SchoolMotokiofKnowledgeMiura
Science
JapanAdvancedInstituteofScienceand
Technology
1-1Asahidai,Nomi,Ishikawa
923-1292,Japan
miuramo@jaist.ac.jp
ABSTRACT
Portablehandhelddevicesinherentlyinvolvedifficultieswithmethodsofinputduetotheircompactsize.Severalap-proachestoattachextrasensorshavebeenproposed,butthesehavenotenabledsizeorexteriordesigntobemini-mized.Weproposeanovelandsimpleinputtechniqueforhandhelddevicesthatmakesuseofastylusinaholderthatistwistedandpushed/pulledlikeaknob.Bothrotatingandslidingthestylusinsidetheholdercansimultaneouslyad-justtwoparameters.Weimplementedaprototypesystemwithaninexpensiveimagesensor,andevaluateditsinput.AnANOVAtestrevealedthatourmethodcouldscrollasfastastap-and-dragoperationsonascreen.
CategoriesandSubjectDescriptors
H.5.2[Informationinterfacesandpresentation]:UserInterfaces—interactionstyles,evaluation/methodology
GeneralTerms
Design,HumanFactors,Performance,Experimentation
Keywords
interactiontechnique,handhelddevice
1.INTRODUCTION
PortablehandhelddevicessuchasPDAsandmobilephonesinherentlyinvolvedifficultieswithinputmethodsduetotheircompactsize.MostPDAshaveadisplaywithatouch-sensitivepaneltoenabletapping,holding,anddraggingwithastylus.TheseoperationsareexploitedtointeractwithGUIobjectssuchasicons,buttons,andknobsonascreen.Althoughpen-basedoperationsareversatile,theyrequireprecisecontroltospecifysmallerGUIobjects.Also,thepre-andpost-actionsofremoving/storingthestylusdis-ruptsmoothtransitionsintheinteractionmode.
Permissiontomakedigitalorhardcopiesofallorpartofthisworkforpersonalorclassroomuseisgrantedwithoutfeeprovidedthatcopiesarenotmadeordistributedforprofitorcommercialadvantageandthatcopiesbearthisnoticeandthefullcitationonthefirstpage.Tocopyotherwise,torepublish,topostonserversortoredistributetolists,requirespriorspecificpermissionand/orafee.
MobileHCI’06,September12–15,2006,Helsinki,Finland.Copyright2006ACM1-59593-390-5/06/0009...$5.00.
SchoolSusumuofKnowledgeKunifuji
Science
JapanAdvancedInstituteofScienceand
Technology
1-1Asahidai,Nomi,Ishikawa
923-1292,Japan
kuni@jaist.ac.jp
Variousideasandapproachestoachievenaturalinterac-tionwithportablehandhelddevicesbyutilizingtheirpo-sition,posture,andmovementhavebeenproposed.Fitz-mauriceetal.[4]introducedamethodemployingpositionandorientationtonavigatevirtualspaceusingaconceptinvolvinganaugmentedrealitytechnique.Rekimoto[14]appliedtiltingtonavigatemapsandselectmenuswithacombinationofmodekeys.HyperPalette[1]enableduserstocollectvirtualcardsby“scooping”withthedevice.Toss-it[16]recognized“tossing”gesturestoenabledatatobetransferred/distributedamongneighboringPDAsandinfor-mationappliances.Theseapproacheseffectivelyachievednaturalandintuitiveoperationsbyexploitingthephysicalmovementsofthedevices.Althoughtheschemesbasedonposition,posture,andmovementwereundoubtedlysimpleandintuitive,detectingthethree-dimensionallocationwithsensorshighlydependsontheenvironment.Thetiltingap-proachalsotendedtorestrictthepostureofthedeviceinuse,andwasnotsuitableforcontinuousorprecisecontrol.Othertechniqueshaveincludedtheattachmentofpres-sure/touch-sensitivesensorstocapturenaturalgesturesandintentions.Harrisonetal.[6]describedinteractionsthatemploypressure,posture,andtiltsensorstoachieverealis-ticpage-turningandscrollingtasks.Hinckleyetal.[7]pre-sentedseveralcontextualinteractionsbyintegratingthesesensorstoimproveoperationssothattheywouldbemorenatural.Althoughallthesetechniquesarepromising,ex-posedpressure/touch-sensitivesensorsessentiallyinfluencetheexteriordesignofthedevice.
