欢迎来到天天文库
浏览记录
ID:54924937
大小:522.86 KB
页数:6页
时间:2020-05-04
《基于改进模糊综合评价的电影情感分类-论文.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第41卷第2期计算机科学Vo1.41No.22014年2月ComputerScienceFeb2014基于改进模糊综合评价的电影情感分类林新棋(福建师范大学数学与计算机科学学院福州350007)(福建师范大学网络安全与密码技术福建省高效重点实验室福州350007)摘要为了提高电影情感分类精确度,以模糊数学理论为基础,建立电影底层特征和高层认知情感之间的关系,提出一种基于改进模糊综合评价的电影情感分类新算法。首先,选择了反映情感变化的场景亮度、镜头切换率和颜色能量作为视频场景底层特征,介绍了相应的特征提取方法。其次,引入和改进模糊综合评
2、价模型,并给出特征对情感的模糊隶属函数,建立了单因素评价矩阵。最后,对于不同情感,采用层次分析法确定底层特征之间的相对权重,根据改进的模糊综合评价模型计算出电影场景的情感模糊特征向量,用最大判决值和阈值原则确定待识别场景的情感类型。实验结果表明,所提出的算法能有效地提高电影场景的情感分类精确度。关键词情感模糊特征向量,模糊综合评价,单因素评价矩阵,视频情感内容中图法分类号TP391.4文献标识码AFilmAffectiveClassificationBasedonImprovedFuzzyComprehensiveEvaluation
3、LINXin-qi(DepartmentofMathematicsandComputerScience,FujianNormalUniversity,Fuzhou350007,China)(KeyLaboratoryofNetworkSecurityandCryptography,FujianNormalUniversity,Fuzhou350007,China)AbstractInordertoimprovetheclassificationaccuracyofthefilmsceneemotion,anove1algorithmw
4、asproposedbasedOntheimprovedfuzzycomprehensiveevaluationinthefuzzymathematicstheorybyestablishingtherelationshipbetweenthelow-levelfeaturesandhigh-levelcognitiveemotion.First,thesceneluminance,shotcutratesandcolorener-gYwereselectedasthelow-levelfeaturesfortheirsspecia1
5、characteristicsthatcanbeusedtobetterdistinguishdiferenttypesofhumanemotionalreaction.Further,theextractivemethodswereputforward.Secondly,afterintroducingandimprovingthefuzzycomprehensiveevaluationmodal,fuzzymembershipfunctionswereformedtomeasurethefuzzyre-lationshipbetw
6、eenlow-levelfeaturesandemotion,andthenthesinglefactorevaluationmatrixwasbuilt.Finally,themethodoftheanalytichierarchyprocess(AHP)wasusedtodeterminetherelativeweightmatrixbetweenthefeatures,andtheaffectivefuzzyfeaturevectorwascomputedbytheimprovedfuzzycomprehensiveevalua
7、tionmode1.Andtheaf—fectivetypeofthefilmscenewasobtainedbythemaximizationvalueofthecomponentsoftheaffectivefuzzyfeaturevectorandthresholdatlast.Theexperimentalresultsshowthattheproposedalgorithmcanefectivelyimprovetheac—curacyofthefilmaffectiveclassification.KeywordsAffe
8、ctivefuzzyfeaturevector,Fuzzycomprehensiveevaluation,Singlefactorevaluationmatrix,Videoaffee—tivecontent题,对研究视
此文档下载收益归作者所有