Model-Based System Development for Asynchronous Distance Learning

Journal of Distance Education Technologies, 1(4), 39-54, Oct-Dec 2003 39

Copyright © 2003, Idea Group Inc. Copying or distributing in print or electronic forms without writtenpermission of Idea Group Inc. is prohibited.

Model-Based System Developmentfor AsynchronousDistance Learning

Shu-Ching Chen, Florida International University, USASheng-Tun Li, National Kaohsiung First University of Science and Technology, Taiwan

Mei-Ling Shyu, University of Miami, USA

INTRODUCTION

The recent advances in multimediatechnology such as the high-speed com-munication networks, large-capacity stor-age devices, digitized media, and data com-pression technologies have drasticallychanged the way learners communicatewith their instructors and with each other,especially in distance education. With theinnovation of new network and Internet in-frastructures and the development of mul-

timedia technology, the distances perceivedby the learners have been virtually dimin-ished and distance learning has become oneof the most interesting new directions foreducation.

Multimedia information has been usedin several applications including education,computer-based training (CBT), manufac-turing, medicine, entertainment, videoconferencing, etc. In particular, distancelearning has become the mainstream ofcomputer-based training and education. For

ABSTRACTThe innovation and diversification of development in multimedia technology and networkinfrastructures have brought a significant impact to education, especially for distance learn-ing. This paper presents a model-based asynchronous distance learning system developmentthat consists of a presentation semantic model called the multimedia augmented transitionnetwork (MATN) model and an asynchronous distance learning system called the Java-basedIntegrated Asynchronous Distance Learning (JIADL) system. The MATN model is powerful inmodeling the synchronization and quality-of-service (QoS) for distance learning multimediapresentations. The JIADL system can support diverse asynchronous distance learning servicesby integrating RealPlayer and Java technology to augment the superiority of both models. Acourse sample is used to illustrate and validate the effectiveness of the system. How to use theMATN model to model the diversity requirements of a distance learning multimedia presenta-tion is also discussed. Furthermore, the initial experimental results show that our system is costeffective and has a wide range of applications.

Keywords: multimedia augmented transition network; distance learning; Java media frame-work

40 Journal of Distance Education Technologies, 1(4), 39-54, Oct-Dec 2003


example, there are 247 articles with thedistance learning subject published on IEEE/IEE Library in 1997 and 1999(IEEELibrary), and different surveys showthat there are one million students takingdistance learning classes via the Internetand predict that the number of college stu-dents enrolled in online courses will reach2.2 million by 2002 (Syed, 2001). More-over, the seamless integration of the over-whelming World Wide Web (WWW) andthe emerging Java technology further en-dorses the universal accessibility to diversedistance learning services. The former pro-vides a cross-platform consistent visualuser interface for accessing informationwhereas the latter allows the applicationcode (applets) to be downloaded over theInternet and run on any Java-compliant Webbrowsers.

In general, distance learning servicescan be delivered in three ways: synchro-nous (real-time), asynchronous (on-de-mand) and hybrid of both. Synchronous dis-tance learning systems provide live lecturecontents as in the traditional classroom.Asynchronous distance learning systemsoffer archived lectures by using Web and/or streaming technologies and try to pro-vide the most of the capabilities and expe-rience that an in-class participant can haveto a remote participant. Hybrid systemssupply complementary services to thoselisted above.

For the asynchronous systems, anearly development is to use television tobroadcast courses (Egan, 1993). The maindrawback of this type of television instruc-tion is the lack of interaction between thestudents and the instructors. Later, the Webis used to support asynchronous activities.One example is to develop the online pro-grams using Web pages to access coursematerials, announcements, and other infor-mation for a course. Another example is to

provide the online activities that includeforums, chat rooms, and emails. Also, thestudents can submit their assignmentsonline in multimedia formats and receiveonline reviews of the assignments in thesame formats from their instructors.

Maly et al. (1997) proposed a syn-chronous distance learning system. In theirsystem, a virtual classroom is developed toallow the students to have a conventionalclassroom experience through a worksta-tion since the students and the instructorsinteract with video conferencing. Such anenvironment incorporates an X-Windows-based group-collaboration system calledXTV (Abdel, 1994) so that any participantcan take control of a window to multicasthis/her inputs to the distance participant.Under this environment, those tools suchas Netscape and PowerPoint can be sharedthrough a window-sharing engine. How-ever, there are two disadvantages in suchan environment, namely speed and band-width. In other words, an extraordinary loadis put on the reliable multicast protocol andthe environment works only on high-speednetworks. Another system proposed inSaini (1999) studied how to immerse thestudents in a virtual environment that pro-vides them with feedback. In their system,the student interactions are computed andthe system reacts like an educational game.However, the issue of how to share a vir-tual world among the students is not dis-cussed.