WeproposeanovelandsimpleinteractiontechniqueforhandhelddevicesutilizingRodDirect,whichusesastylusthatistwistedandpushed/pulledlikeaknob.Styliarenor-mallyassociatedwithsmall-screeneddevicessuchaspalm-topcomputersandotherhandheldappliances,andaregen-erallymanipulatedbytapping,holding,andstrokingthemonatouch-sensitivepanel.Ourapproachexploitsthestylustoachieveinputthroughphysicalmetaphorsratherthanbyconventionaltapping.
2.
RODDIRECT
Stylusholdersaretypicallyonlyusedtostorestyliwhendevicesarenotbeingused.RodDirectexploitsthepositionandmovementofthestylusinitsholdertogenerateinput.Thedegreesoffreedom(DOFs)ofastylusstoredinitsholderdependbothonitsownshapeandthatoftheholder.Assumingthatitscrosssectioniscircular,itcanbeslidandrotatedasshowninFigure1.Consequently,whenthedevice
SlideRotationFigure1:MovementsofstylusinholderFigure2:TypicalinteractionwithRodDirect(con-ceptualimage)
sensesrotationandsliding,thesemovementscanbeutilizedtocontroltwoparameters.Wethereforeemployedthesetwoparameterstofacilitateinteractionwithhandhelddevices.Although,therearenolimitationsonrotation,therangeofslidingmovementsisphysicallylimited.However,thesecanbeexpandedbyapplyingvariousinteractiontechniques.Figure2showstypicalinteractionwithRodDirect.Theuserisholdingthestylusbetweenhisthumbandforefinger(themiddlefingercouldalsobeused)andisoperatingthePDAbyrotating/slidingthestylus.Eventhoughhewouldbeabletomaintainitspositionifhewerenotholdingit,thiscanonlybeachievedwhenthereissufficientfrictionbetweenitandtheholder.ThesamefigureFigure2alsoshowsbothhandsbeingusedformanipulation;thenon-dominanthandtoholdthePDAandthedominanthandtotwistandpush/pullthestylus.IfthePDAhadbeenplacedonacradle,hewouldhavebeenabletocontrolthestyluswithonehand.Furthermore,rotatingandslidingwiththenon-dominanthandarerelativelyeasierthantappingbecauseofthephysicalconstraintsimposedbytheholder.
2.1Comparisonwithregularstylusoperations
Fromourobservationsoftappinghandhelddeviceswitharegularstylus,wecansaythatusersusuallyholdthesty-luswiththeirdominanthandandthehandhelddevicewiththeirnon-dominantone.Inthesecases,therearenophysi-calconstraintsaffectingtheuseofthedeviceorthestylus.This“free”situationissuitablefortakingnotes,drawing,orscribbling.However,whenusersaremobile,theyhavetobe
carefultostabilizetheirhandhelddevicestotapaccurately.Thisisachievedbyaligningtherelativepositionofthesty-lustothetouch-sensitivedisplay.Usershavenaturallybeenstabilizingtheirhandhelddevicesbysupportingthemwiththeirdominanthandorbyproppingthemupbyplacinganelbowontheirside.Thesebehaviorshaveparticularlybeenobservedinunstablesituationssuchaswhentravelingonbusesortrains.AlthoughRodDirectrequiresbothhandstoholdandop-eratethehandhelddevicewhenusersaremobile,therearemorephysicalconstraintsbetweenthestylusandholdertoachievestability.Thus,maintainingtherelativepositioniseasierthanwithtapping.Further,sufficientfrictionhelpsuserstofreetheirdominanthandwhenthedeviceisnotbeingoperated.Thiscanbeadvantageouswhenusersfre-quentlyswitchbetweenoperatingandnon-operatingmodes.Whenconventionalstyliarebeingoperated,ontheotherhand,usershavetoreplacethemintheirholdersorelse-whereandretrievethemtousetheirhandhelddevicesagain.Indeed,conventionallytappinganddraggingscreenswithstyliismoreeffectiveinadvancedtasksthatrequirecontin-uousdirectpositioningsuchasinputtingalphanumericsandtakingnotes.Thus,RodDirectdonotwhollyreplacescon-ventionalstylusoperations.Wecancategorizeinteractionsonhandhelddevicesasfollows.
