notes on laptop computers or Personal Digital Assi stants. (PDAs). These types .... However, pro ducing a reasonable quality webcast requires a grea t deal more effort. First, a ... cheaper solutions are on the way (e.g., NCast. - www.ncast.com ...
Deploying an Infrastructure for TechnologicallyEnh
anced Learning
SamiRollins andKevinAlmeroth Departmentof Computer Science University oCalifornia f SantaBarbara,CA,USA 93106-5110 {srollins,almeroth}@cs.ucsb.edu
Abstract: A number of university campuseshave undertakenthe developmentof digital classroomsthatenable presentation odigital f media anddigi tallecture recording. Deploying the infrastructure for daigitalclassroom isdiffi cultatbestevenfor taechnically savvy person. Aspeople from many disciplines become interestedi nbuilding similar digitalclassroom spaces,there inaseedtoproduce useful a setof designandimplementation guidelinesto reduce the projectrisk and steepnessof the deploy mentcurve. The goalof this paper isto report on the experienceswe havehadidneploying the UCSB digitalclassroom. The two maincontributionsof this paper are:(1)a phased deploymentmodel;and(2) daiscussion of how the proposedtechnology enablesnew educational modelsand techniques.
Introduction Advances in technology coupledwith increasedfamil iaritywith technicaltools havepavedthewayfor new paradigms in teaching andlearning. Instructor are s nowusingmediasuch aPowerPoint s slides and digital videoitnheir lectures. Students can takedigital notes on laptopcomputers or PersonalDigitalAssi stants (PDAs). Thesetypes of technologies allow students andinstructors tocommunicatedigitally across ti meand space. However,whilethesetools arereadily avai lable,using them in coherent a manner is still ca hallenge. A number of university campuses haveundertaken theg oalofdeveloping digitalclassrooms thatenable presentation oinformation f using cutting edgemult imediatools aswellthecapability todigitally re cordan accountof theclassroom activity. Theaccountcan beusedirnealtimetoenabledistancelearning or realtime collaboration,or can baerchivedandreviewedaat later time. Anumber of universities havedeployeddigitalcla ssroomsfor both teaching andresearchonew learningmethodologies andtools. Oneothe f earli estexperimentswith thiskindotechnology f was th A e T&T Learning/Teaching Theater attheUniversity oMary f land(Schneiderman eal. t95). More recentexample s include405Soda aUC t Berkeley (Wu,Swan,&Rowe9 9)andGeorgiaTech's eClass (Abowd99). Whilethe research thathas comeoutof theseprojects has fo cusedlargely ounser-levelissues,thepieceoth f peuzzlethat hasyettobseolvedithe s question owhat f functi onality theseclassroom spaces shouldsupport,and more importantly,how thatcan baechieved. Without ua sefulmodeltodraw from,thereian senormous lear ning curveinvolvedidneterminingfirst,whatfunctiona lity classroom a shouldsupport,andsecond,what technology exists torealizethedesign. Ahugenu mber of tradeoffsneedtobceonsidered. Itis dif ficultatbest for taechnically savvy persontoundertakethecha llengeodeploying f classroom. a As peoplefrom ac ross disciplines becomeinterestedibnuildingsimilar d igitalclassroom spaces,thereinaseedtoproduce uasefulset of design andimplementationguidelines for easeof deployment. Thegoalof this paper is toreporton theexperie nceswheavehaddeploying digital a classroom. By drawingfrom our experiences,futureclassroom arch itects can reduceproject risk aw s ellas thesteep ness of the deploymentcurve. Weidentify four classroom funct ions,andsuggestthat calassroom shouldbdeeploy edin four phases corresponding tothosefunctions. In Juneo2000 f wteook otnhechallengeodeploying f daigital classroom atUC SantaBarbara. Todate,wehavesp entapproximately $70,000 oonur classroom setupbr oken down aroughly s $14,000for phase1$42,000 , for ph ase2and , $12,000for phase To 3d. ate,our phase 4 deploymenthas leveragedtechnology purchaseditnh perior phases. Itis difficultif notimpossible todeploy a fully functionaldigitalclassroom infrastructureb eforetesting ousing r Therefore, it. itis impera tivetosupport incrementaldevelopment,deployment,use,evaluatio n,andmodification.
