2005 Fiscal Year Final Research Report Summary
Dynamic Modeling of Higher Order Impression of Faces and its Application to Image Engineering
Project/Area Number |
15300076
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Research Field |
Sensitivity informatics/Soft computing
|
Research Institution | Hosei University |
Principal Investigator |
AKAMATSU Shigeru Hosei University, Faculty of Engineering, Professor, 工学部, 教授 (50339503)
|
Co-Investigator(Kenkyū-buntansha) |
GYOBA Jiro Tohoku University, Graduate School of Arts & Letters, Professor, 大学院・文学研究科, 教授 (50142899)
WAKAHARA Toru Hosei University, Faculty of Computer and Onformation Sciences, Professor, 情報科学部, 教授 (40339510)
|
Project Period (FY) |
2003 – 2005
|
Keywords | KANSEI information processing / facial impression / Principal Component Analysis / face recognition / face image synthesis / semantic differential method / factor analysis / 3D face model |
Research Abstract |
The human face conveys more than identity. It also provides important impressions of various personal attributes. Results of subjective evaluation obtained through psychological experiments suggest that a causal relationship exists between these impressions and the physical features of a face. If the relationship between the physical parameters that represent variations in the face's appearance and the impression of the corresponding images perceived by humans could be formalized in a mathematical model, applications could be designed to manipulate a facial image to produce a more favorable impression or to estimate such personal attributes from the facial image. In this study, we mainly directed our attention to defining an effective way to synthesize gracefulness of the face which we selected as an example of higher order impression, by the impression transfer vector method that was proposed in our previous work, and finally confirmed validity of the impression transformation by psyc
… More
hological rating experiments. Face images are physically represented in high-dimensional vectors by separating the shape and texture information, and variations in their appearance are coded into a smaller number of parameters by applying PCA, and thus a morphable model of face images is constructed with these parameters. Based on preliminary psychological experiments, some of the parameters were found sensitive to gracefulness of the faces. Image manipulation method based on the impression transfer vector was then applied using the selected parameters. Psychological evaluation of impression transformation was conducted by applying semantic differential method and factor analysis. The obtained results indicated that the impression transfer vector utilizing the selected components can better modulate gracefulness of the face image corresponding to the psychological rating. A similar approach was applied to 3D facial data, and experiments were conducted to compare human perceptual evaluations of the processing of the 3D face data and that of 2D face images. Less
|
Research Products
(65 results)