1536-1268/02/$17.00 © 2002 IEEE PERVASIVEcomputing 49

Wireless User InterfaceComponents forPersonal Area Networks

Although the compelling vision of a“wearable computer” is over 50 yearsold,1 it is still only sporadically real-ized. Some obstacles are transient lim-its of current technology (weight, cost,

and so on) and can be overcome eventually, but otherproblems are more fundamental and structural. Inparticular, requiring that wearable computer com-ponents be connected via wires has imposed severeconstraints on wearable design and acceptance. The

recent emergence of cheap, low-power hardware and viable wire-less transceivers lets us removethese constraints. Wearable sys-tems composed of wireless mod-ules, known as Wireless PersonalArea Networks,2 are arriving.

Given all the aspects of a wear-able system, this seems a rather small change. How-ever, it can have a significant effect on the range ofPAN designs. In describing our WPAN, the SpartanBodyNet (SBN), we demonstrate aspects of this newdesign landscape that let us introduce two new wear-able components. The first is TiltType, a wrist-mounted user interface device,3 and the second arephicons, transferable physical icons that conveysemantic information.4 These components havebeen employed in other user interfaces, but not in aPAN, wired or not.

Why wireless?When a wearable system is wired, the wires limit

the number of components that can be pluggedtogether, the distances between the components, andthe number of layers of clothing that can be crossed.Without wires, these constraints disappear. In 1993,Olin Shivers wrote a visionary paper proposing awireless wearable system that he called “BodyNet,”5

whose benefits were:

• Size and location. When the number of compo-nents increases, the responsibility of each decreases.Instead of a one-size-fits-all set of components,each component can tailor itself to one particulartask and optimize its shape, size, and locationaccordingly. For example, medical monitoringcomponents such as heart monitors can be locatedclose to the appropriate area of the body without“dragging” the rest of the system with them, andthe shape need only accommodate the monitor-ing function.

• Customization. Because components are special-ized and removable, they can also be tailored—users can hot swap user interface components tofit their particular needs or preferences. For exam-ple, a user’s WPAN might have several differentfeedback modules (tactile, visual, or auditory) fordifferent contexts (as Rebecca Hansson and hercolleagues suggest6). The user can tailor each ofthose modules to the user’s preferences.

• Consolidation. Components can be easily sharedamong applications. For example, there is no rea-son for both a cell phone and pager to contain alist of phone numbers—they should consolidate

Wireless wearable systems allow new user interface components. Wehighlight two—TiltType, a wrist-mounted device for text I/O, andphicons, objects whose physical appearances are metaphors for theirelectronic capabilities. Artifacts like these can create smaller, simpler,easier wearable systems.

W E A R A B L E C O M P U T I N G

Kenneth P. FishkinIntel Research Seattle

Kurt Partridge and SauravChatterjeeUniversity of Washington

their information. I/O devices such asheadphones, displays, speakers, andmicrophones could serve any applica-tion that needs them.

• Kinesthetics. Consumers generally preferwireless devices because wires can tangle,restrict movement, be tripped over, andget caught on other objects. Devices suchas WPAN-participating wristwatcheswould most likely not be accepted com-mercially if wired to other wearables.

We propose three additional benefits:unobtrusiveness, multiplicity, and trans-ferability. First, making the system smaller,wireless, and less obvious to other peoplelessens the “cyborg look” and makes thesystem less obtrusive. So, the perceived dis-tancing effect of having a wearable systemwould not come into play as often or asdeeply. Second, because we can have manymore components, we can readily supportmultiples of any particular component (forexample, multiple user interface compo-

nents). Because the components don’t usephysical connectors, they can simultane-ously interact. Third, when the compo-nents become multiple and wireless, userscan capriciously pick them up, set themdown, or hand them off, making themmore transferable.

These benefits can be synergistic andtheir combinations provide new artifactsand use scenarios. For example, later wecombine transferability, customization,and multiplicity benefits into phicons, anew type of UI artifact for WPANs. Simi-larly, by combining the location, kines-thetics, and unobtrusive benefits, we obtainTiltType, a second new type of user inter-face artifact for WPANs.

