视觉纹理的美学情感感知建模研究

项目名称： 视觉纹理的美学情感感知建模研究

项目编号： No.61203364

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 刘建立

作者单位： 江南大学

项目金额： 24万元

中文摘要： 如何借鉴人类情感感知的神经机制以构建图像低层特征与其传递的情感之间的关系模型，进而实现图像情感理解已经成为视觉认知计算研究的一个前沿课题。针对此，本项目将视觉情感感知的神经机制转化为模型内部结构间的逻辑关系，并实现函数解析，构建视觉纹理美学情感感知模型。首先，通过情感实验使被试按照特定的8对情感词汇对视觉纹理进行美学情感量化评价；其次，通过眼动实验获取感兴趣区域，并根据注视时间确定感兴趣区域和非感兴趣区域特征的权值，强化感兴趣区域对情感的影响；然后，采用生物视觉模型对视觉纹理进行特征提取，使低层特征表示更加符合视觉刺激激发美学情感体验的神经机制；最后，将美学情感在感知模型内部结构之间的传递机制转化为合理的函数表达，架起视觉纹理低层特征与其美学情感之间的数学解析之桥。本项目的研究将为基于特定美学情感的视觉纹理设计、融合和挖掘奠定理论基础，推动情感认知和情感计算在人机交互中的应用。

中文关键词： 视觉纹理；美学情感；纹理分析；感知建模；情感计算

英文摘要： How to use the neural mechanism of human emotion perception for reference to model the relationship between low level features and affective semantics for affective understanding of images has been a frontier topic in the field of visual perception computing. Aiming at this problem, we will model the aesthetic perception of visual texture by transforming the logical relationship between the internal structures of the neural circuit of visual emotion perception into mathematical function. Firstly, the quantitative evaluation of the aesthetic emotion for each visual texture image is limited to the 8 specific antonyms in the individual experiment. Secondly, the area of interest of visual texture is captured through eye movement experiment, and the weights of the features extracted from area of interest and non area of interest are decided by the corresponding total fixation time to strengthen the affect of the area of interest on emotion. Then, the low level features are represented based on biological visions model that is used to simulate the simple and complex cells in the primary visual cortex, which makes the feature extraction more suitable for the neural mechanism of aesthetic experience stimulated by visual texture. Finally, a mathematical bridge will be built to span the gap between low level features and

英文关键词： Visual Texture；Aesthetic Emotion；Texture Analysis；Perception Modeling；Affective Computing