D08--左右手运动想象脑电采集和特征提取方法初探

华中科技大学硕士学位论文左右手运动想象脑电采集和特征提取方法初探姓名：叶坤申请学位级别：硕士专业：生物医学工程指导教师：骆清铭20080527I*(BCI)BCIBCIPresentationVisualGuide83%C3C410Hz10HzC3C4Morlet*IIAbstract*TheadvancementofBrain-Computerinterface(BCI)studybringshopesofrehabilitationtothehandicapped.AndtheBCIbasedonelectroencephalographhasbeenahotspotfortheBCIstudy,becauseofitssimplestructure,safetyandconvenienceofsignalacquisitionandportage.Inthisthesis,thecharactersoftheElectroencephalogram(EEG)signal,theconceptoftheBCI,thedomesticandinternationalresearchstatusofBCIwereintroducedfirst.Thentheparadigmofourownwasgropedforaftertheparadigmofthemovementimaginationstudiesofthegroupsoverseabeingstudied,andtheprogram‘VisualGuide’forvisualcuesofthehandsmovementimaginationwasmade.Testeesweretrainedwiththisprogramandthesignalwithenoughdistinctionwasacquiredfinallywiththesuccessfulrateof83%.Forthegoalthatfeaturesofthesignalwouldbeextractedeffectively,thesignalacquiredpreviouslywaspretreated,andthenthesignalwasanalyzedinfrequencydomain,timedomainandtime-frequencydomain.Throughfrequencydomainanalysis,thesignalgotfromC3andC4hadgreatestdistinctionaboutthe10Hz,andthesignalof10Hzhadcorrelationwiththecuesoflefthandandrighthandmovementimaginationthroughtimedomainanalysis.Andthetime-frequencyanalysishadfurthersustainedtheconclusion.Withthefeaturesthatwegotfromthetime-frequencydomainanalysis,theoriginalsignalwasdecomposedandthesignalwiththefeatureswasextractedsuccessfullythroughtheMorletwavelettransformation.Keywords:EEGMu-rhythmTime-frequencyanalysisFeatureextraction.*SupportedbyProgramforChangjiangScholarsandInnovativeResearchTeaminUniversity2008529□“”20085292008529111.1Brodmann(1909)1.11.1[1]Fig.1.1FunctionareasonhumancortexBrodmann6Brodmann4[1]210Brodmann3121.21875·1903192940[2]193419502030507031998RossDunSeath[3-4]ElectroencephalogramEEGElectrocorticogram,ECOG[5]1.2Fig.1.2MethodsofEEGacquisition1.2[6]10mV100V10~50µV0.5~100Hz4(evokedpotential,EP)(evokedresponse,ER)[7]1.313-40Hz8-13Hz4-8Hz0.5-4Hz1.3Fig.1.3SpontaneousEEG(1)(2)(3)(4)[7]1.3,[8]Brain-Computerinterface,BCIHuman-Computerinterface,HCI[9]BCIfMRPETfNIRS[10]51.4BCIFig.1.4Blockdiagramofenvironmentcontroldevicebasedonthenon-invasiveBCIBCIBCIBCI[9],BCI[6],1.4,[11],6BCICompetitionBCIPatternRecognition2006Mason-BCI:(transducer)(demonstrationsystem)(assistivedevice)[12](1)VisualEvokedpotential(2)Eventrelatedpotential,ERP,P300(3)ERS/ERD;(4)α(5)α[13]BCIBCIBCIBCI[14]BCI[15-16]200510126358.X00815134.299122161.3200520136506.1200510126359.4200510126360.7P300[17]Alpha[1819][20][