•(Phase1)onlybuttons(one-handed)
•(Phasebutuser2)candockedquicklystylusreturnandbuttonstophase(mainly1)two-handed,•(Phasehanded)3)conventionaltappinganddragging(two-Conventionaltappinganddragginginteractionsareun-doubtedlymoreeffectiveforbothcomplicatedandsimplertasksinPhase3.However,insimplertaskssuchasscrollingandadjusting,userscanonlyfinishtheseinPhase1,buttheyarestillindirect.RodDirectcoverstheintermediateinteractionwithPhase2,fillinginthegapsbetweenthefirstandlastphases.
2.2Consideration
RodDirecthasafunctionsimilartoavolumeknobap-pendedtothestylusinadditiontoitsconventionaluseasapen.Typically,theactionof“rotatingaknob”iscommonlyusedtoadjustthevolumeofaudio/visualequipmentortotuneradios.Hence,RodDirectaccommodatestasksinvolv-ingfineadjustments.Inconventionalinteractionsformak-ingadjustmentwithtouch-sensitivepanels,usershavetodragscrollbarsorsliderknobs,ortotapbuttonsup/downtocontinuouslychangetargetvalues.Further,asrotatingknobshavebeenexploitedforwindingandregulatingwrist-watches,theyhaveessentiallybeenappropriateforsmallin-struments.However,regulatorytasksinvolvenotonlyfinetuningbutalsomajoradjustments.Bothour“rotating”and“sliding”operationscanfulfillsuchcontradictorytasks,forinstance,precisecueingofmovies/musicorzooming.Inaddition,RodDirectcancarryoutawiderangeoftaskswithasinglecontroller.
RodDirectinteractionsarealsocompatiblewithconven-tionaltouch-paneloperations;thus,applicationscansimul-taneouslymakeuseoftheseinteractions.Inagrapheditor,forexample,userscanselectanobjectbytappingandhold-ingthetouch-sensitivepanel.Atthesametime,theycan
(a)Backview
(b)Frontview
Figure3:RodDirectprototype.SensorboardofopticalmouseisattachedbehindPDA.
subsequentlychangethecoloroftheobjectbyrotatingthestylus.Interactionssuchas“holdingandcontrolling”canbeappliedtothemanagementofschedulesandzoomingonmaps.Furthermore,RodDirectcanworkincombina-tionwithothersensorssuchasthosethatdetecttiltandpressure.
3.IMPLEMENTATIONOFPROTOTYPE
Wedecidedtoadoptanopticalimagesensorcommonlyinstalledinopticalmicetoimplementtheprototypetoinves-tigatehowusableRodDirectwas.Usuallyanopticalmouseusesanopticalimagesensortodetectitsmovementoveratable.Weusedtheopticalimagesensortodetectthemovementofastylus.Whenthestyluswasrotatedorslid,theimagesensordetectedmovementandreturnedtheac-cumulatedvalueasalocation;thex-axisrepresentedthedegreeofrotationandthey-axistheamountofslide.ThedetectingmechanismwiththeopticalmousewasinspiredbyMouseField[12].
Figure3(a)and(b)arephotographsofthebackandfrontofourprototype.ItemploysaPocketPC(hpiPAQh1930)forthePDA.Weattachedthesensorboardofanopticalmouse(ELECOMM-BG2URLBU,800countsperinchres-olution)tothePDAwherethestyluswasexposed.ThesensorshouldbeplacedneartheedgeofthePDAtofacil-itateasmuchrecognitionofslidingmovementaspossible.Thesensordetectsabout420countsperrotationandabout1200countsperone-wayslideofthestylusofthePocketPC.Thenumberofcountsisinfluencedbymanyfactors,i.e.,resolution,wherethesensoris,andthelengthofthestylus.
