Eye Tracking Methodology Andrew Duchowski Eye Tracking Methodology Theory and Practice SecondEdition AndrewDuchowski,BSc,PhD DepartmentofComputerScience ClemsonUniversity Clemson,SC29634 [email protected] ISBN978-1-84628-608-7 e-ISBN978-1-84628-609-4 BritishLibraryCataloguinginPublicationData AcataloguerecordforthisbookisavailablefromtheBritishLibrary LibraryofCongressControlNumber:2006939204 (cid:2)c Springer-VerlagLondonLimited2007 Apartfromanyfairdealingforthepurposesofresearchorprivatestudy,orcriticismorreview,as permittedundertheCopyright,DesignsandPatentsAct1988,thispublicationmayonlybereproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,orinthecaseofreprographicreproductioninaccordancewiththetermsoflicencesissued bytheCopyrightLicensingAgency.Enquiriesconcerningreproductionoutsidethosetermsoflicenses issuedbytheCopyrightLicensingAgency.Enquiries concerningreproductionoutsidethoseterms shouldbesenttothepublishers. Theuseofregisterednames,trademarks,etc.,inthispublicationdoesnotimply,evenintheabsenceof aspecificstatement,thatsuchnamesareexemptfromtherelevantlawsandregulationsandtherefore freeforgeneraluse. Printedonacid-freepaper. 9 8 7 6 5 4 3 2 1 SpringerScience+BusinessMedia springer.com Preface to the Second Edition Sincethewritingofthefirstedition,severalimportantadvancementsinthefieldof eye tracking have occurred in the span of just a few short years. Most important, eye tracking technologyhas improveddramatically. Due to the increased speed of computerprocessorsandimprovedcomputervisiontechniques,eyetrackingmanu- facturershavedevelopeddevicesfallingwithinthefourthgenerationofthefollowing technologicaltaxonomy. 1. First generation:eye-in-headmeasurementof the eye consisting of techniques suchasscleralcontactlens/searchcoil,electro-oculography 2. Secondgeneration:photo-andvideo-oculography 3. Thirdgeneration:analogvideo-basedcombinedpupil/cornealreflection 4. Fourth generation:digital video-based combined pupil/cornealreflection, aug- mentedbycomputervisiontechniquesandDigitalSignalProcessors(DSPs) Often the most desired type of eye tracking output (e.g., for human–computerin- teractionusabilitystudies)isestimationoftheprojectedPointOfRegard(POR)of the viewer, i.e., the (x,y) coordinates of the user’s gaze on the computer display. First-andsecond-generationeyetrackersgenerallydonotprovidethistypeofdata. (For second-generation systems, eye movement analysis relies on off-line, frame- by-framevisualinspectionofphotographsorvideoframesanddoesnotalloweasy POR calculation.)Combinedvideo-basedpupil/cornealreflection eye trackerseas- ilyprovidePORcoordinatesfollowingcalibration,andaretodayderigeur.Dueto theavailabilityoffastanalog-to-digitalvideoprocessors,thesethird-generationeye trackersarecapableofdeliveringthecalculatedPORinreal-time.Fourth-generation eye trackers, having recently appeared on the market, make use of digital optics. Coupledwithon-chipDigitalSignalProcessors(DSPs),eyetrackingtechnologyhas significantlyincreasedinitsusability,accuracy,andspeedwhiledecreasingincost. Thestateoftoday’stechnologycanbestbesummarizedbyabrieffunctionalcom- parisonofolderandnewerequipment,giveninthefollowingtable.If,tothosenew tothisbookanduninitiatedwitheyetrackingdevices,thecomparisoninthetableis VI PrefacetotheSecondEdition Functionaleyetrackercomparison. LegacySystems State-of-the-Art Technology Analogvideo Digitalvideo Calibration 5-or9-point,tracker-controlled Anynumber,application-controlled Optics Manualfocusing/thresholding Auto-focus Communication Serial(polling/streaming) TCP/IP(client/server) Synchronization Statusbyteword APIcallback notimmediatelysuggestive,considertwousescenariosofbotholdandnewtechnol- ogypresentedinthenexttable. Eyetrackerusecomparison. Typicalsessionwitholdsystem. Typicalsessionwithnewsystem. 1. Logintoconsole. 1. Logintoconsole. 2. Turnoneyetrackingequipment. 2. TurnoneyetrackingPC. 3. Turnoneye/scenemonitors. 3. Runeyetrackingprogram. 4. TurnoneyetrackingPC. 4. Calibrate. 5. Runeyetrackingprogram. 5. Run. 6. Turnoncamera. 7. Turnonilluminationcontrol. 8. Adjusthead/chinrest. 9. Adjustpan/tiltunit. 10. Adjustcamerazoom. 11. Adjustcamerafocus. 12. Adjustpupil/cornealthresholds. 13. Calibrate. 