PresentationFacilities Thefirstphaseoclassroom f infrastructuredeploym entfocuses on providing technology toallow an instructor togive multimedia a presentation idan igitalclassroom. Itis impossibletodevelopain nfrastructure thatwillaccommodateevery lecturer or class ever heldidigital an classroom. Someprofessorswill use PowerPointslideswhileothers prefer tousetransp arencieswhilestillothers stick tothestandardc halkboard method. In addition,instructors using digitalpre sentationmediamay require variety a osoftware. f Managing a few pieces ofsoftwareitractable, s however sayst em tomanagelots of softwareinsot. Fortunately, laarge percentageocases f can baeccommodatedwith stan a dardcollection ohardware f andsoftware. Minimall y,a classroom shouldinclude data a projector toshow P owerPointslides or other computer videoiandditio tno providing aInnternetconnection for paresentation laptopodesktop r machine. Selecting presentation a computer anddataprojecto for r purchaserequiressomethoughtaboutthe specificclassroom andthecompletefunctionality t hatwilleventually bseupportedbtyheclassroom. We identify threemain concerns thatneedtobaeddres sedwhenselecting equipment. Thefirstis compatibility. A major concern iw s hether or noteach pieceoequip f mentwillbecompatiblewith theremaining infrastr ucture. For example,if theclassroom willeventually have raoom controlsystem tocontrolvarious components (e.g., power on/off,inputdeviceswitching,etc),does th deataprojector supportthattypeocontrol? f An additional concernwith dataprojectors is how to installourmountthem.Oneoption ito ssimply purchasea mediacart where projector a can bsetored. However, m a edia cartis not sacalableopermanent r solution. The preferred solution ito smountthedataprojectors intothec eiling. Thisrequires thepurchaseocafeiling-mo untkitfor each projector. Additionalconcerns includeprovid ing power a sourceaw s ellas ensuringthatthecei ling is high enough tomounttheprojectors outof theway of sightfor students andother equipment. Oncea data projector and presentation computerhavebeenselec ted,thenextconcern is connectingthem together. The standardsolution ito srun VGA a cablefrom theco mputer totheprojector. However,in digital a cla ssroom the distancemightbetoogreatthus thequality othe f videosignaldegrades. Thesolution ito spurchas signal ae interface -daevicetoboost caomputer videosignalsuch t hatitcan travelgreater distances. Input Sources
Display of other media Display of slides
Production Control
Output Sources
Display of remote site Instructor Camera Projector 1
Projector 2 Projector 3 Display
Audience Camera VCR Video Matrix Switch Computer 1
Archive
Computer 2 Encode Remote A/V Streams
. . .
Floor Control
Webcast
Remote Collaboration Site
Figure1: TheUCSBDigitalClassroom.
TheUCSBdigitalclassroom shown iF n igure1has th on standardprojection screens,twopresentation la room,andonepresentation desktopwith signalinte VCRfor showing standardVHSvideotapes. Each comp can buesedfor showing DVDs,presenting PowerPoint bring their own laptopcomputerswith specializedh provide. Finally,studentsmay connecttotheInte wirelessnetwork. Wearecurrently itnheprocess of identifying othe extensions includeintegratingexisting technology upgrading thedisplay technology tobm e oresophist singlewall-sizeddisplay (Foxeal. t00).
Figure2: TheUCSBdigitalclassroom infrastructure.
