Principle
Theprincipleoftheeyemovementexperimentismainly:thepopulargazetrackingtechnology,mainlybasedonthe"non-invasive"technologyofeyevideoanalysis(VOG,Videooculographic),readytouseThecamerarecordseyemovementsandthenusesimageanalysistodeterminewherethelineofsightfalls,andthesensordoesnothavedirectcontactwiththeuser.Someinstrumentsthatusethistechnologycanallowuserstomovefreely,suchasTobii'sX60/120andT60/120seriesbaremetal,butsometimestheyalsolimittheuser'srangeofactivitiesinsomeways,suchasrequiringuserstoplacetheirchinonafixedOnthebrackettogetmoreaccuratedata.
Weknowthattheuser’slineofsightismovingandtheeyeballisalsoturning,sotocalculatetheuser’slineofsightdirection,wehavetofindacertainfeatureintheeyeimagethatremainsunchangedwhentheeyeballrotates.Andcalculatethevectorrelationshipbetweenitandthecenterofthepupil(thecenterlineisthedirectionofthelineofsight).
Application
Themethodwidelyusedingazetrackingtechnologyiscalledthe"pupilcentercorneareflectiontechnique"(thepupilcentercorneareflectiontechnique).ThecharacteristicofthechangeisthePurkinjeimageontheoutersurfaceoftheeyeballcornea-abrightspotontheeyeballcornea,whichisproducedbythecornealreflectionofthelightenteringthepupilontheoutersurfaceofthecornea.
Becausethepositionofthecameraisfixed,thepositionofthescreen(lightsource)isfixed,andthecenterpositionoftheeyeballremainsunchanged(assumingthattheeyeballissphericalandtheheaddoesnotmove),theabsolutepositionofthePurkinspotdoesnotfollowtherotationoftheeyeballAndchange(infact,smallmovementsoftheheadcanalsobecalculatedbycornealreflex).Butitspositionrelativetothepupilsandeyeballsisconstantlychanging-forexample,whenyoustareatthecamera,thePurkinspotisbetweenyourpupils;andwhenyouliftyourhead,thePurkinspotisonyourBelowthepupil.Inthisway,aslongasthepupilandPurkinspotontheeyeimagearelocatedinrealtimeandthecornealreflectionvectoriscalculated,thegeometricmodelcanbeusedtoestimatethedirectionoftheuser'slineofsight.Basedontherelationshipbetweentheuser’seyecharacteristicsandthecontentpresentedonthecomputerscreenestablishedinthepre-calibrationprocess(thatis,allowingtheusertolookataspecificpointonthecomputerscreen),theinstrumentcandeterminewhattheuserislookingatonthescreen..
Locatingthepositionofthepupilcenterisakeystepinthegazetrackingtechnology,butoneproblemisthatcomparedwiththeextremelyobviousdividinglinebetweentheirisandthewhiteoftheeye,thedifferencebetweenthepupilandtheirisTheboundariesarenotsoclear,especiallyourdarkeyesandyellowskin.Therefore,inordertoimprovetheaccuracyofthisstep,theresearchersdesignedthe"brightanddarkpupildifferencescheme",thatis,alternatelyusedifferentazimuthlightsourcestoemitnear-infraredraystothehumaneye,andthenineverytwoadjacentframes,Obtaintheuser'sbrightpupil(brightpupil,brightpupil)anddarkpupil(darkpupil,darkpupil)respectively,superimposethedifference,soasto"cutout"thepupilmoreclearly,andthencalculatethecenterofmassandshapeofthepupilandotherparameters.
Whetherthe"brightpupil"orthe"darkpupil"iscaptureddependsonwhetherthecameraiscollinearwiththelightsource.Ifthecameraandthelightsourceareonthesameline,thepupilcapturedbythecameraisilluminatedbylight,thatis,the"brightpupil".Thisissimilartotheprincipleof"redeyes"appearingonthephotowhenthecameraflashispointeddirectlyatthesubject.Irrelevant,secondly,thepupilswillbecomesmallerunderstronglight,otherwiseyouwillbeblindedbytheflash).Ifthetwoarenotcollinear,youwillgetanormal"darkpupil".Therefore,theeyetrackerthatsupportsbrightanddarkpupiltrackinghastwosetsofnear-infraredlightsourceswithdifferentpositions.
Thereasonforusingnear-infraredlightisthatthehumaneyecannotdetectit,anditwillnotdazzletheeyesandaffecttheuser.Theselightbeamsareveryweak.Aslongastheresearcherarrangestheusertositatthedistanceindicatedintheeyetrackermanual(forexample,morethan60cmfromtheeyetracker),theuserwillnotbeexposedtoradioactivehazardeveniftheuserstaysfor8hoursinfrontoftheeyetrackeratwork.
Thegazetrackingtechnologycorrespondingtotheabove-mentioned"non-intrusive"technologyrequirestheusertodirectlycontactthesensoronthetestdevice.Forexample,intheearlyeyemovementtest,somethingsimilartoahardcontactlensisinsertedintothetester’seyestomonitortheconstantlychangingmagneticfieldwitheyemovement,soastoknowwhereyouarelookingoraroundthetester’seyes.Attachelectrodestomonitorpotentialchanges.Thesemethodssoundabitintimidatingandcumbersometooperate,butthedataobtainedismoreaccurate.
So,howaccuratecanordinarycommercialeyemovementtestsbe?Thisdependsonthespecificparametersofthetesteyetracker.Itisdividedintotwodimensions:spaceandtime:therelatedparametersinspaceareaccuracy,driftandscreensize,andtheparameterintimeisthesamplingrate(delay).Forexample,theaccuracyofTobiiX120is0.5degrees,andthedriftovertimeiswithin0.3degrees.Iftheuseris60cmawayfromthescreen,theoffsetisabout0.13mm;thesamplingrateis120Hz,andthedelayis17ms.Becausethepupilcanonlybecountedeverytwoframes.However,someresearchershavefoundthatthepositiondeviationintheactualtestismuchlargerthanthevaluecalculatedhere,whichmayberelatedtotheuser'sheadmovementorcalibrationproblems.Ifyouusethetobiiseriesforreadingtests,itmaynotbeabletoaccuratelylocatewhichlineofcharacterstheuserislookingatontheinterface.Therefore,whendoinganalysis,avoidbelievingtoomuchinthegazegivenintheresults.
Atthesametime,whendoingtests,youshouldalsotrytocomplywiththeexperimentalspecifications.Currentcommercialeyetrackerscangenerallycompensateforheadmovement.However,eveniftheeyetrackerallowsuserstomovefreely,thereisaprescribedheadmovementrange.Forexample,theheadmovementrangeofTobiiX60andT60modelsis44×22×30cm.(Length,widthandheight),whilethefrequencyofX120andT120ishigh,andtheallowableheadmovementrangeissmaller,whichis30×22×30cm(length,widthandheight).Theuser'sheadmovementrangeshouldbeensuredduringthetest.Inthecalibration,theusershouldbeallowedtomovetheheadwithinthespecifiedrange,andtheheadmovementshouldbetakenintoconsiderationduringthecalibrationstage.