (FY) |
2005 – 2006
|
| Keywords | Emotion recognition / Affective posture modeling / Affective computing / User modeling / Human-Computer Interaction / Cross-cultural studies / Adaptive models / Non-verbal communication |
|---|
| Research Abstract |
The aim of the project was to build a system to recognize emotions from posture in e-learning environments. The objectives were reached following two different lines :
a) Building an understanding on how people recognize emotion from posture.
1. Video of students performing on-line learning, conference with their tutor and gaming were recorded. The video were analyzed in order to identify body language displays that could be used to infer the affective state of the participants. The results showed more changes in the postural display than in the facial expressions, supporting our hypothesis on the relevance of body postures as an affective communication channel.
2. Video recording and body motion capturing of participants interacting with full-body controller device were performed. A coding system was also proposed to facilitate the analysis and quantification of body behavior. Using statistical analysis, it was shown that controllers which require the involvement of the whole body (e.g.,
… More
guitar shaped controller, Wii Nintendo controller, bongo-shaped controller) to control a computer game favor the use of body to engage at emotional, social and role-playing levels.
2. Gender studies have been carried out to identify differences in the way men and women recognize emotions from the posture of 3D avatars. Our results indicated that females tend to be significantly faster in recognizing affect from body posture. Further FMRI imaging testing are carrying out to identify possibly different brain path ways that could justify this difference in reaction time.
3. 111 avatars were used to understand if people can discriminate the affect conveyed by avatar's static postures using affective dimension scales : arousal, valence, potency and avoidance. Using statistic methods, it was shown that high agreement exists between the participants : 80%ca for the first 3 dimensions and 70% ca for the avoidance dimension.
b) Designing and implementing a computer-based system for affective posture recognition.
Two versions of the system were implemented and tested. The first version is used to process postures collected using a motion capture system. The second one is a computer-vision-based system. The two systems were tested over different set of emotion categories and emotion dimensions obtaining recognition rates comparable to humans.
The results brought to the publication of one international journal paper, 3 international conference papers, 2 workshop papers and 1 doctoral consortium short paper. The papers are attached to this document. Less
|
Research Products
(11 results)
