面向视觉感知的图像情感识别及其在推荐系统中的应用  

Emotional Recognition of Visual-perception Oriented Images and Its Application in the Recommendation System

作　　者：陈芬[1] 何源 汤丽萍 Chen Fen;He Yuan;Tang Liping(School of Economics and Management,Nanjing University of Science and Technology,Nanjing 210094)

机构地区：[1]南京理工大学经济管理学院,南京210094

出　　处：《情报学报》2019年第4期420-431,共12页Journal of the China Society for Scientific and Technical Information

摘　　要：视觉信息是人们获取外界信息的重要来源。作为视觉信息的主要表现形式之一,图像受到了广泛关注。本文首先针对颜色直方图忽略空间信息的问题,基于图像分块思想,利用图像区域不同、引起的关注程度不同的原理,引入Itti视觉注意模型,进行图像显著图的提取,基于显著图计算各分块的加权直方图。其次,根据视觉感知理论,提取多种图像情感信息特征,结合低层的颜色、纹理和形状特征以及高层的面部表情特征,生成复合的图像情感特征描述向量。最后,将本文提出的情感识别模型用于基于情感的电影推荐,结合电影海报及剧情简介文本,进行基于图文结合的电影情感识别,为用户推荐符合其情感需求的电影。本文通过融合基于心理认知的特征及面部表情等诸多特征,实现了更加完善的视觉情感描述,在一定程度上缩小了"语义鸿沟"。Visual information is one of the most important sources of external information.Images are a major form of visual information.In this paper,the authors first optimize the algorithm of color histogram based on image segmentation,introduce the Itti visual attention model,extract the image saliency map according to the principle that different image regions arouse different degrees of attention,and calculate the weighted histogram based on the saliency map.Secondly,different types of visual emotional features are extracted,combining low-level color,texture,and shape,as well as high-level facial emotion features to generate a composite image sentiment feature description vector.Finally,the emotion recognition model is used to make emotion-based film recommendations,combined with movie posters and synopsis texts,and based on the combination of graphic and textual emotion recognition,to meet users’emotional needs.In conclusion,this paper proposes a framework of image emotion recognition that combines visual-perception oriented features and facial emotions.This framework is efficient and exhibits good performance.To a certain extent,the framework proposed in this paper narrows the“semantic gap”in this area of research.

关 键 词：情感分类 特征提取 视觉感知 情感语义 SVM算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]