14. Run. Thedisparityintheuseofoldandnewtechnologiesismainlyduetotheuseofdiffer- entoptics(camera).Newsystemstendtouseanauto-focusingdigitalcamera,e.g., embedded in a flat panel display. Although embedding within a flat panel display mayrestrictauser’sphysicalpositionsomewhat,itisgenerallypresettooperateata comfortablerange(e.g.,50–60cmfocaldistance).Unlikeoldersystems,aslongas theusersitswithinthisdistance,nochinrestsandnofurtherparameteradjustments are needed.In contrast, older devicesrequiredthe use of a pan/tilt unit to position thecamera,thecameratobemanuallyfocusedandzoomed,andsoftwaretobeset toappropriatepupilandcornealreflectiondetectionthresholds.Noneofthesecum- bersomeoperationsarerequiredwithnewersystems. Furthermore,oneofthemostimportantfeaturesofthenewtechnology,especiallyfor applicationdevelopment,isanindividual’sabilitytoself-calibrate.Witholdertech- nology,wheneveradeveloperwishedtotestanewfeature,sheorhehadtorecruit a(verypatient)subjectfortesting.Thiswasquiteproblematic.Thenewersystems’ PrefacetotheSecondEdition VII calibration routines are a much-needed improvement over older (third-generation) technologythatsignificantlyaccelerateprogramandapplicationdevelopment. A third-generation eye tracker was used for most of the eye tracking research on which the first edition of this book was based. The availability of new technology precipitatedthewritingofthesecondedition.Thesecondeditionthereforefillssev- eralimportantgapsnotcoveredpreviously,namely: 1. Client/servermodelfordevelopinganeyetrackingclientapplication 2. Client-controlleddisplay,calibration,datacollection 3. Newprogrammingexamples Beyond updated technical descriptions of client programming, the second edition alsoincludeswhatthefirsteditionlacked:anoverviewof themethodologybehind the use of eye trackers, that is, experimentaldesign issues that are often needed to conduct eye tracking studies. The second edition briefly reviews experimental de- signdecisions,offerssomeguidelinesforincorporatingeyemovementmetricsinto astudy(e.g.,usability),andprovidesexamplesofcasestudies. Finally,thesecondeditionexpandsthethirdpartofthebook:eyetrackingapplica- tions. A great deal of new and exciting eye tracking work has appeared, undoubt- edly drivenby the availability of new technology.In fact, there now appearsto be aratherrefreshingshiftinthereportingofeyetrackingandeyemovementstudies. Authorsnowtendtounderstatethe“gee-whiz”factorofeyetrackersandtheirtech- nical machinationsneeded to obtain eye movementdata and are now emphasizing scientificresultsbolsteredbyobjectiveevidenceprovidedbyusers’gazeandhence attention.Eyetrackingfinallyappearstobeenteringintomainstreamscience,where theeyetrackerisbecominglessofanoveltyandmoreofatool.Itishopedthatthis second edition may inspire readers with the simplicity of application development now made possible by fourth-generationeye trackers and continue on the road to newapplicationsandscientificinsights. AndrewT.Duchowski Clemson,SC,April2006 Preface to the First Edition Thescopeofthebookfallswithinafairlynarrowhuman–computerinteractiondo- main (i.e., describinga particular inputmodality),however,it spans a broad range ofinterdisciplinaryresearchandapplicationtopics.Thereareatleastthreedomains thatstandtobenefitfromeyetrackingresearch:visualperception,human–computer interaction,andcomputergraphics.Theamalgamationofthesetopicsformsasym- bioticrelationship.Graphicaltechniquesprovideameansofgeneratingrichsetsof visualstimulirangingfrom2Dimageryto3Dimmersivevirtualworldsandresearch exploringvisualattentionandperceptioninturninfluencesthe generationofartifi- cialscenesandworlds.Applicationsderivedfromthesedisciplinescreateapower- ful human–computerinteraction modality, namely interaction based on knowledge oftheuser’sgaze. Recent advancements in eye tracking technology, specifically the availability of cheaper, faster, more accurate, and easier to use trackers, have inspired increased eyemovementandeyetrackingresearchefforts.However,althougheyetrackersof- fer a uniquelyobjective view of overthuman visual and attentionalprocesses, eye trackershave notyetgainedwidespreaduse beyondwork conductedatvariousre- search laboratories.Thislack of acceptanceis duein partto two reasons:first, the useofaneyetrackerinanappliedexperimentalsettingisnotawidelytaughtsubject. Hence,thereisaneedforabookthatmayhelpinprovidingtraining.Itisnotuncom- mon for