til‘ the center of the ac.-
`relic» begins with the lfii.‘-§.lll€)I1
`tivateil area. For this pnrpnsel all nonzero pixels are
`taken in be part (if the object. The center and the
`point farthest away frorn the center determine the
`major axis iii" the (ibjeel. The minor
`is taken {Q
`be perpendicular to this. Tliese axes provide an
`object-relative cnorilinale system in which it is pos-
`silale in specify, mughly, the location of bumps and
`iiepressinrls in the image. The botmdirig reeizirigle of
`the object is lalcen to be the smallesl. rectangle, with
`edges parallel to the axes, that cznritnins the image.
`The aspect ratio of the object is taken to he the as
`pee: ratio nf its bounding rectangle.
`
`Mnving the Finger
`
`ll‘ the image read in is not satisfaeiory, as is usually
`the case at first. it is possible in move the linger and
`read another. An important part hi‘ the image analy—
`sis involves annving the finger so that an optimal
`image is sensed. The offset pressure, for example, is
`adjusted in this manner. Optirnally, most (if the
`touched area activates the niisl-range if the sensor,
`nllnwing bumps and tiepressiens to be detected eas-
`ily. This is aecnmplished by rezarling in an irnagge,
`enrnputing the median pressure of all points above
`the noise threshold, and readjiisling the finger pres-
`sure nppt'0prizn;ely. This is repeated several times
`until an acceptable nliset pressure is achieved.
`The finger is also mnved re measure the stability
`(resistance in roll) of an nlsject. To measure the 9-b~
`jecfs stability in El. given <lirei:.tinn, the nhjeel is
`pressed between the linger and the supperiing sur-
`facre by applying 31 fixed for‘-ee on the object normal
`In the plane (if the surliatze. The finger is then moved
`laterally in the desired direction. (The supporting
`surface was a thin layer of soft rubber, to prevent
`sliding.) The stability ef the object is indicated by
`the amount 01’ force necessary in move the linger.
`
`The Matcher
`
`During the liypnthesize step, the program must cle-
`termine which of the objects it knows best matclies
`the knnwn data. For a small set: of possible ohjeets,
`
`such as the fasteners, it is net really irnpnrtant that
`this be alone well. For 3 large set of pessible objects,
`ilie quality of the msteher may he 2: iiete.rininlng
`factor in the speed of recngnitieii. When the hy-
`pothesis is chosen by seleetien til the possibility that
`best matches Elie information given, usually the
`ehnice that has the largest nnrnber of features in
`commnn with the known facts is the best choice. In
`
`a system with a large. number (if parnmeier’s, other
`factors may also be taken into censidemtion. For
`time thing, snme features may he mere impnrtzmt
`than others. either in general or for that pamcular
`possibility. Also, the features tlieniselves may not
`exactly inatch—a btirnp rnay he too large. 3. shape
`elisterted. In cases such as this, we wish to give the
`possibility nnly partial credit for 21 feature match.
`The most Ol’.‘?Vl01lS way to implement such n
`mntcher would he to use a numerical scoring system,
`with the weighting; of factors for feature irnprirtzince
`and periinl inzilelies. This a.pprna.eh was aVe)ie.le<l
`fer the following two reasnns. First, there wnulcl
`have is) be ea degree pl" arbitrariness in assigrii'ng the
`numbers: is a circle 21 53% match to 22. hexagon‘? is
`shape 2.5 times as important as texture, or only
`twice as important ‘.7 It is unwise to trust the sums
`and pX‘()(ll.lCl‘S of nnmlzsers if the numbers tliernselves
`are chosen arbitrarily. The seconzl objection is more
`(if a philosophical one -----converting a complex set of
`syrnl>0li¢:: slrucrtures into a single number caiises us
`tn throw away me much information we quickly. Of
`course, this information must eventually be lost
`the member must termirinie by selecting a single
`item. But the pruning can be. and is. contrnlled in a
`more reasnned. l’E):&)‘ll'l€l".
`
`The impleinenteii rnarclier takes two possibilities
`at a time and compares them on 3. fenturewby—l'eeture
`basis. ll", for a pzwtienlai‘ feature, bell: items rnateli
`the image in silent the same degree, the information
`is ignored. if one of the items is clearly a better‘
`matcli, Elie feamre is counted in favor iii‘ the appre-
`priete item. This prneei;lnre is repeated fer each fee-~
`titre and then the features themselves are compnresl
`in 3 similar manner. A feature cniintetl inward one
`
`item will be cancelled by a feature eonnrecl lnwarral
`anrillier, if they are of apprnximale impnrtanee.
`in a large artilieial intelligence system, the best
`match cnulrl lire computed in parallel. Parallel
`
`£2
`
`The intermztlanai Joumczi :2j‘Roboi'ie:s Rszsecmr,-'1
`
`Page 877 of 1714
`
`

`
`marker-propagation schetnesi, such as the one pro-
`pnsed by Fzalilnian ($79). would do such a task well.
`One important assumption, even for the parallel
`case, is that the binary comparison operator is ta'an—
`sitive. Without this coiistraitit, it would be tiecessary
`to cninpatre each possible pair of items. a task that
`grows as the square of the nunther of items.