However, compared to their wired coun-terparts, WPANs do have some disadvan-tages, the first of which is less available

power. Small devices are constrained bytheir battery size, and practical wirelesstransfer of power between components isinfeasible. Many researchers are investi-gating ways to reduce power consumptionand improve power generation. There isalso an association problem.7 If two PANcomponents are near each other, are theypart of the same PAN? In a wired system,the answer is simple—if they are connectedto each other, they are. In a wireless sys-tem, the answer is no longer clear—withwhich PAN is a component associated?This is an active research area, with sev-eral solutions under investigation.8,9

Another disadvantage is that data ratesare reduced. Maximum data rates for wiresfar outstrip wireless connections. Manyapplications, however, have modest datarate requirements.

Finally, we need a wireless protocol stan-dard. Wireless protocols have many moreconstraints than their wired counterpartswith respect to range, frequency spectrum,

and so on. The Bluetooth wireless commu-nication protocol has recently emerged toaddress this. It is ideal for WPANs becauseits nominal range is only 10 meters. It con-sumes less power and has much less rangethan other standards that target wirelessLANs such as 802.11. Bluetooth is expectedto be available in over 70 percent of mobilehandsets by 2006.10 Bluetooth’s lower lay-ers have recently been adopted as the IEEE802.15.1 standard.2 For our early proto-type, we did not use Bluetooth, as it wasnot quite ready for quick prototyping, butconversion would not be difficult.

So, although there are disadvantages,they are tractable and diminishing, and thebenefits outweigh the negatives.

Spartan BodyNetWe wanted to design a small, simple,

wireless system that was easily reconfig-urable and would let us explore WPANuser interface possibilities. We have usedSBN to build a simple prototype applica-tion involving stock trading. SBN’s fourcomponents are a server, phicon, notifica-tion ring, and TiltType (see Figure 1).

The server is a 3650 iPAQ, with 32Mbytes of RAM, 16 Mbytes of ROM, andan optional 802.11b wireless LAN card(used for LAN connections, not theWPAN). The iPAQ’s screen, buttons, andstylus are not used in the SBN. The iPAQcould be replaced with a smaller device thatcontains only the processor, storage, andwireless connectivity, such as the PersonalServer.11

A phicon is a manipulated object whosephysical appearance serves as a metaphorfor the object’s electronic capabilities. Inour sample application, a phicon shapedlike a lock gives an additional level of secu-rity to the application.

The notification ring’s purpose is toattract the user’s attention. It has twoLEDs, red and yellow, which flash in dis-tinctive patterns to indicate a new event’smeaning and priority. For example, thering could be used to notify users that theyare receiving a cell-phone call in a way thatis both subtle and public. A touch on thering lets the user turn notifications off. Inour sample application, the notificationring notifies users that the system wouldlike them to enter a stock symbol.

TiltType is a wrist-mounted device fordisplaying and entering messages. Its cur-rent form is the size of a large watch.

University of California at Berkeley’s dotmotes12 (see Figure 2) allowed these com-ponents to communicate to each other. Themotes were originally designed as a mini-mal hardware device to prototype wirelessad-hoc networks, although we did not usetheir multihop capabilities for this com-munication. Each mote contains an AtmelAT90S8535 microcontroller with 512 bytesof RAM, 512 bytes of EEPROM and 8Kbytes of flash ROM storage. For I/O, eachmote has two LEDs, a light sensor, a tem-perature sensor, and an RF MonolithicsTR1000 916.5 MHz short-range radio. A3V CR2032 lithium coin cell powers the

50 PERVASIVEcomputing http://computer.org/pervasive

W E A R A B L E C O M P U T I N G

When the components become multiple and

wireless, users can capriciously pick them up, set

them down, or hand them off, making them

more transferable.

mote, which consumes 15 mW of currentnominally and about 40 mW while theradio operates. A dot mote is circular, mea-suring 2.6 cm in diameter and is 8 mm tallincluding the battery. The server and Tilt-Type connect to motes over serial links,while the functionality of the indicator ringand phicons are implemented directly in amote’s microcontroller.

