A Study on Kansei Imagery of Game Character Design Using Support Vector Machine

• Main Authors: Feng-Yueh Kao, 高豐岳
• Other Authors: Ming-Yuhe Chang
• Format: Others
• Language: zh-TW
• Published: 99
• Online Access: http://ndltd.ncl.edu.tw/handle/04633297849627554437

碩士 === 南台科技大學 === 多媒體與電腦娛樂科學系 === 99 === In this study, proposes a Kansei engineering approach to research a case study of the most popular in Taiwan market in the massively multiplayer online role-playing game "World of Warcraft ", to analyze the role of modeling and art design. First, collected samples of the game's monster role. According to type Analysis of samples from deconstruct shape. Through focus groups survey discussion, the selection of key design modeling. The definition of each feature point and draw a monster form features of the role of the table. The second step for a monster character design Kansei survey, through questionnaires to samples of all subjects were evaluated. The questionnaire data through factor analysis (FA) extraction of representative adjectives of the response scales. The representation of adjectives as category labels. For each sample to be classified under the category labels. Using SVM (Support Vector Machine, SVM) classification model established. Questionnaire results of factor analysis. The selection of the five adjectives, as: adj11 heavy, adj3 nausea, adj6 ferocious, adj5 vigorous, adj9 huge. The five adjectives that best represents the player the feeling heart of the monster samples. Through focus group discussions, and according to the literature monster character design and style analysis. Making a monster feature design system for game art and design staff at the reference. In this study, Gaussian kernel and polynomial kernel function of support vector machine classification model of emotional response, and using the results of confusion matrix analysis and forecasting. We find that the Gaussian core function of the average accuracy rate (93.9%) than polynomial kernel function ( 71.4%).