Static Hand gesture recognition

《Static Hand gesture recognition》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、StaticHandgesturerecognitionWuyiWengEECSUniversityofMichiganAnnArbor,MI,USA,48105wengwuyi@umich.eduAbstract—Thispaperproposeamethodforhandgesturerecognition.Itdividesthisgoalintotwosubtopics:segmentationandclassification.Foreachpart,thepaperimplementedtwomethods.Forsegmentation,awayb

2、asedonthresholdandanotherwaybasedoneigen-spectralanalysisweretested.Forclassification,acustomclassifierbasedonnumbersoffingersandanotherclassifierbasedonsubspaceclassificationareproposedinthispaper.Keywords—handgesturerecognition;imagesegmentation;eigen-spectralanalysis;eigenvector;s

3、ubspaceclassificationFig1:8gesturesinthedatasetPreprocessingprocessisshownbelow.FirstItakepictures1.INTRODUCTIONusingmycellphone.Tomakeeachpicturehavesimilarsize,IGestureandGesturerecognitionareheavilyencounteredinmakeaboundinthesettingenvironment.Thenmanuallycuthumancomputerinteract

4、ion.Gesturesarethemotionofthethetrainingimage.Afterthis,Iusedthesegmentationbodyorphysicalactionformbytheuserinordertoconveyalgorithmtoextractforeground.Thenusetheforegroundsomemeaningfulinformation.Gesturerecognitionistheimagestoproducetrainingdataset.processbywhichgesturemadebytheu

5、serismadeknowntothesystem.Throughtheuseofcomputervisionormachineeye,thereisgreatemphasisonusinghandgestureasasubstituteofnewinputmodalityinbroadrangeapplications.Inthispaper,statichandgesturerecognitionisdevidedintotwoparts:segmentaionandclassification.Segmentationistosegmenthandpart

6、fromtheinputimage(Assumingthatinputimagehaswell-localizedhandpart).Itriedtwomethodsfor2.2.1Segmentationusingthresholdsegmentation.Oneisanintuitiveway,whichfirstturntheinputimageintoagrayimagethenchooseathresholdtosegmentthehandpartandthebackground.Anothermethodis[1]basedoneigenvector

7、sandeigen-spectralanalysis.YairWeiss[2]andJianboShiproposedthismethodintheirpapers.Forclassification,Ialsotriedtwoways.Oneisanon-trainingintuitivemethod.ItcountsthenumberoffingersinthepictureAsshowninthediagramabove,iftheinputimagehaveaandthenusethenumberforclassification.Anothermeth

8、odisverysimp