`The trntniitivity of the predicate described above
`can be e2..sil_v demonstrated, given the trzmsitivity of
`indiviitinai fea.tLn'e cnlnparisons. Assume that there
`exist three items, A. B. and C, such that A :>~ B and
`B :=- (I. Let fix, _v} be the set of t"entLn‘es counted in
`favor nifx when compares} with
`Since the individ-
`ual feature comparisons are transitive.,
`
`f(A, C} = €f(r‘\, 3) U 17.3‘ C)?"
`
`fit? Al = (f(C.~ Bl L.if{B.. Al)-
`
`is the fezltnre set comparison predicate (the sec-
`lf
`ond stage of the aigorithtn 2tl.‘P0\'€),
`then A .> B im»
`plies_f(A, B) $> f(B. A). Also, for any sets a, h, c
`and Li such that ti > b and > rt. it must be that
`in U C) > nib U (i), because features that cancel in
`the individual sets will aim cancel in the union. The
`
`assninptions A > B and B
`{J inipiyftA, B} S»
`f€B. A} and f(B, C) ?>‘»f(C. 8) anti. by the union rule
`tft'A, B) U ftici. (1)) :>- (j’{C, B) U f(B. AD. This
`may be rewritten as f(A, C) ?2>_f'(€f, A). which is the
`criterion for A :> C. Therefore. the matching predi-
`cate IS trantsitive.
`
`This member is really overkill for a gaossiliiiity set
`of six objects with three parameters each, but it may
`be necessary if the ;r3rogt'ani is to he extended to a
`large range of objects.
`
`Fro posed Efforts
`
`A program that ciistlngiiishes ziniong six itthjects on
`the
`of three parameters is not too impressive.
`Even if it only got one hit from each pnrnarieter, it
`szhouid have cot'rei:.tiy t'ecogni.zed eight objects. in
`the future, tactile recognition programs will have
`nniszzh more complex and more precise represen-
`
`tations of tactile iniages. 'I‘l'iree improvements can
`help bring thin nbnut.
`The first is texture recognition. The resolution of
`the tactile array sensor, while high, is not not nearly
`stifticient for measuring textural ciifl’erences between,
`say, {inner and glass. Texture seaming requires
`suting bulk effects nf many tiny surface fezittares. It
`is most easily accomplished if something is slid over
`the
`and a pattern of viinaiions in detected.
`This can be likened to sliding a phonograph neetiie
`over a record. Sensors of the future may use em-
`laecideci mezoeiectric devices, an it may be possible
`to use the ACS directly as sort of a carbon micro»
`phone. iiowever the information is deriveti, it must
`be _i;)ro«::e::<seci into it nsefnl €i'ii3t3'21C1i€l"i2‘Zi1l.l-lZl3?l of the
`texture of the surface. 01’ interest is the intensity and
`periodicity of the signai. These featuren may be seeit
`directly in the frequency domain. 'l"extnre processing
`may hear more similarity tn the analysis of sounds
`than to the analysis of visual iniages.
`Another improvement might involve thermal rec«
`ognitinn: the dil‘ferenc.e between paper and glass is
`that glass feels cnlti. This is not actually because
`glass is iower in temperature, but because it is a
`better conductor of heat and so it is; more qtiicltiy
`able to carry away the heat generated by the body.
`We have constructed a small thermal conductivity
`sensor that worl-cs on this pririeiple. in the sensor, 2.
`resistive heating element is sandwiched between two
`temperattn'e—sensii:ive ttnrrent nnurces. Any tliil
`ference in the temperature of the twez sensors is incli-
`cateii by an easy--to-measure difference in the cur-
`rents. The sensnr is designed to be mounted on the
`finger in such a way that nne teniperatnre sensor
`iney contact the device being tested. As the heat is:
`drawn from the object into the sensor, 3 <iil‘fereni:e
`in tem_neratnres will develop, The pmnnry clisadvarn
`tags nf this first prototype is that it is large
`(‘(3.1 X 0.3 X 0,2 in.)i resniting in a relatively high
`thermal
`This iimits both the response time and
`the rninimnm size of the object that may he tisefuily
`tested.
`The third area that shows immediate potential for
`further research is the coorriinntinrl of multiple inc"
`tiie images into 3. global picture. This is probably the
`most nsefnl next step in tactile processing. This
`problem was tleliberateiy zwoitletl in the progrztni
`
`Hir’:’:'.r
`
`Page 878 of 1714
`
`

`
`described through the choice of smali objects that
`ssouid be read in 21 singie impression. Such size iimi—
`fatimm are pmbabiy aiiireaiisiis: (sutside the iab0ra--
`wry enviranment. The first reai-wurid appiicatimis 01'
`tactile smsixig WE}? not be in mccigiiizing mhjscts {hat
`{ii on the: tip of ihe finger, bui rather in orienting
`Ecmawn objects grasp:-.d with an {mime hrmd. This wiii
`requirc coordinaaing images from muitipic: se.nsor's.
`We are enthusiastic about the future prospects of
`autorriated tactile sensing. Wings. has been described
`here »»»»»-the SEEESOE‘, the: finger, and the program ~~~~~-is
`onfiy an initiai app:'0as:.h.