Dot motes greatly simplified our imple-mentation by providing a small general-purpose radio platform that we were ableto implement quickly. Unfortunately, theradio runs at 19.2 kbps. A newer mote canrun up to 115 kbps but uses a much largerAA battery pack. Although some wearableapplications require faster speeds, 19.2kbps is adequate for our current research.

The communication protocol betweenmotes uses the mote’s standard packet for-

mat, which provides addressing and CRCchecking. Discovery, the general mecha-nism by which nodes in a communicationnetwork announce themselves, is presentlyonly partially implemented in a non-general way.

TiltTypeFigure 3 shows TiltType, a wrist-

mounted device that serves as SBN’s pri-mary user interface. 3 TiltType gives usersthe convenience of a wrist-mounted I/Odevice without the space and power con-straints that wrist-mounted computers mustendure. TiltType can display short messageswhile on the user’s wrist. To enter text, users

OCTOBER–DECEMBER 2002 PERVASIVEcomputing 51

Figure 1. Spartan BodyNet components: (a) the server, (b) a phicon, (c) the notification ring, and (d) TiltType.

Figure 2. A dot mote.

(a) (b)

(c) (d)

must remove the device from their wristbecause TiltTyping on the wrist is too fatigu-ing. In our prototype, Velcro keeps TiltTypeon the wrist, but a snap-on mechanismwould be better for a real system. The userthen uses a combination of tilts and buttonpresses to enter specific characters (see theTiltType sidebar).

Because of our desire for a cheap, small,easily prototyped display, TiltType’s dis-play is only eight characters wide and twolines deep. Although limited, we found itsufficient for delivering short messagesand asking the user simple queries. Forexample, in our sample application, usersused TiltType to enter a stock symbol. Weaugmented the normal text-entry mecha-nism with a Choice command, which letsusers use tilting to select from a range ofchoices.

PhiconsThe user interface community has

recently investigated tangible user inter-faces. Their physical appearance suggestsnatural manipulations with correspond-ingly intuitive actions, harnessing ametaphor’s power to create a more com-pelling user experience.4 A real-worldexample is a car seat adjustment controlwhose shape resembles a car seat. Whenthe seat control is pressed forward, the seatitself moves forward.

BenefitsPhicons are easy to learn and remember.

Also, because they are tied to a particularpurpose, they can be faster and less error-prone than general-purpose user interfacesfor frequently performed operations.

In addition, their physical appearance canbe easily customized. People using phiconsmight be more likely to customize how theyinteract with their information; few userscustomize their user interfaces, yet every-

body customizes what they wear and carry. WPAN benefits are particularly syner-

gistic with phicons. Because there are nophysical connectors, there is no physicallimit to the number of phicons that caninteract with a system. WPANs allow thetransfer of components, and physicalobjects such as phicons are also easily trans-ferred from person to person (althoughthey must be reassociated with the newuser’s WPAN). Although association mightpresent problems, the converse, dissocia-tion, presents opportunities because the sys-tem can notice when a device leaves aWPAN. For example, in a WPAN, anauthorization phicon (discussed later)might be hidden in a person’s clothing.When the phicon leaves the WPAN, theWPAN can be notified and can disableaccess to personal data. That mechanismwould let users use other user interfacesfound in the environment, such as largepublic displays,13 without exposing theirdata.

Phicons as an aid in authorization An authorization phicon conceptually

resembles a wireless smart card; it is builtfrom tamper-resistant hardware and pro-vides a form of authentication. When theWPAN detects the presence of an autho-rization phicon, its applications can pro-vide a higher level of functionality. The

52 PERVASIVEcomputing http://computer.org/pervasive

W E A R A B L E C O M P U T I N G

Figure 3. TiltType.

T iltType users enter characters through a combination of tilt-

ing actions and button presses. Nine tilt positions are possi-

ble, corresponding to the eight compass directions and keeping

the device level. Pressing different buttons while tilting brings up

different character maps. With the nine tilting positions, three

character maps cover the English alphabet. A fourth button is

used for backspace. Simultaneous combinations of the four but-

tons select additional character maps for numbers, symbols, and

special functions. For example, to enter the “i” in “ieee,” the

user presses the upper left button and tilts TiltType down and to

the right.