Figure4:MetaphorforScrollingandSlidingFigure5:MapViewerFigure6:Scheduler
WeattemptedtoconnectaUSBmousetothePocketPC.However,aconventionalPocketPCcannotrecognizeaUSBmousebydefaultbecausetheyarebothregardedasperiph-eralUSBdevices.Therefore,wehadtoutilizeaPCasanintermediarytotransmitthemovements.Theopticalsen-sorwasconnectedtotheintermediaryPC.Wedevelopedaserverprogramthatretrievedthelocationsofmultiplemiceandprovidedthelatestvaluesonclientrequests.TheintermediaryPCrantheserverprogramtoprovidemove-mentdata.ApplicationsranonthePDAandcouldretrievethemostrecentvaluesfromtheopticalmousesensorsviaastandardTCP/IPconnection.TheconnectionbetweenthePDAandintermediaryPCwasestablishedthroughaUSBwithMicrosoftActiveSync3.7.
3.1Applications
TheapplicationsdiscussedinthefollowingdemonstrateourconceptandexplainsomefieldswhereRodDirectin-
Figure7:MetaphorforMovingObjectsFigure8:BlockBreaker
Figure9:SpaceInvaders
teractioncouldbeapplied.Theseapplicationswereim-plementedwithMicrosofteMbeddedVisualC++3.0andGapiDraw[2],whichprovidedatwo-dimensionalgraphicslibrary.
3.2MapViewer
RodMapViewer(Figure5)providedmapbrowsingfunc-tionswithscrollingandzooming.Scrollingwasbasedonthemetaphorof“rolling”and“sliding”outlinedinFigure4.Althoughtherangeof“sliding”islimited,wepreparedaclutchbutton.Allslidingoperationsarecanceledwhilethebuttonisbeingpressed.Thezoomratecanbecontrolledbyrollingthestylus(similartothewheelonamouse),whilepressingthezoom-functionbutton.
3.3Scheduler
RodScheduler(Figure6)enablestheusertomanagesched-uleitems.Hecanscrollthroughthem.Theschedulerem-ploysthesamemetaphorof“rolling”and“sliding”outlined
Figure10:ApplicationSwitchingTask
inFigure4.Therefore,itsbasicoperationissimilartothatofthemapviewer.However,theschedulerallowsuserstoholdanddragascheduleitemtomodifyitbyusingtheirthumb.Theycanfreelyscrolltoalterthedateandtimeofaneventwhileholdingtheitem.
3.4Games
Wedevelopedtwogames—RodBlockBreaker(Figure8)andRodSpaceInvader(Figure9).Thesebothemploythemetaphorofa“movingobject”outlinedinFigure7tocon-trolobjectssuchasapadandacannon.Theprimaryoper-ationinthesegamesis“movingtheobjecthorizontally”byslidingthestylus.However,therotationcanbeassignedforeachgame.Intheblockbreakergame,userscanmanipu-latethepadverticallyorrenderamassiveboundingimpact.Inthespaceinvadergame,rotationisallocatedtoshootingoutlaserbeams.Userscancontrolthebeam’sspeedthroughtheamountofrotationandcanaccumulatelaserpowerbyrotatingintheoppositedirection.
3.5
Utilities
ThemovementsofastyluscanefficientlybeutilizedinregularPDAuse.WedevelopedaconceptualdemosystemthatemploysRodDirectinteractionasameansofapplica-tionswitching(Figure10).Userstypicallyswitchbetweenapplicationssuchasthescheduler,contacts,andemailbypressingahot-keyonaPDA.Ourdemosystemenablesthemtoswitchbetweentheseapplicationsbyrotatingthestylus.Whentheydrawsoutthestylus,thesystemrecognizestheslidingmovementandinformsthemthattheswitchingmodehasterminated.Therecognitionofthe“drawingout”eventcanbeallocatedtoactivateaparticularscreenmodesuchasthatforinputtingalphanumerics.RodDirectinteractioncanbeappliedtootherfieldsthatrequirecontinuousad-justmentsuchasvolumeandbrightnesscontrols.
4.RELATEDWORKS
DualTouch[13]isaninteractiontechniquethatenablesuserstooperateaPDAbytappingandstrokingthescreenusingapenandtheirthumb.Itemploysacharacteristicofthepressure-basedtouch-sensitivepaneltodetectthecom-binedmovementsoftwopoints.Therefore,DualTouchcanbeappliedtomostPDAswithouttheneedforadditionalhardware.However,thedegreeofinteractionisrestrictedbythedisplaysizeandresolutionofthetouch-sensitivepanel.