`
`Ackmzwiedgmemg
`
`I wouid like In iiaank the faiiowing pmpie foe‘ ihair
`help, ideas, and enthusiasm: Mike: Brady, Tom Cai-
`Eahzm, Fred Drenciahahn, Richard Greenbiati, John
`Pi1rbris.:k, Geraid Susszman. John H0i‘:er'b;:iL;h,
`Bzriiciiaei Deriouzos, Margaret Minsky, Laurei
`Simmons, Patrick Winston, and, most of aii. Marvin
`Minsky.
`
`REFERENCES
`
`'E‘he utiiizatism of an “‘ariiii;'iai
`Emit. M. 3979 {Marcia}.
`skin" sensor fur the idenziiicaiicm of semi objects. Proc.
`9!}: 1:21, S;.‘m,z3.
`i’rz(z'zis!.re1’z/ Re:-b0zi¢'s.
`
`Harmon, L. D. 1932. Automated tactile sensing. int. J.
`R0.r'mz‘ic.s~ Res. E{2):O{3-—(i{).
`E-Iiilis, W. D. ESE (April). Active much sensing. A. I.
`Memo 62‘), Masaachusem Irafiiimse nf Technulogy Ariin
`ficia] intelligence L.abor2atory.
`Fzahiman, S. 1979‘. NETL: A .*;_‘»'.:<I2-‘.rr:_,’i'H“ :'e,IJre3:»'r:!:'rs,r{ (Md
`izséizg rc=(::’~\m:~!d »’;::«:wiw'gr:*. Cambridge. Mass; MIT
`Press.
`E976 (March). Tziciilea Sr”:fiS%i)E’S sonar
`Larcnmbe. M. H.
`and parisiiax sensors fur robe! app‘:is:ati0ns. Papcsr de;iiv-
`{tied at 3rd Ccmf Inciusariai R«‘)i)0E Tech.
`Cicada, T. 1979, Gbjrsct handling system for manna! in~
`ciuairy. IEEE Tmams‘. 5§‘_vxI.. Man, Cyizern. 9(2).
`Okadea, T.. and Tsuchiya. 5. E977 (Oct). On 2:. versaiiie
`finger system. Pmc. 75.’; Inf. 5_wrzp. I:zdu5zrim' Ra!»-t.~I5.
`Purbrick, 5. A. (1981). A force Hzzns-siiuccr cmpiuying com
`ductive siiicoiie rubber. Papa‘ delivered at 15: Robot Vi~
`sion and Steiisorsi i..'orif.
`isategnatcad b<:i3;avic:r <33’
`Stajiijkuvic,
`and Ckit. 1., I977.
`arfiificial skin. IEEE Tnms. Bizi.-mm’. Elzgirwyrizzg
`24(4);396—399.
`Storace, A., and Woif, B. 2979. Fimctioizai anaiysis of the
`mie of finger Iendans. J. Biorm-*c'izc.':11'c5 i2.
`Wcinreb, D,. and Moon, D. 39793, Lisp z’§.;f(:(‘kine* ;'iir.m.'m'f.
`Cainbndge. 1Vias.<;.: Massachuseits Institute: of Tcc.hnoi~
`ogy Artificial Emeiiigence Laburatory.
`
`The Irm?;"izaiiorml .imm2z:1 ofRobo£ic:: Rasearsiiz
`
`Page 879 of 1714
`
`

`
`APREL 1385
`
`fin &§E£§..T§»T@i§@H THREE §iMEi§S§CiN&E. °¥°Gi.5$i*i-$§i\‘i$i?i\°'§ "i'A§LE?
`
`SK. Lees, W. Simon, KC. Smiiii
`Gompmnr sysiamss Piesearsaia Ensiiiuis
`University of Tomnio
`Toronto, Oniario
`Canada, M58 ws
`
`(4163-9?6-$320
`
`.&$S'i'i‘§iM§’§°
`
`A prrzioiyoa misch-sensitive iaiiiai is prosieniod. This tai:iiai’s main
`irinovaiizm is that ii
`is capabio oi sensing more ihan om pain’-i of
`ooriiziitzi at a time.
`in addition io bssing shin its orszvine pesssiiion coor-
`ziinzsiets, mo iahisi also givos a nmasam oi degreo oi ooniaci,
`incieoenoaniiy ior oasis swim of ozsmzsci.
`in order in enaisio imiii~
`ioiacii sensing, the iabisi ammo is oiviooti inio as grid of diaeresis
`points. Tim gsoinis are iszzannoci wing a recursive arm 3i.bdiVi§I0n
`aigoriihm.
`in order in minimize me ressoiuiicm iassi dues to mes
`oiscrssis natures oi Sh-B grid, & nova! inierpeiaiion schema has boon
`siaveiopen.
`Finaiiy.
`the passoor briefly oiiscisses how muimoucii
`zaensing, inierszoiaiion, and degree :2! oomaci sensing can be oom-
`binsiti to saxpaend our vcsesainuiary in iiL.irnan~::ornpL3ier inieraniion.
`
`"ii. ENTRQQQQTBQN
`
`Rapid advanmnism oi computer ioonnoiogy has ononoci a variety
`of new appiicaiions. New oppiicationo and igsero mean sienvainzio for
`new {nodes of iniersxciion. Ono mnsnvquenoa of this is 3 growing
`soorociaiion of mo imporianoe oi L.-sing aaporszprimes inoia iesoiinoio
`gins (Eiwnon. 1932}. Fosiiioningg devices am: man is: be essiseniiai its
`graphics apssiicaiiams,
`image iimnsifiucam am required for paiiorn
`recognition in meoicai diagriosia, iczmsah screens are iseiui ior ins
`education of young; chiidren, and ihe QWERTY keyboard resmains
`iiaaa usmi standard for sex: processing. However, the range oi‘ inpui
`zievicos swaiiabies is siiii quite: iimiiean, as is our unooramnriing oi now
`io uso them in the mosi oiieoiivo mamas’.