enthusiastic purchasersof eye tracking equipmentto become discouraged withtheirnewlyboughtequipmentwhentheyfinditdifficulttosetupandoperate. Only a few academic departments (e.g., psychology, computer science) offer any kindofinstructionintheuseofeyetrackingdevices.Second,toexacerbatethelack oftrainingineyetrackingmethodology,evenfewersourcesofinstructionexistfor system development.Settingup aneye trackinglab andintegratingthe eye tracker intoanavailablecomputersystemfordevelopmentofgaze-contingentapplications isafairlycomplicatedendeavor,similartothedevelopmentandintegrationofvirtual realityprograms.Thusfar,itappearsnotextbookotherthanthisoneexistsproviding thistypeoflow-levelinformation. X PrefacetotheFirstEdition The goal of this bookis to providetechnicaldetails for implementationof a gaze- contingentsystem,couchedinthetheoreticalcontextofeyemovements,visualper- ception,andvisualattention.Thetextstartedoutastheauthor’spersonalnotesonthe integrationofacommercialeyetrackerintoavirtualrealitygraphicssystem.These technicalconsiderationscomprise the middle chaptersof the bookand include de- tailsofintegratingacommercialeyetrackerintobotha3Dvirtualenvironment,and a 2D image display application. The surroundingtheoretical review chapters grew fromnotesdevelopedforan interdisciplinaryeye trackingmethodologycourseof- feredtobothundergraduatesandgraduatesfromfourdisciplines:psychology,mar- keting, industrialengineering,and computerscience. An early form of these notes waspresentedasashortcourseattheAssociationforComputingMachinery(ACM) SpecialInterestGrouponGraphics’SIGGRAPHconference,23–28July2000,New Orleans,LA. Overview Asofthesecondedition,thebookisdividedintofourparts,presentedthematically in a top-down fashion, providing first an introduction to the human visual system (Part I), then briefly surveying eye tracking systems (Part II), then discussing eye tracking methodology (Part III), and finally ending by reviewing a number of eye trackingapplications(PartIV). In the first part, “Introductionto the Human Visual System (HVS),” the bookcov- ers the concept of visual attention, mainly from a historical perspective. The first chapterfocusesonthedichotomyoffovealandperipheralvision(the“what”versus the“where”).Thischaptercoverseasilyobservableattentionalphenomena.Thenext chaptercoversthe neurologicalsubstrate of theHVS presentingthe low-levelneu- rologicalelementsimplicatedin dynamichumanvision.Thischapterdiscussesthe primarydualpathways,theparvo-andmagno-cellularchannels,whichlooselycor- respondtotheflowofvisualinformationpermittedbytheretinalfoveaandperiph- ery.Following this description of the visual “hardware”,observablecharacteristics ofhumanvisionaresummarizedinthefollowingchapteronvisualperception.Here, resultsobtainedmainly frompsychophysicsare summarized,distinguishingfoveal and peripheral visual perception. The first part ends by discussing the mechanism responsible for shifting the fovea, namely eye movements. Having established the neurologicalandpsychophysicalcontextforeyemovements,the followingchapter onthetaxonomyandmodelsofeyemovementsgivesthecommontermsforthemost basic of eye movementsalong with a signal-analyticdescription of recordableeye movementwaveforms. Thesecondpartofthebook,“EyeTrackingSystems,”presentsabriefsurveyofthe main types of available eye tracking devices, followed by a detailed technical de- scriptionoftherequirementsforsysteminstallationandapplicationprogramdevel- opment. These details are mainly applicable to video-based,corneal-reflectioneye PrefacetotheFirstEdition XI trackers, the most widely available and most affordable type of eye trackers. This part of the book offers information for the development of three general systems: oneforbinocular3Deyetrackinginvirtualreality,oneformonocular2Deyetrack- ingovera2Ddisplay(e.g.,atelevisionmonitoronwhichgraphicalinformationcan be displayed),and one for binocular2D eye tracking on the desktop. The descrip- tionsofthefirsttwoformersystemsareverysimilarbecausetheyarebasedonthe samekindof(older)eyetrackinghardware(ISCANinthisinstance).Thelattersys- temdescriptionisbasedonmoderneyetrackingtechnologyfromTobii.Bothsystem descriptionsincludenotesonsystem calibration.Thispartof the bookendswith a descriptionofdatacollectionandanalysisindependentofanyparticulareyetracking