ScrollDisplay[15]andScrollPad[3]havebeenproposedthatinvolveascrollingtechniquethatutilizesapositionsen-sorsimilartoamouseinstalledbehindthedevice.PeepholeDisplays[17]adoptsaspatiallyawaretechniquetoenhanceinteraction,andpresentsexamplesofdrawings,editing,andbrowsingthroughzooming.Althoughthelocationandspa-tiallyawaretechniquesareeffective,especiallywhenworkingwithlargecontinuousvirtualscreens,theyrelyonaninfras-tructuretotrackthedevices,suchasaflatsurfaceanda3Dposition-trackingenvironment.OurRodDirecttechniqueissimple,andworksunderanycircumstances.
BehindTouch[8]hasatouch-sensitivepanelattachedbe-hindacellularphone.Usersemploytheirthumbtopressnormalbuttonsandtheirforefingertotouchthepanelin-stalledonthebacktoselectmenusandinputalphanumer-ics.Thisapproachhasgreatpotentialtocoexistwithcon-ventionalapplicationsusingbuttons.However,thetouch-sensitivepanelthatisattachedmayinfluencetheappear-anceanddesignofthecellularphoneitself.RodDirectdoesnotaffecttheouterappearancewhenasensorisembeddedinadevice.
Jogdialandasimilardevicereferredtoas“ascrolljoy-stick”havebeeninstalledonsomePDAs.Mostofthesehavebeenattachedtotheleft,enablinguserstoscrollthroughdocumentsandselectmenusbyrotatingandpushingwithonehand.AlthoughRodDirectrequiresbothhandsforcon-trolinmostcases,itoffersslidingaswellasrotation.
5.EXPERIMENT
Weconductedanexperimenttoevaluateboththechar-acteristicsandperformanceofRodDirectinput.
5.1StudyDesign
TofindouthoweffectiveRodDirectinputis,“sliding”shouldbeconsideredaswellas“rotating.”Wethuschoseatwo-dimensionalscrollingtaskthatentailedsimultaneousadjustmentstotwoparameters.Wedevelopedanapplica-tionfortheexperimenttocollectinteractiondatafromthescrollingtask.
Theapplication(seeFigure11)displayedacomet(thetarget),atail(indicatorfortargetdirection),andacentercursorinafield.Thefieldwas1280pixelsinwidthby1200inheight,butthecometwasintheinnerfieldof960pixelsinwidthby960inheight(Figure12).Subjectswereaskedtoscrollthefieldtoplacethecomet’scenter,andthenpressabuttonontheleftbezel.Ifthecentercursorpointedtothecomet,thecometdisappearedandthenreappeared.Oth-erwise,theapplicationwhistledtonotifythatthecomet’spositionwasnotacceptable.Thedirectionthecometwasheadedwasrandom,butthedistancewascontrolledbytheapplication.Theinitialdistancewas100pixels,andthiswasincreasedby80pixelspertrial.Whenthe10thtrial(820pixelsdistance)wascompleted,thediameterofthecometwasreduced—50pixelsforthefirst10trials,40pixelsforthenext10trials,and20pixelsforthefinal10trials.Whenthediameterwasdecreased,thedistancewasinitializedto100pixels.Hence,subjectscouldanticipatethedistance.Evenifthetargetwasplacedoutofview,subjectscouldperceiveboththedirectionandthedistanceofthetargetbythepathandthicknessofthetail.Also,thefieldwastexturedtohelpsubjectsrecognizescrollingoperations.Al-thoughtheapplicationcollecteddatafrom10trialsforeach
Figure11:Screenshotofexperimentalapplication.Diameterofcometdecreasedfrom50to20pixels(targetwas50pixels)
1280 (320x4)960 (320x3)1200(240x5)960(240x4)Figure12:Screensizeandinnertargetfield
diameter,wedestroyedthefirsttwosincethetargetcometwouldappearintheinitialview.