In this paper, we present a projectthat aims at digitizing and distributing videotapes recorded in a synchronous distancelearning classroom to improve curriculums,to provide another channel for learning, andto complement synchronous/asynchronouslearning. In order to broaden the functionsand the effectiveness of such service, anumber of interactive and cooperative ser-vices are integrated by mainly applying Java



technology, which results in the paradigmof Java-based integrated asynchronous dis-tance learning (JIADL) system. Unlikemost related work in the literature, we in-tegrate RealPlayer and Java technology sothat the superiority of both models can beaugmented in the JIADL system. Thisproject is supported by the Ministry of Edu-cation, R.O.C. and implemented at NationalKaohsiung First University of Science andTechnology (NKFUST).

However, due to the variety of mediadata (such as video, audio, text, and imagedata) that needs to be transmitted and syn-chronized for multimedia presentations, adistance learning system development in-volving multimedia data becomes morecomplicated than the traditional system de-velopment. In addition, for a multimedia sys-tem to be useful, reliable, adaptable andeconomic, it must be based on a sound datamodel so that this model is a valuable com-munication tool between the system devel-opers and the users. Hence, it will be ofgreat benefit to have a presentation modelthat can represent and model the differentrequirements for multimedia data in a dis-tance learning system.

Towards these demands, a model-based distance learning system that con-sists of a presentation semantic model calledthe multimedia augmented transition net-work (MATN) model to model the distancelearning multimedia presentations for theJIADL system is proposed in this paper. AMATN has a finite set of nodes (states)connected by labeled directed arcs. A la-beled arc represents the transition function,where an arc represents an allowable tran-sition. The multimedia input strings that cancapture the temporal relationships of themedia streams and model the concurrentand optional displaying of the media streamsin a distance learning multimedia presenta-tion are used as the inputs for the MATN

model. In addition, the synchronization andquality-of-service (QoS) of each presen-tation are maintained by a condition/actiontable associated with a MATN. In the con-dition/action table, a set of conditions forthe synchronization and/or QoS is checkedfor each input symbol. Each condition hasits own set of actions. When a certain con-dition is satisfied, then the correspondingset of actions is activated.

An example course is used to illus-trate how a distance learning course canbe modeled by the MATN model in theJIADL system. The MATN model is pow-erful in modeling the synchronization andquality-of-service (QoS) for distance learn-ing multimedia presentations. In addition,our initial experimental results show thatthis JIADL system that is based on theMATN semantic model is cost effectiveand has a wide range of applications.

The remainder of the paper is as fol-lows. The next section briefly describes thekernel technologies employed in develop-ing the JIADL system.We then introducethe multimedia augmented transition net-work (MATN) model, the JIADL paradigmand its components, and how to use theMATNs to model distance learning multi-media presentations in the JIADL systemwith an example course. The multimediainput strings and the condition/action tablesused in the MATN model are also pre-sented in this section. Finally, the conclu-sion to this paper follows.

INFORMATIONTECHNOLOGIES FORJIADL

Java-Centric Computing

The Java programming language, ini-tially designed for small consumer devices,



became more prevalent when it wascoupled with the WWW for developing dis-tribution applications including distancelearning. For example, the Virtual Labora-tory at the University of Catania is devel-oped in Java to support remote users toaccess tutorials, perform simulations, andcontrol an educational industrial system(JavaJMF). Reed and Afjeh (1998) de-signed and implemented a Java Gas Tur-bine Simulator to provide an interactiveenvironment for constructing and analyz-ing gas turbine systems. De Santo andChianese (Arcelli, 1998) proposed a Dis-tributed Learning Environment SupportTool employing Java language in an intranetenvironment (Stefano, 1997). TheJavaGram application developed atClemson University allows the students toturn in their homework diagrams such asthe signal flowcharts in signal processingclasses (Patterson, 1998). The GVA sys-tem presented in Min (1998) is a Java-basedmultimedia distance education system,which provides a distance lecture environ-ment and a study environment.