hardware. The fourth part of the book surveys a number of interesting and challenging eye trackingapplications.Applicationsidentifiedinthispartaredrawnfrompsychology, humanfactors,marketingandadvertising,human–computerinteractionandcollab- orativesystems,andcomputergraphicsandvirtualreality. Howto ReadThis Book The intended audience for this bookis an interdisciplinaryone, aimed particularly at those interested in psychology, marketing, industrial engineering, and computer science.Indeed,thistextismeantforundergraduatesandgraduatesfromthesedis- ciplines enrolled in a course dealing with eye tracking, such as the eye tracking methodologycoursedevelopedbytheauthoratClemsonUniversity.Inthiscourse, typically all chapters are covered,but not necessarily in the order presented in the text.Insuchacourse,theorderofchaptersmaybeasfollows. First,PartIVispresentedoutliningvariouseyetrackingapplications.Normally,this partshouldgivethereadermotivationfordesignandimplementationofasemester- longeyetrackingproject.Coverageofthispartofthebookisusuallysupplemented by readings of research papers from various sources. For example, papers may be selectedfromthefollowingconferences. • TheproceedingsoftheEyeTrackingResearch&Applications(ETRA)confer- ence • TheproceedingsoftheACMSpecialInterestGrouponHuman–ComputerInter- action(SIGCHI)conference(HumanFactorsinComputing) • Transactionson Graphics, the proceedingsof the annualAssociation for Com- puting Machinery (ACM) Special Interest Group on Graphics and Interactive Techniques(SIGGRAPH)conferenceseries • TheproceedingsoftheHumanFactorsandErgonomicsSociety(HFES) Tospeedupdevelopmentofaneyetrackingapplication,PartIIfollowsthepresenta- tionofPartIV,dealinginthetechnicaldetailsofeyetrackerapplicationdevelopment. XII PrefacetotheFirstEdition Thetypesofapplicationsthatcanbeexpectedofstudentswilldependmainlyonthe programmingexpertiserepresentedby membersofinterdisciplinarystudentteams. Forexample,intheeyetrackingmethodologycourseatClemson,teamsareformed byjoiningcomputersciencestudentswith oneormoreoftheotherrepresentatives enrolled in the class (i.e., from marketing, psychology, or industrial engineering). Although all group members decide on a project, students studying the latter sub- jectsaremainlyresponsibleforthedesignandanalysisoftheeventualeyetracking experiment. Givencommencementofaneyetrackingapplication,PartIIIisthencovered,going overexperimentaldesign.Inthecontextoftheusabilitymeasurementframework,the eye tracking methodology course advocates performance measurement, and there- forefocusesonlaboratoryexperimentsandquantitativedataanalysis. PartIofthetextiscoveredlast,givingstudentsthenecessarytheoreticalcontextfor the eye tracking pilot study. Thus, although the book is arranged “top-down”, the courseproceeds“bottom-up”. The book is also suitable for researchers interested in setting up an eye tracking laboratory and/or using eye trackers for conducting experiments. Because readers withthesegoalsmayalsocomefromdiversedisciplinessuchasmarketing,psychol- ogy,industrialengineering,andcomputerscience,notallpartsofthebookmaybe suitable for everyone. More technically oriented readers will want to pay particu- larattentiontothemiddlesectionsofthebookwhichdetailsysteminstallationand implementationof eyetrackingapplicationsoftware.Readersnotdirectlyinvolved with such low-level details may wish to omit these sections and concentrate more onthetheoreticalandhistoricalaspectsgiveninthefrontsectionsofthebook.The latterpartofthebook,dealingwitheyetrackingapplications,shouldbesuitablefor allreadersinasmuchasitpresentsexamplesofcurrenteyetrackingresearch. Acknowledgments ThisworkwassupportedinpartbyaUniversityInnovationgrant(#1-20-1906-51- 4087),NASAAmestask(#NCC2-1114),andNSFCAREERaward#9984278. Thepreparationofthisbookhasbeenassistedbymanypeople,includingKeithKarn, RoelVertegaal,DorionListon,andKeithRaynerwhoprovidedcommentsonearly editionsofthe text-in-progress.Laterversionsofthedraftwerereviewedbyexter- nalreviewerstowhomIexpressmygratitude,fortheircommentsgreatlyimproved thefinalversionofthetext.SpecialthanksgotoDavidWoodingforhiscarefuland thoroughreviewofthetext. IwouldliketothanktheteamatSpringerforhelpingmecomposethetext.Thanks goto BeverlyFordandKarenBorthwickforeggingme onto writethe textandto