JIPDEC
Informatization
Quarterly
1995
一r「豆
本
-「
1
冤
η7egt/7Jaρan-Germany
Forumon,nformat'on
Techno,ogy
JIQNo.101
●1
JIPDECInfOrmatization
Quarterly
1995
JIPDECInfbrmαtizationQuαrterly
(JIQ)ispublishedquarterlybytheJa-
panInformationProcessingDeveloP-
mentCenter(JIPDEC),KikaiShinko
KaikanBldg.,3-5-8Shibakoen,Minato-
ku,Tokyo105Japan.
Publisher:EijiKageyama,President
Editor:YujiYamadori,Director
Research&International
Affairs
JIPDECisanon-profitorganizationfounded
in1967withthesupportoftheMinistryof
InternationalTradeandIndustry,theMin-
istryofPostsandTelecommunicationsand
relatedindustrycirclesf()rthepurposeof
promotinginfbrmationprocessingandthe
informationprocessingindustryinJapan.
NOTE:Theopinionsexpressedbytheva百 一
〇uscontHbutorstotheJIPDECInforma-
tizationQuarterlydonotnecessarilyrenect
thoseviewsheldbyJIPDEC.
CONTENTS
*FromtheEditor』 1
*MultimediaandHumanCommunication ._.3
*CurrentActivitiesforMultime(lia
Communications
*DevelopmentofMultimediaServicesfor
9
CurrentandFutureBroadcastingMedia__14
*CollaborationandCommunication
-Avideoconferencesystemwith
facilitiesformultipleeye-oontact- 20
*Multi -MediaPCCo㎜unications
*SuperHighDefinitionImageApplicationfor
MedicalUse
27
*PersonalInformationTeTminals
30
Copyright1995byJapanInfbrmationPro-
cessingDevelopmentCenter.
Nopartofthispublicationmayberepro-
ducedwithoutwrittenpermissionofthe
publisher・
*PortableMultimediaMachinefor
TheBusinessPerson
36
42
*MultimediaHomeApplianceanditsTrend ..48
*TheRealWorldComputingPmgram51
*MassivelyParallelComputerRWC-1
*CooperativeInteractionsinFuzzyDecision
SupPortSystems
60
TranslatedandPrintedbyTheTranslation
InstituteofTechnology,Science&Culture
PrintedinJapan,March,1995.
*Neuro.FuzzyApplicationforIntelligent
System
'64
71
κEtR'NThisworkwassubsidizedbytheJapan
oo KeirinAssociationthroughitsPromo・
tionfundsfromKEIRINRACE.
*ConsumerDecisionSupPortSystemin
VirtualSpqceUsingKanseiEngineering.._77
*InterSpace:TowardsNetworkedReality ..._88
*CurrentNews 91
No.101
●
●
FromtheEditor
Variousmediacreatedbythedevelop-
mentofinformationandcommunica.
tiontechniquesarecallednewmedia
inJapan,asoPPo3edtoconventional
mediasuchasnewspapers,TV,and
radio.Newmediaisconsideredtoplay
ahimportantroleindevelopingthein-
fbrmatizedsociety,andtheJapanese
governmenthasbeentakingvarious
measuresinitsregard.MITI,fbrex-
ample,ispromotingaNewMediaCom-
munityplan,andtheNewMediaDe-
velopmentAssociationwasestablished
asabasefbrpromotingthisplan.The
MinistryofPostsandTelecommunica-
tionsconstructedaTeletopiaplan,and
theMinistryofAgriculture,Forestry
andFisheriescreatedaGreentopia
plan.UndertheNewMediaCommu-
nityplan,certaincitiesinJapanwill
bedesignatedasmodelareas,where
highlyinformatizedsocietieswillbe
constructedthroughnewmediasuch
asopticalfiber,theinteractiveCATV
usinghigh-speedcommunicationciト
cuitssuchassatellitecommunication,
Videotex,anddatabaseservices.This
projectbeganin1984toreducethe
informationgapamongdifferentareas.
Thehomeelectronicsindustry,suchas
computer,te】ephone,andTV,isbeing
integratedwiththeelectronicsindus-
tryusingdigitalizationandcreating
mediawithnewfunctions.Thismedia
iscalledMultimediaandhasbeenat-
tractingmuchattentionin士ecentyears.