TiltTypeA B

D E F

G H I

JC K L

M N O

P Q R

S T

V W

X Y

U

Z

1 2 3

4 5 6

7 8 90

& . '

! , ?

; , ;

( - )

$ %

` @ _

CL

authorization phicon requires little hard-ware, so it can be small. Because of this itcan be located in housings and places thathave the appropriate affordances, such askeys, key rings, earrings, eyeglass frames,cards for wallets, belt buckles, and so on—items that are already personal, easily andfrequently worn, carry the appropriatecognitive affordance, and can be aug-mented comfortably.

For example, in our prototype applica-tion, we built an authorization phicon thatlooks like a lock (see Figure 1). When thisphicon is present (the WPAN discovers it),the WPAN allows an additional level offunctionality. We implement this usingWPAN discovery—when the WPAN dis-covers the phicon’s presence, it conveys itssemantics. A real system would requirecryptographic security. The phicon can alsocontain set data, which can be automati-cally and passively transmitted to theWPAN either upon discovery or in responseto some explicit user action.

Authorization levels need not bebinary—by having more than one autho-rization component, or an authorizationcomponent with levels, it can communicatea range of authorizations to entities. Toexpand on the wireless smart-card exam-ple, a cell phone might allow calls to 911only if there is no authorization, or mightnot allow roaming calls if there is only lim-ited authorization. However, it does allowa call if there is full authorization.

Jalal Al-Muhtadi and his colleagues haverecently discussed the concept of storingcryptographic keys in a wearable device,14

although in the context of a ubiquitouscomputing environment rather than aWPAN. All the keys are stored in a singleplace (in the user’s watch) rather than beingdistributed and externally invisible.

Current power technology makesextremely long-lived WPAN phiconsimpractical, but projected increases in bat-tery power and decreases in radio energyrequirements will address this. For exam-ple, the 802.15.4 IEEE task force is inves-tigating ways to reduce the power require-ments of radio protocols by an order ofmagnitude over Bluetooth.

Conceptually, Alice giving Bob an

authorization phicon is no different fromAlice giving Bob some data virtually. How-ever, the use scenario is very differentbecause the capability is now made into aphysical object. This is easier for Bob toremember and use, more difficult for oth-ers to copy, and is language independent.Of course, this choice inherits the disad-vantages of a physical object—the phiconmust be stored, is harder to transfer overlong physical distances, and so forth. Userscan use physical techniques such as phiconsalongside virtual techniques such as pass-words—the system designer can choosewhen to employ which. This is anotherexample of how WPANs expand the rangeof design choices.

Orchestrating the userinterface

One advantage of WPANs—the factthat components can dynamically comeand go at will—complicates user interfaceprogramming. At any time, there is someset of user interface components known tothe system. Their capabilities might dis-joint, overlap, or even be identical (for

example, a user might have multiple “but-ton” widgets scattered all over their body).When the system wishes to perform someUI task, how does it decide which compo-nents to notify of that task and what to tellthem to do?

One option is for the system to assumethat all user interface components canaccommodate all requests, each user inter-face component being responsible for han-dling the request as best it can. For exam-ple, the system request might be “notifyuser of a high-priority event.” Modulesmight vibrate, flash, beep, and so on, toaccomplish this. While this approach is thesimplest for the server, it is not appropri-ate for WPANs. Different user interface

modules have hugely different functional-ities; it wastes bandwidth and energy toinform all modules of all tasks. In addition,if the user has many user interface moduleson their person, the effect of all of themgoing off simultaneously could be quiteundesirable.

A second option is to have user interfacemodules send the application a checklistindicating which features they can support(for example, “I can display color, 8 bit, 60by 80 resolution, and can play audio filesin .wav format of <= 64KB”). This elimi-nates wasted notifications but requires thatthe set of possible capabilities and thedescriptors for each be well understoodand regularized. We felt this was not yetappropriate for such a flexible and emerg-ing area.

