Social Network Extraction and Analysis Based on Multimodal Dyadic Interaction 英文参考文献.docVIP

Sensors2012,12,1702-1719;doi:10.3390/s120201702 OPENACCESS sensors ISSN1424-8220 /journal/sensors Article SocialNetworkExtractionandAnalysisBasedonMultimodal DyadicInteraction SergioEscalera 1,2,?,XavierBar′o 1,3 1,2 1,2 1 ,JordiVitri`a ,PetiaRadeva andBogdanRaducanu 1 CentredeVisi′operComputador,CampusUAB,Edi?ciO,Bellaterra,08193 Barcelona,Spain; E-Mails:xbaro@(X.B.);jordi@cvc.uab.es(J.V.);petia@cvc.uab.es(P.R.); bogdan@cvc.uab.es(B.R.) 2 3 DepartmentMatem`aticaAplicadaiAn`alisi,UniversitatdeBarcelona,GranVia585, 08007Barcelona,Spain Estudisd’Inform`atica,Multim`ediaiTelecomunicaci′o,UniversitatObertadeCatalunya,Rambladel Poblenou156,08018Barcelona,Spain ? Authortowhomcorrespondenceshouldbeaddressed;E-Mail:sergio@maia.ub.es; Tel.:+34-93-581-18-28;Fax:+34-93-581-16-70. Received:24December2011;inrevisedform:21January2012/Accepted:22January2012/ Published:7February2012 Abstract: Social interactions are a very important component in people’s lives. Social networkanalysishasbecomeacommontechniqueusedtomodelandquantifytheproperties of social interactions. In this paper, we propose an integrated framework to explore the characteristicsofasocialnetworkextractedfrommultimodaldyadicinteractions. Forour study,weusedasetofvideosbelongingtoNewYorkTimes’BloggingHeadsopinionblog. TheSocialNetworkisrepresentedasanorientedgraph,whosedirectedlinksaredetermined bytheIn?uenceModel. Thelinks’weightsareameasureofthe“in?uence”apersonhas overtheother.ThestatesoftheIn?uenceModelencodeautomaticallyextractedaudio/visual featuresfromourvideosusingstate-of-theartalgorithms. Ourresultsarereportedinterms of accuracy of audio/visual data fusion for speaker segmentation and centrality measures usedtocharacterizetheextractedsocialnetwork. Keywords: social interaction; audio/visual data fusion; in?uence model; social networkanalysis  Sensors2012,12 1703 1.Introduction Social interactions play a very important role in people’s daily lives