TheMinistryofPostsandTelecommu-
nicationsestimatesthatthemarketfor
Multimediawillbecome123trillionyen
in2010,reaching5.7%ofthegrossna-
tionalproduct.TheMultimediamar-
ketwillalsoprovidenewjobsforsome
2.4millionpeople.Problemsremainto
besolved,however,suchasthehuge
expensesrequiredfbrdevelopingthe
infrastructure,suchasopticalfiberfa-
cilities,a(ljustmentatministriesand
agencies,andthecomplexintellectual
propertyrightsissuesrelatedtoMulti-
media.
TheJapan-GermanyForumonInfor-
mationTechnologywasheldonceayear
from1984,thenlaterheldonceeach
yearandahalf,inJapanandGerma-
ny.Leadersfrombusiness,academic,
andadministrativefieldsinthetwo
countriesparticipateinthisforumto
deepenmutualunderstandingand
friendlyrelationshipsintheinforma-
tiontechnologyfieldinbothcountries.
The9thforumwasheldinBeppu,Oita
Prefecture,inNovember1994.Atthis
forum,whichbeganonNovember8,
greetingsbychairpersonsandgovern-
mentrepresentativesfrombothcoun-
triesandkeynotespeechesweregiven
onthemorningoftheopeningday.
Later,thefbrumwasdividedintothree
workshops_computer,newmedia,
1 JIQNo.101,1995
andsemiconductor.Ateachworkshop,
expertsgavepresentationsfollowed.by
discussions.Onthemorningofthe
thirdday,Japanesechairsateach
workshopsummarizedthepresenta-
tionsandcompletedtheworkshops.At
thisgthforum,apanelsessionby
speakerswasheldforthefirsttimeat
thecomputerworkshop.Astopicsfor
presentationsateachworkshop,the
mostcontroversialissueswerechosen
bycorememberswhoareprofessionals
ineachfieldinbothcountries.Atthis
f()rum,presentationswereheldregard-
ingthemultimediacommunicationser-
vice,apPlications,andterm{nalsfbrthe
newmedia6eld.Also,inthecomputer
field,maintopicswererea|-worldcom-
puting(RWC),whichisbeingdevel-
opedcurrentlyinJapan,paraUelpro-
cessingasabasisforRWC,neuro-fuzzy
theory,andvirtualreality.
AtthisJIQ,weintroducedpresenta-
tionsgivenbyJapaneseparticipants
atthenewmediaworkshopandthe
computerworkshopattheabovefo-
rum.Japan'sleadingtechnQlogiesare
introducedinthesereports.Wehope
thisissuewillbehelpfultoreaders
promoteunderstanding.Wewouldalso
liketotakethisopportunitytothank
theJapanesespeakerswhopermitted
ustopublishtheirpresentations.
YujiYamadori
Director
Research&InternationalAffairs
LlQNo.101,1995 2
MultimediaandHuman
Communication
MasahideKANEKOandHiroshiHARASHIMA
DepartmentofElectronicEngineering
FacultyofEngineering
TheUniversityofTokyo,JAPAN
Abstract 1.Introduction
"Multimedia"hasbeenusuallydis -
cussedfromthetechnologicalpointof
view,however,wethinkthatmultime-
diashouldbealsodiscussedfromthe
viewpointofitsrelationtohumans.
Multimediawilloffertheinformation
environmentwherehumanscanfreely
makefulluseofvariousinformationl
media.Thisnewenvironmentisclose-
lyrelatedtotheframeworkof"human
communicationtechnology"whichsup-
portsbetterandω αrmheαrtedcommu-
nicationamonghumans,andfurther,
betweenhumansandcomputers.In
ordertomakethispointelear,weaim
toward"human(e)media"ratherthan
theso-called"multimedia."Thispaper
discussesmultimediafromtheview-
pointofitsrelationtohumans(human
communication),andpresentsacon-
ceptofa"virtualhumanecity"which
isconstructedasanalmostcomplete
metaphorofarealcitybymeansof
multimedia.Thisconceptwillhelpus
toputinordervariousproblemsin-
volved .inmultimediatechnology.
