| Project/Area Number |
04650328
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
情報工学
|
|---|
| Research Institution | WASEDEA University |
|---|
Principal Investigator |
HASHIMOTO Shuji WASEDA Univ., School of Science and Engineering, Professor, 理工学部, 教授 (60063806)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MATSUSHIMA Toshiaki TOHO Univ., Faculty of Science, Assistant Professor, 理学部, 講師 (30190458)
|
|---|
| Project Period (FY) |
1992 – 1993
|
| Project Status |
Completed (Fiscal Year 1993)
|
| Budget Amount *help |
¥2,000,000 (Direct Cost: ¥2,000,000)
Fiscal Year 1993: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 1992: ¥1,200,000 (Direct Cost: ¥1,200,000)
|
|---|
| Keywords | Face Image / Image Retrieval / Ambiguous Feature / Feature Extraction / Human Model / Human Interface / あいまい語 / 感性 / 表情 / スプリングフレームモデル / あいまい検索 / データベース / ニューラルネット |
|---|
| Research Abstract |
In order to construct the face image retrieval system based on the ambiguous feature description, an automated facial feature extraction and a facial image retrieval using a neural network are investigated.
Two face images are used to extract features. One image is eyes-close scene, and the other is eye-open scene.The frame difference and the regional pattern matching technique are used for the eye detection, then the other parts of the face such as a nose and a mouth are detected. Even for the face images which are slightly shifted form each other, the eyes and the other parts of faces are successfully detected.
In the research on face image retrieval, the geometrical features of faces are measured from the face images, and the impression of faces are evaluated by the questionnaire research. Using these data, we trained the neural network to associate geo-features with the impression for the faces. The neural network is successfully trained for the impression of physical features such as position of eyes, size of mouth, and roundness of the face. Although output for some faces are different from the answers of the questionnaire for the emotional words such as "active" and "intellectual, " we can get the promising results to build the facial image retrieval system using the ambiguous feature description words.
|