Wepreparedfourinputtechniquestonavigatethroughthefield—[Rod],[Drag],[Walk],and[Tilt].[Rod]mapsthemovementsofthestylustoscrollwithametaphor,asshowninFigure4;rotationwasassignedtoverticalscrollingandslidingtohorizontal.[Drag]enablespanningaroundthefieldbyenablingholdingwithtapping,andpanningwithdragging.[Drag]canbereferredtoas“panningbypushingthebackground”[9].[Walk]providescontinuousscrollingbymappingthedisplacementofdraggingtothevelocityofnavigation;thedirectionanddistanceofthelocationofthedraggingpointfromthetappingstartpointarefrequentlyaccumulated.[Walk]canbereferredtoas“Touch-n-Go”[10].[Tilt]mapstheinclinationofthedevicetothescrolldirectionandvelocity.Toenabletilting,weattachedaPhidgetAccelerometer[5](ADXL320,2dualaxisaccelerom-eterthatcanmeasure±49.0m/speraxis)behindthede-vice.ThesensordataweretransferredviatheintermediaryPC,thesameaswiththe[Rod]method.Theaccelerometerdata(acc)rangedfrom−1to1(−90to90degrees),andthescrollingratio(velocity)wascalculatedusing200×acc2where|acc|>0.01,otherwiseitwaszero.
Figure13:Experimentalsetting
WeemployedaPocketPC(hpiPAQh1930),andplacedthedevicesideways;thedisplaywidthwas320pixelsanditsheightwas240.Theresolutionofthetouch-sensitivedisplaycorrespondedtothedisplaysize.Thestylusholderwas92mmdeep,andtheimagesensorwasattachedatthe37mmpointfromthetop.Therefore,thesensorcouldscan55mmofsliding.Theoriginallengthofthestyluswas94mm,buttoeliminatelimitationswithsliding,weadoptedalongerone(110mm).Onerotationofthestyluswith[Rod]resultedinabout420counts,andone-wayslidere-sultsinabout1880counts.Withthelongerstylus,subjectscouldcompletetrialswithouttheneedforspecialopera-tionssuchasclutchingeven1280pixelswidthofthevirtualscreen.Theoriginalstyluswasmadeofaluminumandplas-tic,whereasourstyluswasmadeofcastiron.
Ninegraduatestudents(right-handedmales,aged23to35)wererecruitedassubjects.Figure13showsthesettingfortheexperiment.Tocounterthebetween-subjecteffect,weconductedawithin-subjectexperiment.Wecategorizedtheinputmethodsintotwogroups:[Rod][Drag]tomovethebackground,and[Walk][Tilt]tocontrolthevelocity.Wethenchangedtheorderwithinandbetweenthegroupstocounterbalancethem.Thesubjectsweretrainedforabout5minuteseachforbothinputmethodspriortotheexperi-ment.Theapplicationrecordedthetimetocompleteeachtrial,fromwhenthetargetappeareduntilthetargetwasclicked.
5.2Result
Sincetheexperimentinvolvedwithin-subjectfactors,weanalyzedthetimingdatabyrepeatedANOVA(fourin-putmethods×eighttargetdistances).Themostsignifi-canteffectoftheinputmethodwasfoundacrossdistances,F(3,81)=195.9,p<.001.Further,Bonferronipair-wisecomparisonsrevealedthatthe[Rod]methodwassignifi-cantlyfasterthaneither[Walk]or[Tilt],and[Walk]wassignificantlyfasterthan[Tilt].However,nosignificantdif-ferenceswerefoundbetween[Rod]and[Drag].Themostsig-nificanteffectofdistancewasfoundacrosstheinputmeth-ods,F(7,189)=53.81,p<.001.