Java has a number of distinct featuressuch as architecture neutrality,concurrency, distribution, user interaction,persistence, mobility, security, andinteroperability, which make it popular inmany applications. Moreover, accompaniedwith a comprehensive set of APIs, Javaoffers a potential framework for efficientlydeveloping a wide range of applications in-cluding image processing, multimedia, da-tabase management, electronic commerce,information security, telephony, embeddedsystems, to name a few. A Java industryera is expected to emerge in the near fu-ture.

Streaming Archived Lectures

To transmit the archived lecture con-

tents, the Java Media Framework (JMF)from SUN (JavaJMF) and RealPlayer fromReal Networks (RealPlayerJava) are twocommonly used approaches. JMF providesa platform-neutral framework for buildingmedia players and supports a variety ofmedia content types, including the high-qual-ity MPEG-1. It specifies a unified archi-tecture, messaging protocol and program-ming interface for playback, capture andconferencing of compressed streaming andstored timed-media including audio, video,and MIDI across all Java enabled plat-forms. On the other hand, the RealPlayerG2 supplies an efficient service for playingstreaming video (RealVideo) and audio(RealAudio) objects. Choosing MPEG orRealVideo for delivering archived lecturesis a tradeoff between network infrastruc-tures and constraints on efficient compres-sion. Recently, RealPlayer for Java, a wrap-per around the RealSystem G2 released byReal Networks, provides a comprised so-lution that allows the ability to play G2 ob-jects via the JMF API (RealPlayerJava).By utilizing RealPlayer for Java, the goalof delivering archived streaming contentsefficiently can be achieved without loss ofthe salient features of the Java computingmodel.

A MULTIMEDIAPRESENTATION MODELFOR JIADL

Benefits of HavingSemantic Models

In any software system design, it isnecessary to have a conceptual model ofthat software system provide a high-levelview of the system once the initial analysisof the user’s requirements for the softwaresystem have been identified. Such a con-



ceptual model is a valuable communicationtool between the system designer, softwaredevelopers, and users since the involvingcomponents, functionalities, and relation-ships of the system are documented in thisconceptual model.

A data model is a symbolic or abstractrepresentation of a system, which can high-light the important facts by efficiently elimi-nating uninteresting details. Symbols, arcsand notations are used to represent func-tions, objects and relationships for a sys-tem. For a trivial and small system, a con-ceptual model may exist only in the mindof the developer responsible for establish-ing the specifications since the developerunderstands the details of each function orcomponent together with the limitations,conditions, and actions. However, this doesnot work for the non-trivial systems suchas a multimedia system.

A multimedia system may involve dif-ferent media types such as videos, audio,images, texts, graphics, and animations. Toprovide a better multimedia system, sev-eral factors need to be considered. Forexample, the temporal relationships amongmedia streams need to be modeled. Syn-chronization and QoS need to be maintainedin order to have a high-quality presenta-tion. User interactions should be providedso that the system allows two-way com-munications between the user and the sys-tem. User loops also need to be providedso that the users can continue watching thesame part of a presentation more thanonce.

In such a non-trivial system, the sys-tem to be modeled is usually complex andmany situations need to be considered andhandled. A mental model is not suitablesince the model in the developer’s mindnormally is incomplete and imprecise.Therefore, it is necessary to establish anexplicit and precisely defined data model

at an early design stage so that the systemcan be understood.

The MATN Model

The benefits of using a semanticmodel for a software system were pre-sented in the last subsection. Specifically,there are several factors that need to beconsidered for a multimedia presentationin a multimedia system. In order to modelthe complex multimedia presentation, weproposed to use the multimedia augmentedtransition network (MATN) model as theunderlying data model for the distancelearning multimedia presentation in theJIADL system.

A MATN can be represented dia-grammatically by a transition graph. TheMATN consists of a finite set of nodes(states) connected by labeled directed arcs.An arc represents allowable transition fromthe state at its tail to the state at its head,and the labeled arc represents the transi-tion function. An input string is acceptedby the grammar if there is a path of transi-tion that corresponds to the sequence ofsymbols in the string and leads from a speci-fied initial state to one of a set of specifiedfinal states. Multimedia input strings areused as the inputs for the MATN model.The multimedia input strings have the ca-pabilities to capture the temporal relation-ships of the media streams, and to modelthe concurrent and optional displaying ofthe media streams in a distance learningmultimedia presentation.