Asweobtaininformationfromtheout-
sideworldthroughvarioussensessuch
asvision,hearing,touchandsoforth,
humanshavemadefulluseofmultiple
informationrepresentationmeans.In
thissense,theuseofmultimediais
quitenaturalforhumans.Sincethe
technologyhadnotmatureduntilre-
centdays,wehadtodowithmonome-
dia.Forexample,thetelephonecould
onlytransmitvoice,andcomputerhan-
dledonlycharacters.Remarkablede-
velopmentoftechnologyinthefieldsof
computerandtelecommunicationhas'
changedthissituation.Theuseofmul-
timediahasrapidlyprogressedinvari-
ousfields;computer,telecommunica-
tion,broadcasting,consumerelectron-
ics(especiallyaudio-visualdevices),
andpublishing.Now,"multimedia"is
expectedtoconstructthenewinfprma-
tionenvironmentwherehumanscan
makefulluseofvariousinformation
representationmedia.Heretheterm
"informationTepresentationmedia"
meansimages,video,sound,texts,and
soforth.Thisdiscussionmaybesum-
3 JIQNo.101,1995
marizedbythefollowingequation:
E=MC2,
whereE:mediaenvironment,m:mul-
timedia,andc2:computerandcommu-
リ コ
nlcatlon.
Thisnewinformationenvironmenthas
acloserelationtotheframeworkof
"humancommunicationtechnology."
Conventionalcommunicationtechnol-
ogybasicallyhandlesthetransmission
ofsignalsbetweencommunicationter-
minals.Ontheotherhand,thehuman
communicationtechnologyintendsto
supPortbetterandω αrmheαrtedcom-
municationamonghumans.
Thispaperdiscussesmultimediafrom
theviewpointofitsrelationtohumans,
andpresentsaconceptof``virtualhu-
manecity"whichisconstructedasan
almostcompletemetaphorofarealcity
bymeansofmultimedia.Thisconcept
willbeusefultoputinordervarious
problemsinvolvedinmultimediatech-
nologyandtheirapPlications.
2.BriefOverviewofMulti・
media
2.1Whatismultimedia?
Adefinitionoftheterm``multimedia"
maybedescribedasfollows:
-technologytohandleseveraltypesof
differentinformationrepresentation
mediaconsideringtheirmutualrela-
tionship.
・fusionor皿ifiedusageofmultiple
mediathroughthedigitaltechnolo-
gy.[Alltypesofmediacanberepre-
sentedbybits(Oor1)andcanbe
treatedwiththesamemanner.]
Multimediaisacompoundtechnology
anditissurelyexpectedtocausethe
followingtypesofintegrationamong
differentcomponentsinthenearfu-
ture.
(a)integrationofvariouskindsofinfbr-
mationrepresentationmediasuch
asimage,video,sound,text,andso
fbrth.Notonlythesimplecombina-
tionofmultiplemediabutalsomu-
tualsynchronizationofdifferentme-
dia,andfurther,manipulationand
mutualconversionareimportant.
(b)integrationofinformationtransmis・
sionchannelssuchastelecommuni-
cation,broadcasting,computernet-
work,andsoforth.
(c)integrationofdifferentindustriesor
businessworldssuchastelecommu-
nicationcompanies,broadcasting
stations,industriesofconsumerelec-
tronics,andsofbrth.Thismeansthe
removalofexistingbordersamong
differentindustriesandbusiness
worlds.
2.2Whatwillmultimediagive
us?
Multimediaisattractivesinceithas
hugepotentialstogiveusmuchmore
real,human-friehdlyandconvenient
informationenvironmentasshownbe.
JIQNo.101,1995 4
10W.
(a)morereal:reproductionofrealistic
sensationofprese'nceorfeelingof
sharingthesamespacewithother
persons.Here,thespacemeansphys-icalspaceandalsopsychological
space.Thishascloserelationtothe``virtualreality.,,
(b)morehuman-friendly:reductionof
human'sloadtocarryoutcommuni-
cationwithotherpeopleorcomput-
ers.Thishasacloserelationtothe"humaninterface ."
(c)moreconvenient:enablingustomake
fulluseofvariousmediaatanytime,
atanyp}ace,andwithanybody,Im-
portantfactorsareinteractivity(bi-
directionalcommunication),mobili-
tyandpersonaluse.Here,the"con-
venience"meanstwodifferentas-
pects:oneisthatausercandoevery-
thingwithoutmovingandtheother
isthatausercandoeverythingeven
ifmoving.