Weinvestigatedthecharacteristicsofthemethodofinputbyfittingthemtothepredictionmodel:
MT=a+blog2(A/W+1)
7000
’Fitts_Rod.txt’Rod(x)
6000
5000)cesm( e 4000
miT tnem 3000
evoM 2000
1000
0
5 5.5 6
6.5 7 7.5 8 8.5
log2(A/W+1) (bit)
Figure14:Rod
7000
’Fitts_Drag.txt’Drag(x)
6000
5000)cesm( e 4000
miT tnem 3000
evoM 2000
1000
0
5 5.5 6
6.5 7 7.5 8 8.5
log2(A/W+1) (bit)
Figure15:Drag
reportedbyMacKenzieetal.[11],whichwasa2DextensionofFitts’law.Intheirformula,Aindicatedthedistancebetweentheinitialpositiontothetarget’scenter,andWindicatedthediameterofthetarget.Figure14–17showsthescatterplotsandregressionlines.Figure18summarizestheregressionlines,andTable2liststheestimatedparametersofaY-intercept(a)andaslope(b).Theslopevalue(b)of[Rod]waslessthanthatfortheotherthreeinputmethods.Thus,[Rod]tendedtoreducethemovementtimewhentheindexofdifficulty(ID)wasincreased.
5.3Discussion
Theresultsoftheexperimentrevealedsubjectscouldper-formscrollingtaskswith[Rod]asfastastheycouldwithdraggingoperations.RodDirectalsohasthepotentialtoreducethemovementtimewhenthescrollingdistancein-creases.Consideringthetimerequiredtodrawoutthesty-lusandthenholditbeforetapping,[Rod]mayhaveanad-vantageinbrowsing-onlytasks.
Accordingtointerviewsconductedafterthescrollingex-periment,mostsubjectspreferred[Drag]becauseitenabledprecisecontrol.However,dragoperationsexhaustedthembecausethebestperformancerequiresfrequentbackand
7000
’Fitts_Walk.txt’Walk(x)
6000 5000)cesm( e 4000miT tnem 3000evoM 2000 1000 0
5 5.5 6
6.5 7 7.5 8 8.5
log2(A/W+1) (bit)
Figure16:Walk
7000
’Fitts_Tilt.txt’Tilt(x)
6000
5000)cesm( e 4000
miT tnem 3000
evoM 2000
1000
0
5 5.5 6
6.5 7
7.5 8 8.5
log2(A/W+1) (bit)
Figure17:Tilt
forthdragging.Although[Rod]causedlessfatiguethan[Drag],trainingisnecessarytoachieveaccuratecontrol.Somesubjectspointedoutthattasksweredisruptedbyre-flectionsfromroomlightingonthescreenunder[Tilt]con-ditions.
Wedidnotcompareourmethodwithscrollbarsinthisexperimentbecauseitdidnotpermitobliquescrolling.Weintendtoassesswhateffecttrainingsubjectsintheinputmethodsaswellasscrollbarswillhaveontheirfutureper-formance.
6.CONCLUDINGREMARKS
WeproposedanoveltechniqueofinteractionutilizingRodDirecttomanipulatehandhelddevicesusingthemove-mentsofastylusinsideaholder.Weimplementedaproto-typeandseveralapplicationstodemonstratefieldsinwhichitcouldbeused.
Weevaluateditsinteractioninscrolling.Theresultsre-vealedthatRodDirectcanscrollasfastasthetap-and-dragoperationscan.
Astheopticalimagesensorisrelativelysmallandinex-pensive,itiseasytoattachtoPDAs.Furthermore,itismorestablethanvariousaccelerationandpositiontypes.
7000
Rod(x)Drag(x)Walk(x) 6000
Tilt(x)
5000
)cesm( e 4000
miT tnem 3000
evoM 2000
1000
0
5 5.5 6
6.5 7 7.5 8 8.5
log2(A/W+1) (bit)
Figure18:Regressionlines
Table1:Estimatedmarginalmeansoftime(sec)
95%ConfidenceIntervalMethodMeanStd.Err.LowerBoundUpperBoundRod2.555.0612.4302.681Drag2.620.0422.5342.706Walk2.967.0792.8563.130Tilt3.811.0703.6673.954
Table2:Estimatedparameters(basedonformulasbyMacKenzieetal.)
Methoda(msec)b(msec/bit)Rod-1196.4547.1Drag-2655.0769.1Walk-2843.4844.2Tilt-1790.8785.2
Althoughthesensorboardisexposedinthecurrentproto-type,itcanbefullycovered,andthischaracteristiccancon-tributetomoreflexibleexteriordesign.Webelievethatthis
interactiontechniquewillenhancetheusabilityofhandhelddevicessuchasPDAs,cellularphones,andsmart-phoneseveniftheydonothaveatouch-sensitivedisplayinstalled.