reeceilingmounteddataprojectors thatproject ptopcomputerswith signalinterfaces atthefront of the rfaceathe t back othe f room. In addition,wepro videa uter has D a VDplayer andaInnternetconnection an d slides,and/orwebbrowsing. In addition,speaker m s ay ardwareand/or softwareandusethedataprojectors we rnetusingtheir own laptops or PDAs though 10Mbp a s extensions r totheinfrastructure. Short-term such adasocumentcamera. Longer-term extensions icated. Insteadothree f separatescreenswceould
include havea
A WebcastingStudio Oncethebasicinfrastructureiin splacetosuppor lectures t in classroom, a thenextstepiexpandi s ng theinfrastructuretoincludesupportfor webcastin Awebcasting g. environmentcaptures audioandvid eofeeds generateditnheclassroom andsends them,potentia lly accompaniedboythermaterialssuch aslides, s over the Internetto remote a location. Therearetwopiece involved s inmaking this happen. Thefirstis to provide supportfor capturing videoandaudioothe f presen tation. Theremoteaudienceshouldbaebletosee andhear theinstructor. Second,multimediamaterial(e.g., slides)presentedtothelocalaudienceshouldals obe presentedtotheremoteaudience. Theminimalrequirementsfor w a ebcasting environme ntareonecamera, m a icrophonefor the instructor,andaenncoding computer. However,pro ducing reasonable a quality webcastrequires garea deal t moreeffort. First, sainglecameracan bleimitin igyou f hopetocaptureallof theactivity thato ccurs in a classroom. Tocaptureallof theclassroom activit iyncluding instructor andstudentmovement,multip le camerasmustbemountedivnarious locations in the room. In addition,capturing slides or w a ebpage can be donesimply bfyocusing camera a otnheprojection screen. However,capturing thecomputer videofeed straightfrom thecomputer can produce higher a qua lity image. Determiningwhich streams areencoded and webcastatany given timeadds naon-trivialbitof complexity tothesystem. Generally,a producer mustbe availabletomanually selectthestream forwebcast A producer . is generally student a or staffmembe who r produces thewebcastby controlling theencoding to ols andselecting theappropriatevideoandaudios treams. Managingmultiplestreams simultaneously introduce hasostof complexities. First,devices such as cameras haveaanssociated controlinterfaceandmay bceontrolled(e.g.,zoomed)fro m raemotecontrolor computer interface. However,whennumerous,hetero geneous devices areinstallediroom, an itbecomes difficulttocontrolallof theindividualelements Rather . than havingmultipleinterfaces such acomputers s and remotecontrols,itis desirabletosupport saingl e,integratedinterfacethatsupports controlofma ny oall rof the devices in theroom (Yu eal. t01). Also,itis un likely thatany hardwareconfigurationwouldsuppor encoding t of allpossiblestreams simultaneously.Thegeneral protocol is to select saubsetof allavailablestreamsfor encoding. In order toaccomplish this,theinfrast ructuremustinclude video a matrixswitch. Avide omatrix switch idasevicethattakes asinput saetof vid eosignals andallows routing othe f videosignals tooneor moreothe f switch outputchannels. By routing all videothrough video a switch,thearchitecturebec omes muchmoremodular. Audiosignalcaptureposesmany of thesameproblems encounteredbvyideostream capture. If an infrastructuresupports only sing a lemicrophoneusedbtyheinstructor,thesignalca bndeirectly connectedtsound oa card. However,thismodelbeg ins tobreak down relatively quickly. Capturing ot her audiosources such aaudience s discussion othe r au diotrack from V a HSvideoiimperative. s Theulti mate solution ito sinstall parofessionalquality audio system thatis capableomixing f audiosignalsfro m the instructor,microphones placedtocaptureubiquitou audience s discussion,audiostreamsfrom remotesi tes,and other sources such avideo. s Finally,videoformat compatibilityicasoncern. Videocapturehardware generally expects caompositevideosignal. Theref ore,the computer videosignalsentfrom paresentation computer cannotbedirectly encodedapart s of vai deostream. Thesolution ito susea scanconverter to convertthehigh-quality computer videosignaltoa compositevideosignal. Developing ainnfrastructuretomanageandselectm ultiplemediastreams in various formats is extremely complex andrequiresmuch thought. Howev er,oncethearchitecturetocapture selected a str eam is in place,thenextstepito sdeterminehow toenco detheaudioandvideostreams into format a thatc an beasily distributedandviewedbryemoteparticipants. The rearetwoprimary concerns thatneedtobaeddress ed. The firstis expense. Hardware-basedencodingsolutions provideeffici ent,high-quality encoding. However,while cheaper solutions areotnheway (e.g.,NCast www.ncast.com andVBrick www.vbrick.com), current solutions can bveery expensive. Software-baseden coding solutions can alsobexpensive,but raange of lower-costsolutions existaswell. While encodingformats such aMPEG-4 s may seem tobtehehighest quality solution,thechosen encoding formatshould have widely a available,cross-platform decoder/vi ewer. Ideally,studentswouldhavetheviewing tools alre ady installedotnheir desktopthus avoidinghaving to downloadopurchase r them. Themostcommon tools c urrently otnhemarketareRealPlayer andWindows MediaPlayer both owhich f havefreely availablean deasy-to-installviewers andencoders. TheUCSBdigitalclassroom infrastructuresupports webcastof sainglevideostream selectedfrom a setcameras or other videoinput. In addition,we haveimplementedtheaudiosetupsuggestedbtyheA ccess Gridspecification ( www.accessgrid.org)which includes naumber ofmicrophones topickupambientsound