Conditions and actions in the arcs inMATNs maintain the synchronization andquality of service (QoS) of a multimediapresentation by permitting a sequence ofconditions and actions to be specified oneach arc. The conditions are to specifyvarious situations in the multimedia presen-tation. A condition is a Boolean combina-



tion of predicates involving the current in-put symbol, variable contents and the QoS.A new input symbol cannot be taken un-less the condition is evaluated to true. Moreelaborate restrictions can be imposed onthe conditions if needed. For example, ifthe communication bandwidth is not enoughto transmit all the media streams on timefor the presentation, then the action is toget the compressed version of mediastreams instead of the raw data. In thisway, synchronization can be maintainedbecause all the media streams can arriveon time. In addition, QoS can be specifiedin the conditions to maintain synchroniza-tion. The actions provide a facility for ex-plicitly building the connections among thewhole MATN. The variables are the sameas the symbolic variables in programminglanguages. They can be used in later ac-tions, perhaps on subsequent arcs. Theactions can add or change the contents ofthe variables, go to the next state, or re-place the raw media streams with the com-pressed ones, etc. If a condition is

matched, then the corresponding action isinvoked. Different actions can generatedifferent presentation sequences which aredifferent from the original sequence.

The JIADL Paradigm

The high-level architecture of theJIADL paradigm is outlined in Figure 1.The archived lecture contents are convertedinto RealVideo files by RealEncoder in ad-vance and are stored in the Coursewaredatabase. Mobile learners may downloadthe system applets from Web Server andretrieve the courseware of interest via theJDBC mechanism. The student databasekeeps records of students’ learning progressand personal registration information. RMIServer supports the collaborative learningin chat rooms. Currently, the platform forApplication Server, Web Server, and RMIServer is on NT4.0.

There are three subsystems in theJIADL system. An example course offeredby the Center of General Education at

Figure 1: The JIADL paradigm



NKFUST, Science and Technology hadbeen experimented on the JIADL Para-digm. The lecture was broadcast in livevideo from NKFUST to four neighboringuniversities via three ISDN-dedicated linesat 384K.• StudDBMS: Any students, including ones

in the classes, can log into the JIADLsystem by registering it in advance asshown in Figure 2. The StudDBMS sub-system allows learners to query andmodify personal information besides reg-istration. As shown in Figure 2, studentsneed to register into the system by giv-ing personal information such as StudID, Password, Name, SSN, and so on.Three options are provided to let stu-dents register, query, or update their per-sonal information.

• CourseDBMS: The CourseDBMS sub-system endows the lecturers or admin-istrators to manage the digitizedcourseware. The control layout of theCourseDBMS subsystem is shown inFigure 3. Student ID and Password need

to be specified in order for them to ac-cess the corresponding courses.

• CyberLearn: The CyberLearn subsystemis the kernel part of the JIADL para-digm which allows the student to browsethe static lecture notes and the corre-sponding video lecture. After studentsspecify their student ID, Password, andthe course information, they click on theOK button in Figure 4 to activate thenext window as shown in Figure 5. Thereare two options — Show_Outline andShow_Video. When the Show_Outlineoption is highlighted, the lecture noteswill be displayed. The same situation ap-plies for the Show_Video option. If bothoptions are selected, both lecture notesand video (audio) lecture will be dis-played synchronously. Therefore, thestudent can browse the static lecturenotes and the corresponding video lec-ture by simply clicking on the Play but-ton, which results in a RealPlay windowpopped out as shown in Figure 6. As wecan see in this figure, both lecture notes

Figure 2: The control layout of the JIADL StudDBMS subsystem



Figure 3: The control layout of the JIADL CourseDBMS subsystem

Figure 4: The control layout o the JIADL CyberLearn subsystem



and video lecture are displayed sinceboth Show_Outline and Show_Videobuttons are highlighted in Figure 6.

Empowered by Real Player G2, the

student is granted the capability of con-trolling voice and lecture progress. Furtherfunctionality like lecture clips can also beeasily implemented. One important supple-

Figure 5: The CyberLearn subsystem

Figure 6: The archived lecture played by RealPlayer for Java.



mentary function for mobile learner is theNotes Area (the bottom area in Figure 5),in which the learner may write down thenotes and send them via the SMTP proto-col. The capability of attaching MIME ob-jects from the local learning environmentis incorporated. From the learning theory,learners of distance learning are not ex-pected to sit still and stare at the lecturesconsistently. To improve the learning per-formance, an interactive multi-userchatroom and a collaborative real-timewhiteboard are offered, which are basedon the Java RMI-distributed computingmodel. Figure 7 depicts the layout of suchservices, where a number of vivid imagescan be added into the whiteboard.