`
`The imam oi iiwe rosaarciz presemod in inis paper is in incraasss mes
`vmabiiiary than can be uiiiizasz in numan~coinouier imoiramioni Cur
`approach isas been to eieveiop as now ingzui iaciinoiogy that oniarges
`iho domain oi human physioai gssisxieo ins’: can no cssziurmi ior
`mmroi })L.'i’;?4?3$€S2!5.
`in what ioiiows, we wiii oewibo iho iechnoiogy.
`what ii avowed from. and some aspascizs of how ii man be usoati.
`
`2. OVERVBEW
`
`The iranssdmar that we have -iiavaioped is it ioLscn~son5iiivo iabiei;
`Eiiai
`is, a 1153 suriaoe than nan seense whores ii is being ioucheoi by
`the operaior’s finger. This in iiseii is mi now. Savorai ssmh dw-
`iwss are ssomrnerciaiiy avaiiaibio from as nornizer of manufacturers
`cfsieo Appisnszix A). What is unique about was iaoioi is inn: is ooni»
`
`Permission to copy wiihoui fee ail or part of this mmeriai is granted
`pmvierled that the copies are um made or distributed for dizec:
`comm-:rc§s1i advaniage‘ the ACM copyrighi notice and the title of ihfi
`pubii-zaiion and its dam appear, and notica is given that copying is by
`pttrriiissiiiyi of the Associaiion for Computing Machinery. To copy
`mherwis-3, or so ropubiisia, requi:'e§ 22 fee asnd,/or sipecific permission.
`
`I935 AC5/i $8979i—i4*}~€i,"85/(}i)4/i}{}2i
`
`00.75
`
`bines iwo acisiisiemai ieseaiuress. First. it can sesrms the degree oi con»
`east in a ooniinuous manner, Socorro, if can sonso iho simouni and
`ioczaikzn oi as riwnitzer oi simuiianeoao ooims of oomaci. These: iwo
`iosaiasrea, whoa oombim-mi wiih ioucii sensing, are wary impcriam in
`ressaeci to the iypos oi‘ interaction aim we can suppon. Sam of
`ineso are discussed beiow. iiizi sea amen, i-iiii. and Rowiey (193%)
`and Brown, Bastion and Murtagh {$985} for mom cieiaii. The iatsioi
`which we preasni is 3 oominisaiion of work done in our iais by
`Sasaki set 22! (ions) and iieieiha {E382}.
`in this gsresotiization which inflows. was iocus mainly on issuas rniai—
`ing in tho transdisoam impiomamaiion. Twas important oontribuiionsi
`discussed are ow method :1! ooanriing the this’: surface. anti our
`rrueihod oi misinisiining nigh iresoiution nespiio ins zsurioco inning
`oariiiionevs inio ea oiscrmo grid. émdiiionaei ins-niznioaai séeisiizs cars am
`foumri in L36 (1 $384).
`
`3. WKV i§l.Pii.'i‘i-=T0%3$3H‘P
`
`Tomb sensing has a nunwbar oi imponam iznammorisiiss. There is
`no niiyslcai swims or puck in gen iosi, broken, or vibrmo out oi posi-
`tion. Touch tiaizaiois can be moideo so as to make them easy to
`cisan (iiiearoiaro making inern issiaiui in cinema Si"i‘i4'ii‘i2‘:i1i"."iv&i'i%SiiK& iiosr
`piiaiss, or
`sciirty onvironmemis iikss
`iazmsrios). Since mm is no
`mechanical intermediary Between hand nmi iaizioi. iiiero is mining
`is: pr-wont :|‘i‘i'i.ii’i5~S&3i.b!3§'i assnsing. Yompiatos can be piacoci over me:
`moist so define special regions and. oinoo the hand is being used
`-fiireoiiy, iiiesa regions mix be manuaiiy sensed, iiioreisy aiiowing fine
`irainod usazr so niiociivoiy ‘mach type" on tho iabiot.
`wmmm gnossum sensing, however, the uiiiiiy of ‘imam iabiois is
`nuns iimitessé. One can move as smoking syrrxixoi nmsami the scresaan,
`for exampiss, bu‘: when tho iirigor is mass’ at iigm taxation, inesw i$ ?10ii"i"
`Eng oouivaiom io ins bimon on & i'‘:‘'£)i.S% So push in orefiez in make a
`aesiesciiesn. Yass. we on-uizi iiii Rho iingor oii iiaes ‘isiisissi. imi iiissi wouici
`be rmra iiius puiiing (rather ihan gzushinggi the biifitsn. Anti wiiai if
`we wanted to drag an iiam being poinioo oi, or in indicaio mat we
`wanted to siari inking’? Lifting our fingers’ wmixi ioav-e our {ingest :31?