1 and high a quality soundmixer The . heartof theUCSBclassroom infrastructurei 1as2input,8outputvideo matrix switch. Theswitchsupports both composite andcomputer video. Thefirsttwoswitch inputs (s ee Figure2are ) thecompositevideosignals generated from theclassroom camerasmountedtocaptureboth instructor andaudienceviews. In addition,compos itevideofrom theVCRis alsoroutedthrough thes witch. Allcomputervideosources from thepresentationma chines,aswellas remotesources,arealsoinputs tothe switch. However,in order to captureandencodean oythese f feeds,thefeedmustberoutedthrough a scan converter andconvertedfrom computer videotocomp ositevideo. Theresulting stream is then fedbac iknto theswitch andmay bseelectedfor encoding. Thes treamsselectedfor encoding arefedinto video a c apture cardinstalledostandard an PIn C. addition,the mixedaudiostream isfedintothesoundcardoth f P eThe C. audioandvideostreams arethen encodedusing eith erRealMediaformator Windows Mediaformatand webcastto remote a audience.
Remote Collaboration Theone-way distancelearning scenariosupportedby w a ebcasting studiodoes notcapturethetrue learning experience. For distancelearning totrul byeffective,wehavetoenableremotestudents t oask questions,participateidniscussion,andotherwise appear to baethe t localsite. This requires two pieces. First, theremotesitemusthavefacilities similar totha of tthelocalsite. Second,thetechnology tocom municatein realtimebetween thetwositesmustbeipnlace. Ideally, raemotesitewouldbaeenxactreplicaof theprimary site. In reality though,theinfrastr ucture of raemote siteigsenerally subset a of theinfrastructurede ployedapatrimary site. Minimally, raemotesite mustinclude camera, a m a icrophone,an encodingma chine,and decoding a machine. This couldtakethe form of satandardwebcam andmicrophoneconnectedtoa student's homePCwherethePC is both theencoding and decodingmachine. However,if raemotesiteides s ignedtosupportmultiplestudents (e.g.,an extens ion campus) m a orecomplex infrastructureinecessary. s For example,displaying videoothe f instructor o PC an screen iprobably s notsufficient. In addition,th ceamerashouldbaebletocaptureaanudiencelarg er than a singleperson. Fortunately,theproblems encounter edwhendeveloping theremotesiteinfrastructurea rethe sameproblems encounteredwhen deploying theprimary classroom an dthuswcean apply thesamesolutions. Whiledeploying theremotesiteinfrastructureits elf is relatively straightforwardandfollows direc tly from theexperienceswheavealready described,dep loying communication a layer on topothe f infrastr ucture ismoredifficult. Therearetwomain topics tob aeddressed. Thefirstis realtimeencoding. Our first inclinationwas tousethesamesoftwareencoding s olutionsfor realtimecommunication thatweusedfo oner waywebcasting. However,off-the-shelf encodingso ftwaresuch aRealMedia s andWindowsMediaintroduc e intolerabledelaysfrom 7to15seconds one-way due tobuffering requirements. Aswithmany othe f pr oblems wehaveencountered,extremely expensiveencoding s olutions exist. But,deploying thesesolutionswit hout knowingwhether or notthey aregoing towork ias risky venture.Thealternatesolution ito suseenc oding softwaredesignedfor videoconferencing such as vicoMicrosoft r NetMeeting. Whileoff-the-shelfvide o conferencing softwareigsenerally easy touse,qua lity isacrificed s tomeetrealtimerequirements. Also,in addition towatchingvideogeneratedapatrimary s ite,students at raemoteshouldbaeble toask que stions (Malpani&Rowe97)andaccess sharedcomponents su ch aw s hiteboards. Additionally,as thenumber of remotesitesgrows,itbecomes necessary tomanage thesites sothatonly oneremotesiteiasking s a question or sourcingvideoagativen time. Someothese f prob lemsmay bseolvedwith standardvideoconferencing software. However,therequirements aredifferent for any given infrastructureanddifferences requir e specializedsolutions. Wehavedeployed test a remotesitewcealla kiosk. Thekiosk has one camera, sainglemicrophone (tobpeassedfrom participanttoparticipant), sa ingleencodingmachine,andtwodecodinglaptopcom puters connectedtodataprojectors for display (seeFigur 3eWe ).useMicrosoftNetMeeting tocommunicateb etween thesites. Theprimary classroom sends slides (usi ng theNetMeeting screensharing facility), vaideo feedothe f speaker,andaanudiostreamfrom thespeakermicro phonetothekiosk. Thekiosksends sainglevideo stream and single a audiostream back totheprimary class room whereiis displayed t oside an mountedprojec tion screen. Figure4showstheflow ofstreams between sites. In building thekiosk,wehavemade numb a er of simplifying assumptions. First,toavoidtheprobl ems of floor controlandremotestream selection,w aessume 1 The Access Gridian sinitiativetoenableresearc meetings over theInternetthrough an
always on
lhabs anduniversities toconductlargescale,dis infrastructure.