USING MATNS TO MODELDISTANCE LEARNINGMULTIMEDIAPRESENTATIONS

A MATN consists of nodes (states)and arcs. Each state has a state name andeach arc has an arc label. Each arc label

Figure 7: The whiteboard subsystem.

represents the media streams to be dis-played in a time duration. Therefore, timeintervals can be represented by transitionnetworks. In this transition network, anew state is created whenever there is anychange of media streams in the presenta-tion. There are two situations for the changeof media streams:

1) Any media stream finishes to display;2) Any new media stream joins to display.

Figure 8 is a timeline to represent themultimedia presentation for Figure 6. In thisexample, part of the presentation is depictedfor simplicity. There are four time instants(t1 to t4) and three time durations (d1 tod3). There are three occurrences of mediastream combinations at each time duration:

1) Duration d1: V1, A1, and T1.2) Duration d2: V2, A2, and T2.3) Duration d3: V3, A3, and T3.

Figure 9 is a transition network forFigure 8. There are four states and threearcs which represent four time instants and



three time durations, respectively. Statenames are in the circles to indicate presen-tation status. State name P/ means the be-ginning of the transition network (presen-tation) and state name P/X1 denotes thestate after X1 has been read. The reasonto use Xi (i = 1, 2, 3) is for conveniencepurposes. In fact X1 can be replaced byV1

*& A1*& T1

*. State name P/X3 is the fi-nal state of the transition network to indi-cate the end of the presentation. State P/Xi represents presentation P just finishes

to display Xi and the presentation can pro-ceed without knowing the complete historyof the past.

Each arc label Xi (i = 1, 2, 3) in Fig-ure 9 is created to represent the mediastream combination for each duration asabove. For example, arc label X1 representsthat media streams V1, A1, and T1 displaytogether at duration d1. A new arc is cre-ated when new media streams V1, A1, andT1 display. A multimedia input string is usedas an input for this transition network, and

Figure 8: Timeline for the distance learning multimedia presentation as shown in Figure 6

Figure 9: Augmented transition network for multimedia presentation as shown in Figure 8



the symbol “&”‘ between media streamsindicates these two media streams are dis-played concurrently. A multimedia inputstring consists of several input symbols andeach of them represents the media streamsto be displayed at a time interval.

When an ATN is used for languageunderstanding, the input for the ATN is asentence which consists of a sequence ofwords with linear order. In a multimediapresentation, when user interactions suchas user selections and loops are allowed,we cannot use sentences as inputs for anATN. In our design, each arc in a MATNis a string containing one or more mediastreams displayed at the same time. Amedia stream is represented by a lettersubscripted by some digits. This single let-ter represents the media stream type anddigits are used to denote various mediastreams of the same media stream type.For example, T1 means a text media streamwith identification number one. A multime-dia input string consists of one or moremedia streams and is used as an input for aMATN. Multimedia input strings adopt thenotations from regular expressions. Regu-lar expressions (Kleene, 1956) are usefuldescriptors of patterns such as tokens usedin a programming language. Regular ex-pressions provide convenient ways of speci-fying a certain set of strings. In our frame-work, multimedia input strings are used torepresent the presentation sequences of thetemporal media streams. A multimedia in-put string goes from the left to right, whichcan represent the time sequence of a mul-timedia presentation as shown in [1].

Multimedia input string: (V1*&A1

*&T1*)

(V2*&A2

*&T2*)(V3

*&A3*&T3

*) (1)

In (1), the “&” between two mediastreams indicates these two media streamsare displayed concurrently. The “*” sym-

bol is used to indicate the media streamwhich can be dropped in the on-line pre-sentation. For example, V1

*&A1*&T1

* rep-resents media streams V1, A1, and T1 beingdisplayed concurrently, but each of themcan be dropped if Show_Outline and/orShow_Video buttons are not highlighted.

