多摄像头视觉系统在社会互动分析中应用研究.pdfVIP

BRIDGINGCOMPUTERVISIONANDSOCIALSCIENCE:AMULTI-CAMERAVISION

SYSTEMFORSOCIALINTERACTIONTRAININGANALYSIS

JixuChenMing-ChingChangTai-PengTianTingYuPeterTu

GEGlobalResearch,NiskayunaNYUSA

takeceduringtask-orientedscenarios.Socialawareness

andtheabilitytounderstandandmanagesocialsignalsareat

Weinvestietheuseofavision-basedsystemcapableof

thecoreofsocialintelligence.Proficiencyinsuchsocialskills

estimatingsocialstatessuchasrapportandhostility.Westudy

isoftenthekeyforsuccessinpracticalscenariossuchasne-

thecorrelationweeninterpretationsautomaticallygenerat-

gotiation,interviewing,marketing,etc.Ouraimistoimprove

edbyoursystemandthosereportedbysocialscientists.Our

trainingbyconsideringsocialsignalsrangingfrombodyori-

multi-cameravisionsystemcollectsvisualcuesincludinglo-

entation,movement,gaze,expression,andposeasdepicted

cation(proximity),motion,pose,gaze,andfacialexpressions

inFig.2(a).Weproposetof