`me iaisioi, just when we wisni it in contact with it the most. Them
`ism ways arounsfl ihis probiom. inn they mo inoirocii
`ii. iiowovar,
`ins Raisin: has orzssssuro sonsing, wa can gzuesix as virissai button by
`giving an own bit oi pressures to ssigmai a change in amiss.
`Prosswe has inner advantages. Cine axampies is ac wnimi tine
`ihinknssss in a paint program. Sui why tie: was wan? muiiigsies gsnini
`sensing? A simpie smrnpie wouin no if we had a iismpiaio piano
`ova: the iaoisi which oeiiniiied iixroa rogians oi 9 can try 2 sin.
`Whore was itsuois each rogion maid onmmi the seizing of o parano-
`ior nssocizsiod with each region.
`if we wanton to fi5mLPi?.&i‘i$OiS§i}i
`aiijusi aii inroe oaremotem. than we wouisz nave to his aisle in sense
`aii ihroa regions. An even oasis? ommpio is using fine iaitaiai in
`amidata a piano koyizsoard iizai man may smiypimnic mimic.
`
`Page 880 of 1714
`
`

`
`:5. H3mE3'W5iRiE !3E£€3¥§EP'E'5§N
`
`rnuiiigzier-imx:h-=
`ihe hardware :3’? Eire fast
`A hriei descripiien ei
`sssrseiiive inpm device (FMTSEEJ) is inirwmee here, The eeeigri ef
`ihe iiareware is izesed er:
`iim reemremenis cf the iesi seeming;
`aigaeriihan arm on irariesziis iagsiween soiiwszre and himiwsarei Mercy
`aerssszsrs have been exarhirseri far cur ersriiesaier eepiieeiien. ixewei/er
`(Hurei,
`‘Q74; Hiiiiss,
`‘E932; TSB, 982; TNSA, Q86; JSRQ,
`‘E981;
`iviethe, $982) names seemed ta have the prnpsriiess the: eeiisiy {he
`reezuirememe M e FMTSIQ. The hardware hasicaiiy mheisis ei a
`sensor mairist board,
`row and eoiumn seieciien regieiesrs, NE)
`eszmveriirag cirems and as mmraiiirig QZFU.
`The design ef aims 5i€5?iS€3i mairix is based on the technique of capa-
`chance rraeasarewieni izeriween a finger
`tip and a metal piaie. Te
`rhihimize i\ars:iwaz'e,‘1i’ie sensers are accessed by raw aiisé miumra
`E§€£ii3€3ii€3P'i.
`Raw seieeiion regissierss seieci esne er more rcswe by sei~
`Ring the etsrressmnding bits is a high state in arder in charge up line
`sensors whiie the eaiuinn eeiectian registers eeieei sane er imre
`eeixmrie by turning en corresponding eriaiw swiiciiee is discharge
`the siansere ihmugh iiming reeisiere. The intersecting regieri ei Eire
`seiecied rows and ihe seiecied coiumne represenis the seiecieci
`senswre as a grease. AID eenvening
`airs.-uizs measure the
`discharging iime irsieniai :2! the seiemesi eehmre.
`A University :21
`Temme ease heard is used as as saszmircsiiing CPR}. The men 31.3--
`isaeses ei the §£ii*i$i3i‘ beard mnsisis iii mmher ei smaii rheiaimeeted
`reeiarigaiar-shapm areas: serving as essmwr plates eapaeiiexre. The
`design ei the mesa: pieie area ei a unit sensor eieeencie an ihe
`measurainie eapaciimoe change that reams when iire grass:
`is
`severed by a iirmr tie, and an the reeeiuiien than can be impis-
`menied.
`
`Ciiergziirigs dicde
`
`a Eofil line
`
`SENSOR,
`
`Dxechsrging dude
`
`IC
`
`5
`Ks eelxam: line
`R
`=
`
`ewcaluznn selactlmz
`switch
`
`Fig. 1 A mzxiei m‘ e aesieezieci seam? in ‘me $€8€'i5&3i' matrix.
`
`in order is seieci a sensor by row and miumr: avmss, twa diedes
`area used with each seriser. (ins: eiodee, mririeeiesé so ‘me row Elna, la
`used ‘:2: charge up ihe sensers in the mm. is is: reisrreri in as ‘me
`Charging; Dime {GD} as shm-km in Figme 1.
`‘Fire {:9 mm serves its
`biock this charge fiszrwing bank is me we line when the raw iirie vai
`isage is dropped in zero, The eiisexr diadss csaiiassi ihe Discharging
`i'.iiodet(Di'.i}, zzenriecieci is the miimh iirxe, enemas eiieehargirig hf iire
`seiecied raw senscsrs is as vimmi greimci. Aim the 123i) hissszisas
`charge firm irem the sensers in the seieeied raw 3:: the eensers in
`this iifisiiiifiitiifisfi rewza during ihe discharging geried. The eeieciisn ei
`i‘£8W$, by the row eeieeiiere prmadme, catasssee the eensmre in has
`charged. “me sensers in ihe eciumrs are then cieciiergm through
`assemcaiaied iiming resistors eenneciw ie has eeiumn seieciiers
`switches.
`
`The ehargess aimed in me selected rams) iimu down thzrongh the
`sesiescziad
`swiiriiree to she viriuai grasnfi 0? 3 ms! operailehai
`amplifier. Ali the discharging currssrris are mrrespenciingiy edaiesei ts:
`produce a signei imm wiwieiz the éischaarging iiime oi aaii the eesierzied
`sensers is ieunci by mrhgsarisen wiih e ifireeheid veitage.