The third option, which puts the mostwork on the application but supports themost flexibility, is for the user interfacemodules to announce their exact ID to theapplication (“I am a model <X> widget”),and have the application understand theset of module IDs. It could then determinehow and when to interact with the mod-

ule. We use this approach in SBN. SBNuser interface modules, when discovered,announce their unique identifiers (“I am aTiltType module”), and the host applica-tion determines when and how to bestinteract with that module.

Sample applicationsTo show how a simple WPAN like SBN

can be used in practice, we demonstratehow our system performs a simple testapplication. It demonstrates the use ofauthorization phicons, multiple user inter-face components, and TiltType. This appli-cation involves trading stocks; a user entersa stock symbol, and TiltType displays thatstock’s current value. If the security phicon

OCTOBER–DECEMBER 2002 PERVASIVEcomputing 53

SBN user interface modules, when discovered,

announce their unique identifiers, and the host

application determines when and how to best

interact with that module.

is present, then the user can specify anaction to undertake with the stock (buy,sell, stop-loss, and so on). If it is not pre-sent, then the stock price alone is displayed.

First, the SBN network discovers what-ever phicons are present in the WPAN. Inthis example, it is informed of the securityphicon’s existence (the lock in Figure 1).By pressing a certain button sequenceusing TiltType, the user invokes the stockapplication.

The application starts by sending a direc-tive to the TiltType module to display astring prompting the entry of a stock sym-bol (see Figure 4), and to the notificationring (shown in Figure 1) to display a noti-fication that will prompt users to interactwith their TiltType module.

Once a user has entered a stock symbolusing TiltType, the string is sent from Tilt-Type back to the application as aname–value pair (tag name, entered value).For example, in this case the name–valuepair is “stock symbol, INTC.”

The application then queries the outsideworld for the stock’s present value anddirects TiltType to display it. If the securityphicon is not present, the application loopsback and prompts the user for more stocksof interest. If the security phicon is present,the application directs TiltType to display aset of choices, one for each action that canbe undertaken on the stock (for example,buy, sell, buy-limit, and sell-limit). UsingTiltType, the user then selects from themenu of choices, and the selected option istransmitted back to the application.

Other scenarios using phiconsA phicon can also be used in a WPAN as

a data reference. For example, when a“Statue of Liberty” charm on a user’scharm bracelet is triggered, the WPAN canretrieve the latest New York news from aWeb site. WPANs with more multimediacapabilities could support more advancedoperations. For example, a smart businesscard could include an image of a phone thatautomatically dials the card’s phone num-ber when pressed. Advertisers could usephicons to promote their products; wheninvoked, the phicon could retrieve locationinformation through the WPAN to give

directions to the nearest location. Theadvertiser might later collect the phicons inexchange for a product discount.

In addition, we could use WPAN phiconsas a remote-control device. Remote-controldevices can operate outside a WPAN, but ifin one, users can modify their operationaccording to their preferences. For exam-ple, a miniature light switch might adjustthe ambient lighting. Users could controlambient music with the help of a model oftheir listening preferences, stored on theWPAN server.

Because phicons can contain data, theycan let users temporarily lend sensitiveinformation. Suppose Alice wishes to letBob buy some groceries using her auto-mated teller machine card. If Alice canencode her PIN into a phicon, she can lendthat physical artifact to Bob, who can thenuse it at a phicon-aware ATM. Once Bobreturns the phicon to Alice, she does notneed to change her PIN as she would havehad to otherwise—Bob was able to use thePIN without learning its value.

Other TiltType scenariosTiltType can assist with entry of pass-

words or PINs (such as at ATMs). Cur-rently, the user must trust the ATM hard-ware to not retain it for use by others.Thieves have taken advantage of the gen-

eral public’s trust by setting up fake ATMsand modifying legitimate ones to recordpasswords.15 An alternative is for the userto bring their own interface, which theytrust to not retain the password. TiltTypeis particularly suited to this; users can carryit anywhere without special effort becauseof its small size and because it can easilybe placed anywhere on the body. Using acryptographic challenge-response proto-col, the password is not directly commu-nicated to the ATM, but the ATM is satis-fied that the user has entered it. On its end,TiltType discards the PIN once the chal-lenge-response is complete.