Table 1 shows a simple example of acondition/action table. A condition/action(transition) table in the MATN can be usedto control the synchronization and quality-of-service (QoS). The first column in thistable contains the input symbols. The sec-ond and third columns show the conditionsand the actions, respectively. When the cur-rent input symbol V1

*&A1*&T1

* is read, thecondition is to check whether the pre-speci-fied presentation duration to display V1, A1,and/or T1 is reached. If it is not, the displaycontinues. The start time (stime) is definedto be the time starting the display of V1, A1,and/or T1, and the difference between thecurrent time (ctime) and the start time isthe total display time so far. The secondcondition is met when the total display timeis equal to the pre-specified duration. Inthis case, a next input symbol V2

*&A2*&T2

* is read. Conditions three to five are usedto check whether Show_Outline and/orShow_Video buttons (as shown in Figure5) are highlighted or not. The correspond-ing actions will be performed for each con-dition. Synchronization can be maintainedby specifying the detailed control in theactions. For simplicity, details are skippedin this paper. Similarly, the process contin-ues until the final state is reached.

PERFORMANCEEVALUATION

Table 2 shows the performanceevaluation of streaming archived lecturesin terms of the packet statistics of streams.



Input Symbols Condition Action

If ctime-stime(X1) < D(X1)

Continue

If ctime-stime(X1) ≥ D(X1)

Get_Symbol and Next_State

If Show_Outline = true & Show_Video = false

Display T1

If Show_Outline = false & Show_Video = true

Display V1 and A1

X1

(V1

*&A1*&T1

*)

If Show_Outline = true & Show_Video = true

Display V1, A1, and T1


Continue




Display T2


Display V2 and A2

X2

(V2

*&A2*&T2

*)




Continue




Display T3


Display V3 and A3

X3

(V3

*&A3*&T3

*)



Table 1: Condition/action table: D(Xi) is the pre-specified presentation duration for Xi (i =1, 2, 3). Display procedure is to display the media streams. Get_Symbol is a procedure toread the next input symbol of multimedia input string. Next_State is a procedure toadvance to the next state in MATN.



In this sample, the codec value of audiopackets is 5 Kbps whereas the value ofvideo packets achieves 15 frames per sec-ond. Both demonstrate an acceptable qual-ity for the learners. Each archived lectureis encoded by RealEncoder at the clip band-width of 56kbps, of which audio and videoaccount for 5,000bps and 51,000bps, re-spectively, as shown in the table. The av-erage bandwidth for rendering the stream-ing lectures at the client side is 5,039bpsfor audio and 51,026bps for video; both arehigher than the clip bandwidth of encod-ing. This guarantees the rendering qualitysmoothly for the streaming lectures. Thetotal packets consumed in the sample lec-ture in Figure 6 for audio and video clipsare 510 and 40,685, respectively, and thereare no packets lost attributed to TCP thatis used as the underlying transport protocolfor the solution of RealPlayer for Java. Thisfurther ensures the quality of presentation,although the efficiency drops slightly. Withthe smoothly rendering performance, thesupplementary lecture notes in HTML, andarchitecture-neutral learning platform byJava, the learners can benefit from the flex-ibility, efficiency, and timeliness supportedby JIADL.

CONCLUSIONS

In this paper, we presented theJIADL system which is based on the mul-

timedia augmented transition network(MATN) model for supporting asynchro-nous distance learning. The inputs for theMATN model are modeled by the multi-media input strings. The multimedia inputstrings have the capabilities to capture thetemporal relationships of the media streams,and to model the concurrent and optionaldisplaying of the media streams in a dis-tance learning multimedia presentation. Inaddition, the synchronization and quality-of-service (QoS) of each distance learningmultimedia presentation are maintained bya condition/action table in a MATN.

As far as the references surveyed,our work presented is the pioneering studyin the literature towards integratingRealPlay G2 and Java JMF in the distancelearning application. The experimental re-sults demonstrate that the paradigm pro-posed is efficient and cost effective. In ad-dition to supporting asynchronous distancelearning, the paradigm has a variety of po-tential value-added applications in culturalheritage multimedia applications, market-ing for electronic commerce, multimediadigital libraries, and lifelong learning. Forexample, The National Palace Museum ofR.O.C. had published a series of videotapes, with a preview function being sim-ply provided the JIADL paradigm.

Table 2. Performance evaluation of streaming archived lectures

CodecClip

bandwidth(bps)

Averagebandwidth

(bps)Protocol Total

packetsLost

packets

Audio 5(Kbps) 5,000 5,039 TCP 510 0

Video

15(Frame

persecond)

51,000 51,026 TCP 40,685 0



ACKNOWLEDGMENT

For Shu-Ching Chen, this research wassupported in part by NSF CDA-9711582and NSF EIA-0220562.