`Preesize wnsiiiviiy is incerszaraiesz by him measures: Firm there is
`the affect, hare miner, oi wmgarsssiari ei the sverieying ineuiairsr.
`S«eceiid iiiere is: the effect cf irriririek: spreading cm’ the eomszreesibie
`finger Sip as pressure is irrcreaeee.
`
`___ii.,§?..E.._i..,L_.i 9 3
`
`The eefiware in the mrsiroiiing CPS) uiiiizee communication with me
`hosi escampmer
`te emcsmmodaisa the imersseiaiien scheme. The
`cioci: raie (ii) MR2) aiimvs aizom fl consists in correspond in ihss
`eeneer eepacziianma chaxige due is as touch. But. :2? accuses, iise c:aar>e~
`dishes at’ eii me eircuiiry smashed in me esesiumn iims during ihe
`discharging periee is much iarger than the eenemr eaeaztiiance.
`Thus izeieres scermings the iabiei {er a teach,
`ii
`is scanned ersmr
`pie-teiy in 52!! possibie reseiuiim modes when mi icmhed. The
`vaaiues so aiaiairied are sicsred as reierenees. Touches are
`irieniified by ihe riiiierences between ihe rerierenm vsaiume am‘: the
`vaiuae measured during use.
`The sepeciienee chazrige eerreepaririing ie the ieuch by mere than
`me finger (er by me wimie hand) is very iargs. Tum iha mirriher cf
`bite in {has mmiesr shsiuiai be enough in smegma ihe rnaximmn
`eapaaciienee. Htswesver ii
`is urrnecesssery either to have euiiicient
`iniis in measure the entire cegssiciisanee irzeziuciing the $i.ii’f£3iiP‘it§§i'i§
`capaniiaznms, er in eiere the eerreepemiing "eoi'ripieie" counted‘
`veiuese see reierersesse.
`ii
`is necessary oniy to have runes mare hit
`ihan the mxnizer rri bits required is cam: iiie vaiue ei change in
`Rhee capaciiariw rather than me eamgaieie trainee in same: ta ¥'i"ie’.$5.Ei‘i.Ei‘é3
`she eiiierences 2:34.‘ cepaciianee due is ioueh. Tim cmiy er: 8 hi!
`mists:-er is impiernenieti. The munier erieicziee the measurement 01' 3
`7 izii eapeaesiieance change regardiese ei ihe rsegree oi evesriiew in
`she cminter.
`
`A ieciiiiy is aise previeeri fa‘ iziemifying iempieiea appiied in the
`surface as? iiie ieisiiesi.
`
`§.. SSRHNINE fiiefiflflfiflfl
`
`Cine idea sari some significance that new has iriireducazsi is '15:; aveieii
`scanning of 5:13 the pixels in me iabiei which corsieir: na iniormaiien.
`Far example, scanning aii 26:38 points of 3 ieeiei having e i°ei$OiLF
`iion 5-i» isy 32 for iewer ihan ‘Si points: is resasiiy quite 8 ridicuioue
`idea.
`in iaci, ii the number of paints is he eearzziied is eemgsambiy
`srraaii, ihers an imgarevesci eigeriihrre, here miiem‘ i'i3<>1ii!’Sii‘d'£$ area subdi-
`vision, can he used, A penimaier
`impiememaiien exampie is
`descritseasfi as iniiews.
`
`itxmsiiier at iahiai with reseiutieri 3 by 8 is: he eeereheri in? as iweia
`paint as sihzzwn in Figure: 2. Firsi, chest-zk the izebiei fer tzsixzh am e
`wheie regieri as eimown by the ewe A863 in Rite ‘iigure.
`ii’ march is
`deieeied, divide ihe tablet irate iwe eqmi regime ehewn by the Sine
`EF and shed; each 9‘!
`the two regions ABEF and EFCE3 for
`ioiichanimeae. Sesieci the icsisehiaei raglan. ?‘€iQ§€3?‘i EFCD in this case.
`and divide ihie imn mm equal regimes as eimwh by the siivisiesri iine
`SH. Gehiiniae this process en the {embed regien urriii no imhef
`divisien is pessibie, aim is, uhiii a uni: E8i‘i3wO?, iieeigheied as this
`region MCMQ in Figures 2. is reached. The figure eien shows ihe
`sequence oi suixiivisien in the rem-rsive suisdivieieri scheme.
`
`5
`
`(n)-Sequence as suhdivisien 1:: binary cperationz.
`
`Fig. 2 Resessrsive eutsdivissieh operation far 8 by 8 iabiei.
`
`Page 881 of 1714
`
`

`
`85 PROCEED§
`
`APFtEL1§85
`
`2%‘ Spatial Ramfiution
`Om possitstss amt immediate: Ehterpetatlan sszhenta is tea Emerphtate st
`”tm.>r£tatt” mint with at: aztjasent values whtch may mt be Eargs
`eamugh to be reported as tnmhect A heat army at 3 by 3 points
`can be used far this tnterpciattnn. Same estarnpfes drawn on as
`Easer printer {censeqtsentty having no intensity watts) are shown in
`Figure: 3. These pictures are streamed withcsut fsssadttasak. that is,
`drawn withaut thee operator looking at the output screen. This times
`not zaitaw the zxperatcr ta camtuensata, that is, tea zseiect paints when
`data are sparse tn mmparisszn with the interested figure, hut rather
`takes attract Engaut tram the tcassatitm at the flgtm drawn an the input
`swim. The first pictura ta) is drawn by moving 3 ttngar in 3
`straight tines Qgutcteszi by a ruler} tar vartcm angt-as and the as-c<>n:£
`me (ha is drawn by rrming as tingar tn 2: Gina guide fly & chats atrawh
`on 3 tssmplate.