WPANs containing a TiltType modulecan be used in conjunction with largepublic displays. These displays are typi-cally located in areas with significant foottraffic and show items of interest such asupcoming talks. Suppose we wish to aug-ment a display to let users easily down-load information into their PDA calen-dars. The display could detect when a userchooses or clicks on a displayed talk, butthis is problematic. The displayed areamight not be physically reachable, severalpeople might want to click on the samearea at the same time, and different usercommunities might prefer different UIs.Instead, the display could broadcast itsevent information to all nearby WPANs.To act on that information, the WPANonly needs a minimal UI because the com-munity display already handles most ofthe information display. The WPAN needonly let the user scroll through the list oftalks and select the talk of interest. Asmall UI module such as TiltType can beuseful in this scenario.

Hidden input scenarioAnother advantage of WPANs is that

modules can be scattered all over the body,heedless of intervening clothing layers. Thiscould be advantageous when at an ATM.At present, the user must enter their PINusing a large keypad and be careful ofnearby prying eyes. In a WPAN, a usercould have a small numeric input modulelocated in their pocket or handbag. Theuser could enter their PIN by feel, obtain-ing a greater level of privacy and security.

54 PERVASIVEcomputing http://computer.org/pervasive

W E A R A B L E C O M P U T I N G

Figure 4. A prompt for entering a stocksymbol using TiltType.

Our Spartan BodyNet demon-strated two new examples ofthe type of novel UI artifactsits advantages (and their com-

binations) allow: phicons and TiltType.Areas for future work include

• Practicality. While our system is work-ing, it is only at a prototype level in termsof industrial design. We could dramati-cally improve the form factors, powerrequirements (such as incorporating the802.15.4 effort), display characteristics,and so on.

• UI intermediary. It would be interestingto explore more scenarios in whichWPAN UI components are used as inter-mediaries between two “faceless” com-ponents, such as a community display,a Personal Server,11 and so on.

• Tangible transfer. We have discussed afew examples of how to integrate thetransfer of part of a WPAN from one userto another into the model. We can gener-alize this in several interesting ways. Whatif more than one user is involved—forexample, when a user needs k phiconsfrom n other users to be granted somecapability? If the granted capabilities aretemporally bounded, how are thesebounds specified, displayed, and modi-fied? What if these transfers are n-way,instead of one-way; for example, if Alice’stransfer to Bob changes semanticsdepending on whether Bob has trans-ferred something to Alice? How do wecontrol transfers of a negative capability,such as “Warning: this bearer is not to betrusted in an e-commerce transaction”?

The Spartan BodyNet is a first step in thisdirection. We hope it points the way to afuture of ubiquitous wearable systems.

ACKNOWLEDGMENTSWe thank Jason Hill for supporting University ofCalifornia at Berkeley’s dot motes and GaetanoBorriello for his continuing support and counsel.

REFERENCES1. V. Bush, “As We May Think,” The Atlantic

Monthly, July 1945, vol. 176, pp. 101–108.

2. T.M. Siep et al., “Paving the Way for Per-sonal Area Network Standards: AnOverview of the IEEE 802.15 WorkingGroup for Wireless Personal Area Net-works,” IEEE Personal Comm., Feb. 2000,pp. 6–13.

3. K. Partridge et al., “TiltType: Accelerome-ter-Supported Text-Entry for Very SmallDevices,” Proc. Int’l Conf. User InterfaceSoftware and Technology (UIST 02), ACMPress, New York, 2002, pp. 201–204.

4. H. Ishii and B. Ulmer, “Tangible Bits:Towards Seamless Interfaces between Peo-ple, Bits, and Atoms,” Proc. Conf. HumanFactors in Computing Systems (CHI 97),ACM Press, New York, 1997, pp. 234–241.

