面部表情识别算法研究

《面部表情识别算法研究》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、AbstractNowadaysmodernscienceandtechnologyarerapidlydeveloping，asaverychallengingCROSS-subject，facialexpressionrecognitioninvolvesimageprocessing，patternrecognition，artificialintelligenceandaffectivecomputing，andmanyotherdisciplinesofknowledgeandtheory．ThefieldofexpressionrecognitionresearchCanpromo

2、tethecomputertotheintelligentandhumaneandbetterdevelopment，hasaveryimportantpracticalsignificanceandwillhaveconsiderablesocialandeconomicbenefits，themarketprospectsarebright．Thethesishasstudiedinseriesalgorithmsofexpressionfeaturesextractionandclassificationdeeply,andproposedanexpressionrecognitioni

3、mprovedalgorithmbasedon2DGaborwavelettransformandsupportingvectormachine，italsohassolvedthekeyquestionsoffeatureregionspositioning，eigenvectorextraction，classifierdesignandetc，asresearchobjectofJAFFEdatabaseexpressionimage，crossedaseriousofresearchtoproveitsvalidityandfeasibility．Thespecificworkandt

4、hemaincontributionofthepaperasfollows：(1)PretreatmentInordertoeffectivelyeliminatetheinterferencefactorsincludingnoisewiththeunrelatedexpressionrecognition，itneedstopreprocesstheoriginalimagefirstly．Paperproposesanewfusionalgorithmforfastandaccurateeyelocation；thenuseshistogramequalizationtocompensa

5、tetheilluminationofexpressionofimage；andaccordingtotheanglebetweentheconnectionlineofthepointsofcoordinatesaroundtheeyesandthehorizontalaxis，rotates，cropsandscalesofgeometricnormalization．(2)Expressionfeatureextractionbasedon2DGaborwavelettransformTheheartoftheexpressionrecognitionisfeatureextractio

6、n．Thedifficultyincomputerthefaceisaflexiblebody，itisverydifficultforaccurateidentifyingfacialexpressions，becausemodeling．Theslightchangesinfacialorganlocation，correspondinglythefacialexpressionchangeswilloccur,howtochoosethecharacteristicsofthedecisionbecomethekeytoidentificationaccuracy．Paperfocuse

7、sontheimportanceoffeatureextractionalgorithm，analysesdentally,finallyselects2DGaborwavelettransformwhichishigh-limitedshieldingilluminationandtheeffectsofdifferencesinindividualcharacteristicsasfeatur