ACKNOWLEDGMENTS
OurresearchwaspartlysupportedbyafundfromtheMin-istryofEducation,Culture,Sports,ScienceandTechnology,Japan,underthetitle“ClusterforPromotionofScienceandTechnologyinRegionalAreas.”
7.REFERENCES
[1]Y.Ayatsuka,N.Matsushita,andJ.Rekimoto.
HyperPalette:aHybridComputingEnvironmentforSmallComputingDevices.In(InteractivePoster)inCHI2000ExtendedAbstracts,pages133–134,Apr.2000.
[2]Develant.com-Cross-platformToolsforMobileGame
DevelopmentforPalm,SymbianandWindowsMobile.http://www.gapidraw.com/.
[3]D.F¨allman,A.Lund,andM.Wiberg.ScrollPad:
TangibleScrollingWithMobileDevices.In
ProceedingsofHawaiiInternationalConferenceonSystemSciences(HICSS’37),Jan.2004.
[4]G.W.Fitzmaurice,S.Zhai,andM.H.Chignell.
VirtualRealityforPalmtopComputers.ACM
TransactionsonInformationSystems,11(3):197–218,July1993.
[5]S.GreenbergandC.Fitchett.Phidgets:Easy
DevelopmentofPhysicalInterfacesthroughPhysicalWidgets.InProceedingsofUIST2001,pages209–218,Nov.2001.
[6]B.L.Harrison,K.P.Fishkin,A.Gujar,C.Mochon,
andR.Want.SqueezeMe,HoldMe,TiltMe!AnExplorationofManipulativeUserInterfaces.InProceedingsofCHI’98,pages17–24,Apr.1998.[7]K.Hinckley,J.Pierce,M.Sinclair,andE.Horvitz.
SensingTechniquesforMobileInteraction.In
ProceedingsofUIST’00,pages91–100,Nov.2000.[8]S.Hiraoka,I.Miyamoto,andK.Tomimatsu.Behind
Touch:BacksideKeyswithTactileSenseInterfaceforMobilePhones.IPSJJournal,44(11):131–138,Nov.2003.(inJapanese).
[9]J.A.Johnson.AComparisonofUserInterfacesfor
PanningonaTouch-ControlledDisplay.In
ProceedingsofCHI’95,pages218–225,May1995.B.MacKay,D.Dearman,K.Inkpen,andC.Watters.
Walk’nScroll:AComparisonofSoftware-basedNavigationTechniquesforDifferentLevelsofMobility.InProceedingsofMobileHCI’05,pages183–190,Sept.2005.
[11]I.S.MacKenzieandW.Buxton.ExtendingFitts’law
totwo-dimensionaltasks.InProceedingsofCHI’92,pages219–226,May1992.
[12]T.Masui,K.Tsukada,andI.Siio.MouseField:A
SimpleandVersatileInputDeviceforUbiquitousComputing.InUbicomp2004AdjunctProceedings(Demos),Sept.2004.
[13]N.Matsusita,Y.Ayatsuka,andJ.Rekimoto.Dual
touch:atwo-handedinterfaceforpen-basedPDAs.InProceedingsofUIST2000,pages211–212,Nov.2000.[14]J.Rekimoto.TilingOperationsforSmallScreen
Interfaces.InProceedingsofUIST’96,pages167–168,Nov.1996.
[15]I.Siio.ScrollDisplay:PointingDeviceforPalmtop
Computers.InProceedingsofAsiaPacificComputerHumanInteraction(APCHI’98),pages243–248.IEEEComputingSociety,July1998.
[16]K.Yatani,K.Tamura,M.Sugimoto,and
H.Hashizume.InformationTransferTechniquesforMobileDevicesbyTossandSwingActions.InSixthIEEEWorkshoponMobileComputingSystemsandApplications(WMSCA2004),pages144–151,Dec.2004.
[17]K.P.Yee.PeepholeDisplays:PenInteractionon
SpatiallyAwareHandheldComputers.InProceedingsofCHI2003,pages1–8,Apr.2003.
[10]
因篇幅问题不能全部显示，请点此查看更多更全内容