tributed
thatthereareonly twosites participating aany t either siteigenerated s from thesinglemicrophone signal. Extensions toour infrastructurewouldinc
given time. Second,weassumethattheonly audio availableathat t sitethuswdeonothavetomix ludeenablingmultiplesites toparticipateaany t
stream for theaudio given time.
Display of Slides Video of Speaker
Functions Multimedia Presentations
Capture/Encoding Station
Microsoft NetMeeting Slides
Webcasting Lectures Remote Collaboration
Classroom
Camera
Archival/Retrieval
Camera
Selected Production Stream
On-Demand User
Kiosk
Windows Media Encoding
Archive
Live User
Audience
Figure3:
Remotesiteinfrastructure.
Figure4: Classroom andremotesite
communication.
Lecture Replay Oneothe f primary advantages of recording lectur a oecourse r ithat s itmay breeviewedafter thefa ct. Studentsmay reviewmaterialattheendocafourse or beforeaenxamination. In addition,lectures g iven by guestspeakers or experts in their fieldmay baerc hivedandwatchedbsytudents foryears tocome(Ts ichritzis 99). Thereare number a of issues involvedwith rec ording lecturesfor replay. However,mostof thed ifficultly lies in providingmorefunctionality over straightf orward,sequentialplayback. Thegoalof theinfra structureis simply toprovide content a basethatmay baeccess edandusedtoresearch newmethods of access. The challengeito srecordandencodethecontentsuch thatitmay baeccessedlater in differentwaysusi ng variety a of tools. Supportof VCR-styleinteractivity aw se llas integration omultiple f mediatypesprovides m a ore effectivereplay experience. Therearetwomain issues in thedeploymentof an i nfrastructurefor lecture replay. Thefirstis the deploymentof a media server. Anhour of lecturecan b5e00Mbytes ormoredepe nding otnheencoding scheme. Therefore,thefirstconcern ito sdeploy m a ediaserverwith enough disk spacetoholdther ecorded lectures. Thesecondissueito sdeterminewhich encodingstandardtouse. Thesimplestsolution ito s simultaneously savethestream already being encode dfor webcast. However,itmay bdeesirabletosup port postprocessing othe f stream such aincluding s sync hronization between videoandslides (Mukopadhyay 99). TheUCSBdigitalclassroom primarily focuses on rep lay othe f webcaststream. Duringwebcast,the encodedRealor Windows Mediastream is savedtoa fileandmay bsetreamedfrom theserver later. We are stillin theearly stages of generating content a b aseandthushavenotyetdeployed server a tosupp ortlarge amounts of data. Wehavealsobuilt taooltosupp ortsynchronization between noteswritten bpyartic ipants and thevideoothe f lecture.