`Theme teams
`shaw that
`Itataswataticen actuaiiy
`increases the sgsatiai resctutten as west as the texaatabiitty 221 a tines
`point an a screen.
`
`
`
`-....).A4[.L)4JJ.!-Ln<---4.;
`
`t 3
`
`raw scale
`
`(a) Straight llnss drawn by tha table: using 5 by 5
`sensor array inzerpulatian.
`Ths scalss shavn regrossnt
`the boundarius cf the
`actual aansorao
`
`‘-
`
`cclumn
`
`t
`
`t3
`3
`
`rcw acals
`
`Eb) A circle drawn by the fiablet using 3 by 3
`sensor array intarpnlatian.
`The scales nhewn represent
`the boundaries cf the
`actual sensoraa
`
`Fig 3 Paints éravm hy tha tazsrat using an trttarpcisaticn mxethad.
`
`Using 1313:: aasgortthm, a search tar one mint em as tattiext having a
`ves<:£1..-{Eon 8-5 by 32, requires & scanning times, that its
`
`2'{E0g3ub 2)($4$“’32)=22
`
`rat: cverheas in the I'&%t2f$hi& aubsitvtstan prcaesesesas and
`is:
`there:
`it
`scanning tretgtrsss at the: "top of the has" (mat is, with a raglan in
`which an gstxeta 3:9 gtauged tagstheevt, than wing thin sschesmss. ths
`msmtzer of tesmzhmt paints that can be kientmed in the ttrm that it
`warm take to detect ans touch directfy (that is,
`if at! pixais are
`scanned one Ezy {ma aequentfaiiyj ts
`
`N = {(54 ° 32} over 22}
`
`138.
`
`This shtzws immssdiatety that the recursive ssutxdtvtsion schema is
`much superlnr to sequential scanning if ma number at pcstmes is: be
`mannm is tawer than ‘$33.
`
`$.
`
`fi’§‘?E§P§S.A"§'5*i§N
`
`It may szaerr-. that the resceiutian at the hardware is tam tcw ta-r use in
`graphics agspttzaatfiona. Hawesvasr touch intensity and mtétttsmh sew
`sséttvtty can be asset} is enhance rasatutien. This ts passstbie tmcauzas
`the gamer at as tsatszh can be most accurateiy astimatext by an tntssru
`gszstattzsn uttitztng the vaiues at the: asdiaeent esesnmr intensities.
`Direct
`intesrpeisttiatt
`schamesss far
`as {aw cessesa hast bean Emma»
`f‘:‘38MP.<5. Cine at interest ts tn tmergmiate an array :3! :8 by 3 sassrsscars:
`using at
`tcassszhefi paint tn the center. Another ts ta interpolates alt
`paints an the tatztet. The taster macs extsvicmty prmrktas that highast
`remiution but as a wsutt it simpigi emutates 3 singie tumh tablst
`with vary high ressotuttsra.
`7. PERFQRWWE
`
`‘M £§ssxta-s
`An Ezflaaf sansor matrix far a FMTSED wand be «me that has uniform
`and smatl retsrence vatues ever a growing bevel, a targa vartatécsn
`at intensity due tn 2: much, am: tssat mesassasnemant time.
`‘mes senwr
`rnatrix at the pretmypa, hzswesver, has a retsettvsty wtctas rsztagss at
`reefssrertcss vaitsem. i-mwevsr these wines :30 net change vary much
`esvssr extezrscsm pertzszflss at ttrme. The rasatts shaw that atasatzttng the
`mamtzer of sensors‘ in a grants in the column direction Imzreasetss tha
`restvarence value by at Sector at atmut L5. This mrzespondg wait to
`thasnrastimt asttmatass. As welt the rasutts Show that increasing the
`nmtber at Mangers in a gmus: tn the: raw dlmcttfian. Eh mntraai. dassss
`not Encraam the vetarenca values in gwmrat, even it tha number at
`the sensmxs Ls doamtaazt
`in :8: graup.
`"fha mfasrersces trains ranges
`mm 49 (tax: a single ssnsar in a grew) 1:: 58¢ (far that entire array
`of M by 32 aansars mnslrtarad as 3 group).
`in arm: is swam: me tima anti anther variations at the rstamnw
`vaimh. a thresssmtzs 5:5 inc!-«tact which musst be ekv-srmvm: tn cards: tut
`a tamh to ba detected‘ The thresmié used ranges: tram 2 in 7
`counts desvgzsndtng can gimp 352.6. ifistng these ttareshcid vaims the
`CPU doses not resgcxart untcsuczheefi paints wrmgiy we? Entarvais at at
`team 3 hours in «either esaszquaentix-ti er recwssiva sssstscttvixtcssa mamas.
