Research on emotion extraction based on motion analysis and its application
| Project/Area Number |
16300038
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Media informatics/Database
|
|---|
| Research Institution | Advanced Telecommunications Research Institute International |
|---|
Principal Investigator |
NAEMURA Masahide Advanced Telecommunications Research Institute International, Group Leader, メディア情報科学研究所, 室長 (70374098)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KOBAYASHI Yuichi Advanced Telecommunications Research Institute International, Resercher, メディア情報科学研究所, 研究員 (70395143)
BERRY Rodney Advanced Telecommunications Research Institute International, Resercher, メディア情報科学研究所, 研究員 (90395145)
|
|---|
| Project Period (FY) |
2004 – 2005
|
| Project Status |
Completed (Fiscal Year 2005)
|
| Budget Amount *help |
¥15,100,000 (Direct Cost: ¥15,100,000)
Fiscal Year 2005: ¥5,000,000 (Direct Cost: ¥5,000,000)
Fiscal Year 2004: ¥10,100,000 (Direct Cost: ¥10,100,000)
|
|---|
| Keywords | emotion extraction / human interface / motion analysis / motion capture / action recognition / visualization / CG generation / CG生成 / ヒューマンインターフェース / 感性処理 |
|---|
| Research Abstract |
Firstly, we developed a method for extracting body motion features that can be utilized for acquiring emotion information. Secondly, in parallel to this development, we also developed a method for specifying emotions from motion features based on results of subjective estimation experiments for emotion classification. Finally, after completing these tool algorithms, we built the prototyped 'Kansei' (sensuous) human interactive interface that has a filter function handling media contents with body movements. The followings are summaries of these research issues.
1.A method for extracting body motion features
This method consists of two parts. These are parameterization of body parts and tracking of thus extracted parameters. The tracking results become motion features for analyzing emotion from motion afterwards. Hough transform on skeleton images makes parameters of body parts. Tracking of the hough parameters is carried out using kalman filters in order to decrease occlusion effects.
2.An algorithm for extracting emotions from motion
The analysis algorithm for specifying the relationship between human motion and emotion is proposed. This algorithm is based on performing higher order SVD. The algorithm analyzes human motion data spanning multiple subjects performing the same action and emotion. Given an enough action data with different emotions for the same person, it yields the distinctive motion data specific to the individual.
3.Construction of the 'Kasei' human interactive interface
We built a new type human interactive interface that present a basic method for applying real-time filters to human motion capture data in order to modify the perceived emotional affect of the body movements. Since the interface employ a commercial realtime 3D package callde Touch Designer, various 'kansei' filtering functions will be able to be plugged in.
|
Report
(3 results)
Research Products
(12 results)