PotentialImpact onEducation Multimediaitnheclassroom presents naumber of op portunities for students andeducatorsalike. Firs t, using technology,learning canbecome more a intera ctiveprocess. Teachers can use variety a omedia f toteach students in new anddifferentwayswhilestudents c an usetechnology such alaptop s computers or PDAs to shareinformation andcommunicatewith oneanother. Moreover,usingwebcasting andremotecollaborati on facilities,wecan removethephysicalbarrier impo sedbay classroom environment. Enrolledstudents can“web commute”rather thanmiss lectures,andstudentswh omay haveotherwisebeen unabletotake class a at all havemoreflexibility tochoosetoattendlecturef rom theirhome. Additionally,thenumber of people whocan fitin room a or bephysically presentnolonger li mits audiencesize. For example,if an expertspea ker visits a university,thenumber of peopleinterestediantte ndingher lecturemay exceedthecapacity othe f la rgest
lecturehallon campus. Finally,aswebcasting ena an expert,lecturearchivalandreplay canmakethe watch andlearn from experts in theirfield
bles laimitlessnumber of peopletowatch lectur a samelecturepersistent. Studentsforyears toco
geiven by mecan
Conclusion Deploying theinfrastructurefor daigitalclassroo m is laongandoften tedious process. In theory, it involves technicalstaff,facilities staff,aswell as researchers. In thepastyear,wehavebrought theUCSB digitalclassroom onlinetosupportthefour functi ons definedboyurmodeltovarying degrees of comp letion. Whiletheprocesshas been slower thanwoeriginall aynticipated,delays in deploymentcan baettribut edtolack of staff aswellas tofactors such aback-ordered s equipmentand equipmentincompatibility. However,weare pleasedwith theresulting infrastructureanditsu sesofar. To date,wehaveusedtheclassroom for threediff erenttypesoevents: f (1)wehavehostedand recorded standard a graduatecourse;(2)wehaveco nducted graduate a studentseminarwith participant s distributedbetween theclassroom andour remoteki osk;and(3)wewebcasted talk a given bNobel ay l aureate toaenlementary schoolclassroom locatedinear an by town. TheNobellaureate's talkwas alsoviewed from other locations in theUS,aswellas in theUK. O verall,ourexperienceshavebeen successful. Whil w e aere working towardautomating theprocess of setupand configuration aw s ellas lectureproduction,our infrastructuresupports nearly allof theintended functionality. Throughouttheprocess of designing andimplementi ng ourclassroom model,wehaveidentifieda number of considerations thatmay notbeimmediatel oybvious tothedesigner or implementer. These considerations rangefrom high-leveldecisionssuch assupportedfunctionality tolow levelchoices su ch as requiredequipment. Wewouldliketoacknowledgem any helpfuldiscussionswithLarry Roweandother classroom architects thathavehelpeduto sdetermi nethecommon properties ofmostdigitalclassrooms We . believethatmostdigitalclassroom implementations supportsimilar functionality. Thus,itis unnece ssary for each design team tostartfromgroundzero. As cam puses aroundtheworldbegin toembracetechnology in their curriculums,itis essentialtobaebletqou ickly andeasily deploy technologically enhancedme eting spaces.
References Shneiderman,B.,Alavi,M.,Norman,K.,& Borkowski Classroom. Communications oftheACM 38 ,(11), 19-24.
E. ,(1995).Windows of Opportunity itnheElectron
Wu,D.,Swan,A.,& Rowe,L.(1999)An InternetMBo ComputingandNetworking 1999, , San Jose,CA,USA.
neBroadcastManagementSystem.
Abowd,G.(1999).Classroom 2000:An Experimentwit Environment. IBMSystems Journal 38 ,(4),508-530.
thheInstrumentation oL af ivingEducational
Multimedia
Fox,A.,Johanson,B.,Hanrahan,P.,& Winograd,T. (2000).Integrating InformationAppliances intoan InteractiveWorkspace. IEEEComputer Graphics andApplications 20 ,(3), 54-65. Yu,T.,Wu,D.,Mayer-Patel,K.,& Rowe,L.(2001). Computing andNetworking,2001,San Jose,CA,USA. Malpani,R.,&Rowe,L.(1997).Floor Controlfor L Seattle,WA,USA.
dc:A LiveWebcastControlSystem.Multimedia
arge-ScaleMBoneSeminars.
Tsichritzis,D.(1999).Reengineering theUniversit
y. Communications ofthe ACM
Mukopadhyay,S.,& Smith,B.(1999). PassiveCaptur Orlando,FL,USA.
aendStructuring oLectures. f
ACMMultimedia 1997, ,
42 ,(6), 93-100. ACMMultimedia 1999, ,
ic