REFERENCES

IEEELibrary. The IEEE/IEE Elec-tronic Library Online, http://www.ieee.org/ielonline.

JavaJMF. Java JMF AFI, http://www.javasoft.com/products/java-media/jmf/index.html.

RealPlayerJava. RealPlayer forJava, http://www.real.com/.

Abdel-Wahad, H., & Jeffay, K.(1994). Issues, Problems, and Solutions inSharing x Client on Displays. Journal ofInternet-Working Research and Experi-ence, 5(1), 1-15.

Arcelli, F., Santo, M. D., & Chianese,A. (1998). Client-Server Architecture forDistributed Learning Environments: A Pro-posal. In Proc. of the 31st Hawaii Inter-national Conference on System Sciences,395-403.

Maly, K. et al. (1997). Interactive dis-tance Learning over Internets. IEEEInternet Computing, 1(1), 60-71.

Egan, M. et al. (1993). Learner’sPerceptions of Instructional Delivery Sys-tems: Conventional and Television. TheAmerican Journal Distance Education, 2(4),47-55.

Kleene, S. (1956). Representation of

Events in Nerve Nets and Finite Au-tomata, Automata Studies. Princeton Uni-versity Press.

Min, D., Choi, E., Han. Y.-T., Hwang,D., & Jung, J.-H. (1998). Distributed Mul-timedia Conferencing System for DistanceLearning. In Proceedings of MultimediaSoftware Engineering, 88-95.

Patterson, E. K., Wu, D., & Gowdy,J. N. (1998). Multi-Platform CBI ToolsUsing Linux and Java-Based Solutions forDistance Learning. In Proceedings ofIEEE International Conference onAcoustics, Speech and Signal Process-ing, 1909-1912.

Reed, J. A., & Afjeh, A. A. (1998).Developing Interactive Educational Engi-neering Software for the World Wide Webwith Java. Computers Education, 30(3/4), 183-194.

Saini-Eidukat, B., Schwert, D. P., &Slator, B. M. (1999). Designing, Building,and Assessing a Virtual World for ScienceEducation. In Proceedings of Int’l Conf.Computers and Their Applications, 59-65.

Stefano, A. D., Fazzino, F., Bello, L.L., and Mirabella, O. (1997). Virtual Lab:a Java Application for Distance Learning.In 6th International Conference onEmerging Technologies and FactoryAutomation Proceedings, 93-98.

Syed, M. R. (2001). Diminishing theDistance in Distance Education. IEEEMultimedia, 8(3), 18-20.

Shu-Ching Chen received his Ph.D. from the School of Electrical and Computer Engineeringat Purdue University, West Lafayette, Indiana, USA in December 1998. He also received Mas-ters degrees in Computer Science, Electrical Engineering, and Civil Engineering from PurdueUniversity, USA. He has been an Assistant Professor in the School of Computer Science, FloridaInternational University since August 1999. He is currently the Director of the DistributedMultimedia Information System Laboratory and the Associate Director of Center for Advanced



Distributed System Engineering (CADSE). Dr. Chen’s research interests include distributedmultimedia database systems and information systems, data mining, databases, and multime-dia communications and networking.

Sheng-Tun Li is a Professor in the Department of Information Management at NationalKaohsiung First University of Science and Technology, Taiwan, R.O.C. He received his Ph.D. inComputer Science from University of Houston in 1995. His current research interests includeJava multimedia systems, knowledge management, and data mining.Dr. Li is a member of the IEEE/CS and ACM.

Mei-Ling Shyu has been an Assistant Professor in the Department of Electrical and ComputerEngineering, University of Miami since January 2000. She received her Ph.D. from the Schoolof Electrical and Computer Engineering, Purdue University, West Lafayette, Indiana, USA in1999. She also received her M.S. in Computer Science, M.S. in Electrical Engineering, andM.S. in Restaurant, Hotel, Institutional, and Tourism Management from Purdue University,U.S.A. in 1992, 1995, and 1997, respectively. Her research interests include multimedia net-working, wireless communications and networking, data mining, multimedia database sys-tems, multimedia information systems, and database systems.

Documents

Model-Based System Development for Asynchronous Distance Learning