5. O. Shivers, “BodyTalk and the BodyNet: APersonal Information Infrastructure,” Per-sonal Information Architecture Note 1,MIT Laboratory for Computer Science.

6. R. Hannson, P. Ljungstrand, and J. Red-strom, “Subtle and Public Notification Cuesfor Mobile Devices,” Proc. Int’l Conf.Ubiquitous Computing (Ubicomp 01),Springer-Verlag, New York, 2001, pp.240–246.

7. T. Kindberg and A. Fox, “System Softwarefor Ubiquitous Computing,” IEEE Perva-sive Computing, vol. 1, no. 1, Jan./Mar.2002, pp. 70–81.

8. L.E. Holmquist et al., “Smart-Its Friends:A Technique for Users to Easily EstablishConnections between Smart Artefacts,”Proc. Int’l Conf. Ubiquitous Computing(Ubicomp 01), Springer-Verlag, New York,2001, pp. 116–122.

9. T. Selker, A. Lockerd, and J. Martinez,“Eye-R, a Glasses-Mounted Eye MotionDetection Interface,” Proc. Conf. HumanFactors Computing Systems (CHI 01),ACM Press, New York, 2001, pp. 179–180.

10. ARC Group, “Bluetooth Personal AreaNetworks”; www.arcgroup.com/reports/bluetooth/bpan-brochure.pdf

11. R. Want et al., “The Personal Server: Chang-ing the Way We Think about UbiquitousComputing,” Proc. Int’l Conf. UbiquitousComputing (Ubicomp 02), Springer-Verlag,New York, 2002, pp. 194–209.

12. J. Hill et al., “System Architecture Direc-tions for Networked Sensors,” Proc. Int’lConf. Architectural Support for Program-ming Languages and Operating Systems(ASPLOS-X 00), ACM Press, New York,2000, pp. 93–104.

13. D. Russell and R. Gossweiler, “On theDesign of Personal & Communal LargeInformation Scale Appliances,” Proc. Int’lConf. Ubiquitous Computing (Ubicomp

01), Springer-Verlag, New York, 2001, pp.354–360.

14. J. Al-Muhtadi, D. Mickunas, and R. Camp-bell, “Wearable Security Services,” Proc.Int’l Conf. Distributed Computing Systems(ICDCS 21), IEEE CS Press, Los Alamitos,Calif., 2001, pp. 266–271

15. Boston Globe, “Fake ATM Plays Gotchawith Users,” 12 May 1993.

For more information on this or any other comput-ing topic, please visit our Digital Library at http://computer.org/publications/dlib.

OCTOBER–DECEMBER 2002 PERVASIVEcomputing 55

the AUTHORS

Kenneth P. Fishkin is aresearcher at Intel ResearchSeattle. He has more than 15years experience in industrialresearch and development.He cofounded the eXtremeUI research effort while atPARC, which helped invent

tangible user interfaces. His research interestsare ubiquitous computing user interfaces andsystems. He holds bachelor’s degrees in com-puter science and mathematics from the Univer-sity of Wisconsin, Madison, and a master’s incomputer science from the University of Califor-nia at Berkeley. He is a member of the ACM.Contact him at Intel Research Seattle, 1100 NE45th St., Sixth floor, Seattle WA 98105; kfishkin@intel-research.net.

Kurt Partridge is a PhDcandidate in the Universityof Washington’s ComputerScience and EngineeringDepartment. His researchinterests include technologyto support novel user inter-action mechanisms for

wearable and ubiquitous computing devices.He received a BS in computer science from theUniversity of California at Berkeley and an MSin computer science from the University ofWashington. He is a member of the IEEE andthe ACM. Contact him at University of Wash-ington’s Department of Computer Science &Engineering, Box 352350, Seattle, WA 98195-2350; kepart@cs.washington.edu.

Saurav Chatterjee is anundergraduate student inthe computer science andengineering departmentand the Honors College atthe University of Washing-ton. Previously, he workedwith Kurt Partridge to

develop TiltType. His research interests includeubiquitous computing and artificial intelligence.He is a member of the IEEE. Contact him atsauravc@u.washington.edu.