Vol.1 No.1 2008

49/76

Research paper : Technologies for the design and retail service of well-fitting eyeglass frames (M. Mochimaru, et al.)−46−Synthesiology - English edition Vol.1 No.1 (2008) 1. Methods for improving fitQuestion (Motoyuki Akamatsu)Authors listed 1) Population Grouping, 2) Adjustable Products, 3) Finding Well-fitting Product, 4) Mass customization, and 5) Traditional customization as methods for improving the fit. I believe that Method 3) is for improving the fit obtained by Method 1), rather than a method located between Methods 2) and 4). Therefore, I believe that Method 3) should be considered as a secondary method to make up for the shortcomings of Methods 1) and 2). Because both methods depend on the subjective judgment of the user, true conformity cannot be attained. In addition, Method 3) is the basis of Method 4).Answer (Masaaki Mochimaru)Thank you for your comments. As you stated, 3) Finding Well-fitting Product does support the shortcomings of 1) Population Grouping and 2) Adjustable Products. Therefore, we revised the text as follows: After describing 1) Population Grouping and 2) Adjustable Products, we state that good fit cannot be obtained if the selection depends on user preference, citing our previous study as an example. We then state that Method 3) Finding Well-fitting Products overcomes these issues.Your last comment indicates that the relationship between the characteristics of users and characteristics of products will be accumulated by providing services for “Finding Well-fitting Products”, and such information will be the basis for “Mass Customization”. We agree with this opinion. As our original manuscript did not mention this point, we now discuss it in the Discussion.2. Grouping subjects and the validity of evaluationQuestion (Motoyuki Akamatsu)Please explain the grouping of subjects into 4 groups. Did you evaluate the validity of each evaluation study with regard to which group the 38 subjects belonged and the differences between each subject and average shape in the 4 groups? Answer (Masaaki Mochimaru)(1) Grouping subjectsAs the shape variation followed the normal distribution, there was no scientific way to divide subjects into a limited number of groups. Theoretically, the more groups, the better the fit. Therefore, in this study, the collaborating company examined the maximum possible number of groups from the viewpoint of profitability in manufacturing and distribution, and decided on 4 groups. We then had to determine how to cover the variation within the limit of 4 groups. In this study, we divided the scores for the 1st scale, which explained the largest variance, into 3, and the group with near average scores for the 1st scale was divided into 2. As shown in Figure 5, Groups A and D cover larger areas than Groups B and C. This is because by setting different ranges of scores for the 1st and the 2nd scales between the 4 groups, we attempted to minimize the difference in shipments between the 4 groups. In other words, numbers of people who would be assigned to the 4 groups are about the same in number. Such reasoning is described in reference [4], and we have added an overview to the manuscript.(2) Evaluation studyWe recruited subjects so that the numbers of subjects assigned for the 4 groups did not differ substantially. Because 3D shape was not measured for the 38 subjects, the shape difference between each subject and the average for each group is unknown. However, the difference was within 3 mm for head dimensions. This is described in the reference cited, but we now mention it in the text.3. Head dimensions measuredQuestion (Motoyuki Akamatsu)Section 3.3 describes “using 15 face dimensions”. How did you decide the 15 face dimensions?Answer (Masaaki Mochimaru)Details are described in the reference cited. We measured about 50 items for all subjects. The 15 measurements were adopted for regression equations to estimate the scores for the 1st and the 2nd scales, using stepwise regression analysis. We intentionally selected measurement items that can be measured easily and reliably, and performed the stepwise regression analysis several times. We did not describe the details in the text.Our initial plan was to use these 15 measurements rather than 3-D shape for selecting eyeglass frames (see references), but this was not accepted by the collaborating company, as retail staff do not possess the necessary skills in anthropometry. Therefore, we decided to develop a low-cost head measurement system. 4. Future worksQuestion (Motoyuki Akamatsu)Please include a discussion of the future direction of the project, areas that need to be improved, and technical breakthroughs yet to be realized.Answer (Masaaki Mochimaru)Of course, there are some areas that require further work. For example, the effects of hairstyle on style recommendation have not been investigated. It is clear that there is an effect based on the results of preliminary experiments, but we currently lack the computer graphics technology to naturally change hairstyles. Future works include designing parts (nose pad and bow) and finite element analysis for estimating the pressure distribution between these parts and the face, as well as the sensation of touch. Another incomplete aspect is the technology for utilizing the information on product selection and satisfaction. Without this technology, simultaneous data accumulation, updating size variation designs and Kansei evaluation are impossible; this system is merely a tool for sales support. This will be studied further after systems such as that presented in this paper are practically used in shops.5. Other applications of this technologyQuestion (Motoyuki Akamatsu)Similar systems may be used for other products, such as clothes and shoes. Will the technologies for other products be the same as with this system, or is there something specific to eyeglass frames?Answer (Masaaki Mochimaru)We believe that this approach for measuring the human body and selecting the best fitting products from size variations has general applications. Such methods are already used for running shoes in shops. However, applications for selecting clothes are not very popular, as few consumers would be willing to undress for a 3-D scan. We believe that improved technology for 3-D measurement for clothes, or other businesses related to body scanning may solve this problem.For recommendations based on Kansei modeling, style design and fit are more independent elements for eyeglass frames and running shoes, while they are closely related with shoes and clothes. Therefore, the present system that recommends size and style independently may not be applicable to shoes or clothes. However, recommendation systems for those products are possible and should be studied in the future.

※このページを正しく表示するにはFlashPlayer9以上が必要です