A KANSEI Model to Estimate the Impression Ratings of Spectacle Frames on Various Faces 2005-01-2693

The aim of this study was to develop a KANSEI model that estimates the impression ratings of various combinations of faces and spectacle frames. Such a KANSEI model can be used by a computer system at a retail shop to recommend a suitable frame, using physical characteristics of a consumer's face and spectacle frames. We selected 14 KANSEI words (e.g., “cheerful,” “intelligent") to describe the impression of each face/frame combination. We generated 96 CG images (8 representative face models × 12 frames). These images were shown to 75 female observers (students at a women's university), and impression ratings for 14 KANSEI words were obtained for each image using visual analog scales. Results of ANOVA indicated significant inter-face and inter-frame variance (p<0.01) for 11 KANSEI words. Thus, we have developed a computational model for estimation of impression ratings from physical characteristics of faces and frames.