`The: rmzststva 8i.th¢.iWiS.§t2fi schema uses 6 difiarent tttreshatdss, cone
`ssescatsentty E:
`is very mitkefy to re-gscrt s wmng mint whataas the
`aimsa! scanning mode using any a singts thmshcttd is iikeiy ta baa
`mam stemsstttve.
`
`‘ms tntssnstty oi as stngie 1-mx:h tar as stngta steamer group varies ever
`the tattle! but usuaity ranges amvts the thmsmid vaiue by as much
`as ‘.5. For a atngie €54
`try 32 sensor gram. the intenaity warm tram
`gsesmon to tzzsrson but it ranges tram the threshoid to "124. This ma»
`irnzum is abtatnm wheat‘:
`18 paint mitten thsan at flngesr twshesa the
`wheat. Aretha!
`Interesting featam 13
`that 936
`raspssnsa tlrm
`tsasemwbss
`tamer ass ms nummsr at sensors in as greases: bacumesss
`imger. and iurthnermnre that for the 5-3 by 32 aensar gmup, it is me»
`mate is 623993’! at a ham: mesnsty planed In thss takcintty at the tzshiast.
`
`Page 882 of 1714
`
`

`
`CHE ’85 KRQCEEDENGS
`
`APRii_ i§3§___§
`
`$inz:e- this spatial reszsazimiczn in me iscai iniarpuiatian sciierm is Hm»
`Red by the number ai bits: avaiiaitaia iram ma intensitias is? an may
`of 3 tag: 3 sensors. other scheme wag wnaidesrefi.
`in ibis seizema,
`ail am poims ircm as oompieia scan as? a issbisi are imesrpoiaim
`aiinwing the gsaieniiei reesesiuiissn in be aimasi infmitei iiawwm ihis
`process simpiy emuiaias 3 prnjamive tiaviae am aecsmiinggiy
`reepzms nniy sirxgies point, wiaich is intergszsissied {mm 4335 ihe minis an
`ihe taitsiai. Hawevvar wiiia ihis scheme, tiwra am 3 gm.-it marry ways
`:2? painting :9 a specific iucaiion on 2: disapsisy scream, a iaaiasre with
`some intriguing eappiicaiion szossiiiiiitiesn
`its Rsspaiws ‘Farm army
`The rassmnesa Sim siaiay is ins time zisiay {mm mes bagiriniiig 61 at
`mum in an campui recesivesd esithar by iocafi iemzinai ur by sen auigsm
`device aiiached E4: the hast esesmpuaat. Far imfiiipie iomma. ibis
`ziesiay wiii incraasa with the member or! imsches. The prdiatypss used
`with a was bauci-rates isrmimi
`ts: msaswe time aways. mnuai
`respansa iimss wera measxxed swam times; am swaragasfi
`iesr
`vazimzs cases and are ixabuiaied in Table ‘I.
`
`E(c) pts/sec
`41.6 E
`1 msacc/pi.
`mam ‘L Aciisaai Rassxanss Time Besiays
`
`The: mass in Tama one are: an be imerpretm a§ fleiiawsz
`a arm sanssr isigshefi esesminuewsiy
`E3 ma sensars iamhad as mes same aim ecminumsiy
`x: mm sansarsa inmheecfl an the same iims eemiramigiy
`
`&.
`
`(§€)§‘6CLi3$K3§£S5
`
`A szroicaiypvs vi a ias:-scanning imziiipia»-10ax:h~sensEs;*m inpui iabiei
`having bash tires ssdazsiasbiiiiy and fiaxiksiiity 70:’ as brand ranges ai assimi-
`miians has been designed and impiemersiad. Gapadianesa msaxw
`uremesm of individual sensoflsi which can be uniquely amdrsssad
`using me die-dssa par sensor, makes it possibia in sense: both time
`possiiions and iniesnsitiesss
`taf ms 02' mars simuiimeem ‘muciwas
`wiiimui ambigssiy. The scsnmr mairix is mraimiiexzi by Universiiy of
`Tammie: S8639 board whases ssariai port is onnrseseyied i<> ones evi ihas
`Elf} parts of 1: hassi CO|'i‘i§iLiE13i'. Softwam iisai miiizem ‘me mcursivs
`subdivision aigoriiiim in: 1%! scannireg an army :2: 54 by 32 sensera
`on the iahise, and ma‘: ccmrmnicama wiiia im imsi assrnpuissr, has
`been imgsiesrrienietci am: tasked.
`
`%. A€2XN¢fiWLE§§€iE§EN‘?§
`
`The research described in flhis paper has izsen imdaxi by {he
`Nsaimai Scisncas and Engineering i§awsza.'s:h Gcuncii ai ifianada.
`This sipped is grsiesfasiiy asnkmwiadgm.
`‘Q. $"IEFERE§88§§
`
`Brawn. E, Bus-ztszsn. W. & Mmagh. K. (1985). Winsiewa an Taisiesis as
`a M93343 <21 Achieving Vimsi iwpua Devicas. C-amputer Sya»
`isms Rss-sarm insiiima, University :31’ Terame.
`
`Emma. W. (1982:. Lexicai and Prsgwmie: cansieiemtiam of mu:
`Simsiures, Conrpuier Graphics, ‘I? {*3}. 31- 3?.
`
`§3i5.ii.i8£