Emotional Cellular-Based Multi- Class Fuzzy Support Vector Machines on Product's KANSEI Extraction
Emotional Cellular-Based Multi- Class Fuzzy Support Vector Machines on Product's KANSEI Extraction
复制标题
DOI：
--
复制
发表时间：
2012
期刊：
影响因子：
--
通讯作者：
Fuqian Shi;Jiang Xu
中科院分区：
文献类型：
--
作者：
Fuqian Shi;Jiang Xu
通过评价关键形态特征（Critical Form Features， CFF）来提取产品的整体感性图像是一种重要的方法。本文提出了基于情感细胞（EC）模型的多类模糊支持向量机（MF-SVM）来提取产品CFF的感性图像。语意细胞是一种非常特殊的语义细胞，它被定义在二维情感空间上。EC的外壳覆盖了反映其不确定性的每个情感词的边界区域，通常采用密度函数来反映这种不确定性。首先，将特征的乘积映射为N维向量。其次，利用单高斯模型（SGM）和高斯混合模型（GMM）的概率密度函数计算向量空间范数和各元素的模糊隶属度；最后，研究了多类支持向量机的One-Versus-Rest （OVR）算法，用于处理多维感性图像。对于新产品，系统将使用mf - svm指定所有cff。一个手机设计的案例研究给出了证明所提出的方法的有效性。
It is an important methodology to extract product's overall KANSEI images by evaluating Critical Form Features (CFF). In this paper, Multi-class Fuzzy Support Vector Machines (MF-SVM) employing Emotional Cellular (EC) model was presented to extract KANSEI images of product's CFF. EC is a very special kind of semantics cell, which is defined on two-dimensional (Valence-Arousal) emotional space. The shell of EC covers the areas of the boundary of each emotional word that reflects its uncertainty, in common, a density function was employed to reflect this uncertainty. Firstly, product from features was mapped into an N - dimensional vector. Secondly, the norm of vector space and the fuzzy membership of each element are calculated by using probability density function of EC including Single Gaussian Model (SGM) and Gaussian Mixture Model (GMM). Finally, One-Versus-Rest (OVR) for multi- class SVMs was addressed to deal with multi-dimensional KANSEI images. For new products, system will specify all CFFs by using MF-SVMs. A case study of mobile phone design is given to demonstrate the effectiveness of the proposed methodology.