2122][2325]BCI1988IllinoisFarwellDonchinP300[26]6×6“”P300BCI[27]a.TubingenNielsBirbaumerDornhegeGSCPSlowCorticalPotential,BCIb.Wadsworth7Mu8-12HzBeta18-26Hz,c.GrazPfurtschellerERD/ERSBCIERS,ERD1.4BCImuβFarwellandDonchin1988(somatosensorymotorcortex)muβ[28],muβTubingenNielsBirbaumerDornhegeGSCPSlowCorticalPotentialWadsworthMu8-12HzBeta18-26HzGraz[27]ERD,ERS1.520066.34%1.5241229.07%[29]10%6.34%820068296,6.34%1.52006[29]Fig.1.5Diagramofnation-wideproportionofthehandicapped,2006BCIBCI2003,,9Morlet1022.1[30]2.1.1(ERP)(checklist)ERP12EEG690%12421.01134418~3556242.1.264Neuroscan32100KHz/32,64,128,256DC+/-200mVAC+/-995uVDC24nV/bit,AC3nV/bitAg-Agcl2.1ab2.1(a)(b).Fig.2.1ProcedureofEEGacquisition(a)testee;(b)dataacquisition(2.1a)(2.1b)12Neuroscan321m3°[31]Neuroscan20~25[32]2.1.3Neuroscan64ERPAg-Agcl2.2Brodmann[1]2.3[1]Fig.2.2Brodmann`sareasFig.2.3MappingofbodypartsonprecentralgyrusBrodmann42.2,2.3Neuroscan3210-202.4(M1,M2)C3CzC42.5C3C413C3C4GNDCzC3C4M1M22.410-202.5Fig.2.4International10-20SystemFig.2.5Placementofelectrodes0-20KΩ5KΩ0.1-100HzFs=1000Hz50Hz(ON):R5KΩ2.1.4BCICompetition2003GrazBCI[33],1304024(section),section40(trial)section3(Rest)160trial802.6141234SectionRest2.6Fig.2.6Timearrangementintheexperiment3trial9sFixationCross(+)VisualCue(←Or→)102389CueBeepTrialDuration=9000msTriggerSignal=0msReadySignal(fixationcross)=2000-8000msCueTiming=3000-8000msCueBeep=3000msTimingTriggerSignalTimeinSecond2.7Fig.2.7Schematicdiagramoftheparadigm12trial123trial2s8s4trial3s2.8(Beep)152.8Fig.2.8Imaginationcuesofleftandrighthandmovement5s(Beep)Trial3sVisualGuidePresentation2.1.5ChecklistNeuroscanScan4.3124/4021632section80trial4562.22.9Fig.2.9Blockdiagramofsignalacquisitionandprocessing2.9NeuroscanERP122.317ERD/ERS2.10NeuroscanC3C4CzERD/ERS122.10Fig.2.10Flowchartofsignalprocessing2.418ERDERS193(Wavelet),(autoregressivemodels,AR)[34-35],(ERS/ERD)[36](FFT),(PSD)(Time-Frequencyspectrum)(Laplacianfilter),(principalcomponentsanalysis,PCA),(Commonspatialpattern,CSP),(Independentcomponentsanalysis)3.1NeuroscanSCAN4.3C3C4Cz3.1.1Matlabn=index(1)-1;C3b=mean(fC3(n-4999:n));C4b=mean(fC4(n-4999:n));20ntrial1mstrial5sa1=mean(LC3(i,1:3000));a2=mean(RC3(i,1:3000));b1=mean(LC4(i,1:3000));b2=mean(RC4(i,1:3000));3saC3bC412c1=a1-C3b;c2=a2-C3b;d1=b1-C4b;d2=b2-C4b;LC3bc(i,:)=LC3(i,:)-c1;RC3bc(i,:)=RC3(i,:)-c2;LC4bc(i,:)=LC4(i,:)-d1;RC4bc(i,:)=RC4(i,:)-d2;3s3.1.2C3C4CzCz1020C3C4CzC3C4CzC3C4CzC3C4C3C4CzC3,C4CzMatlabb1=cov(Cz,C3)/var(Cz);C3Czb2=cov(Cz,C4)/var(Cz);C4CzC3n=C3-b1(1,2)*Cz;C3CzC4n=C4-b2(1,2)*Cz;C3Czb1≈