3.HumanCommunication
Technology
Conventionalcommunicationtechnol-
ogyhasbeendevelopedtorealizefast-
erandmo .rereliabletransmissionof
signalsbetweencommunicationtermi-
nals.Ontheotherhand,theabove-
mentionednewinformationenviron-
mentgivenbymultimediawillleadto
"humancommunicationtechnology,"
whichplaceshumansonthecenterin
thedesignanduseofcommunication
systems.ItsupPortsbetterandwαrm-
hearte(icommunicationamonghumans,
andmoreover,human-friendlyinter-
facebetweenhumansandcomputers.
Principalsuhjectsrequiredinhuman
communicationtechnologyareasfol-
10WS:
(a)multimedia(innarrowsense);ex.
inputandoutputofmultimediain-
f()rmation,analysisandsynthesisof
multimediainformation,multime-
diadatabase,hypermedia.
(b)advancedsemanticcommunication
theory.
(C)intel}igentCOmmUniCatiOn;ex.皿 -
derstandingofmeaningandinten-
tion,acquisitionandusageofknowl-
edge,valueaddedcommunication,
agentsystem.
(d)humaninterface;ex.humanmodel,
man-machineinte㎡face,multimedia
interface.
(e)representationandcodingofinfbr-
mation;ex.structuraldescription,
compression,accumulation,andre-
trievalofvariousinformation,intel-
ligentcoding.
(f)supPortf()rhuman,scommunication;
ex.supportf()rhandicappedoraged
people,supPortforcreativeactivi-
tiesandthought,supportforcollab-
orativework.
4.Virt岨lHumaneCity
Therearemanyaspectsin"multime-
5 JIQNo.101,1995
dia"includingtechnology,services,law
andregulation,humanfactors,psychol-
ogyandsoforth.Untilnow,theuni-
fiedandwideviewedscopeonmulti-
mediahasnotbeennecessarilypre-
sented.Ifwepayattentiontothefact
ofwhatmultimediacanoffertous
throughthisnewinformationenviron-
ment,aconceptof"virtualhumanecity"
willhelpustothinkoverthevarious
aspectsofmultimediasystemat{cally.
Theabovediscussionabouthurnancom-
municationtechnologyis,ifanything,
anitem-by-itemdiscussionanddirects
to・individualpersons.Ontheother
hand,theconceptofa"virtualhumane
city"directstoasocialsystemandre-
1ationshipbetweenhumansandaso-
cialsystem.Theterm"virtualhumane
city"isnewlypresentedhere."Virtu-
al"meansthatthiscityexistsvirtually
incomputersandcommunicationnet-
works."Humane"meansthatthisvir.'
tualcityisconstructedforhumansand
notformachines.
Byusingmultimediatechnologywith
computersandcommunicationnet-
works,itispossibletocarryoutsever-
alSOCialaCtiVitieSOneleCtrOniCnet-
worksinsteadofdoingtheminthereal
world.Typicalexamplesaretelecon-
ferencing,teleshoPPing,videoonde-
mandserviceandsoforth.Several
ideastobuildvirtualspacewhichac-
ceptsmultipleparticipantsonelectron-
icnetworkhavebeenalreadyproposed:
f()rexamplecyberspace(USA)[1],the
VirtualPolice(CarnegieMe}lonUniv.,
USA)[2],andInterSpace(NTT,Japan)
[3].
Theconceptof"virtualhumane
city"proposedhereisoriginatedfrom
thesimilaridea.Butwe'dliketoplace
emphasisond輌fferentaspectssuchas
competitiveprincipleofeconomics,free
economy,chargingmethod,andregu-
lationbylawinadditiontothetechno.
logicalproblems.Thevirtualhumane
citycanbeconsideredasanalmost
completemetaphorofrealcities.Not
onlythecommunicationmechanism
throughcomputersandcommunication
networksbutalsothedesignandcon.
structionofvirtualcityspaceareim-
portant.
Arealcityincludesvariousaspectsof
thehumandailylife.Avirtualhu-
manecityalsoincludesalmostthesame
variousaspectsofhumanlifeinareal
city.Thebasicconditionstoconstruct
avirtualhumanecityarebrieflysum-
marizedasfollows:
(1)GUI(graphicaluserinterface)torep-
resentcityspaceandtosupPorteasy
accesstoit.ThisGUIisusedasan
interfacebetweenauserinareal
worldandavirtualcityconstructed
incomputersandcomputernet-
works.
(2)Ensuringthefreewalkspace(prom-
enadewhereonecanwalkaround
freeofcharge).
(3)Sufflcientfunctionsandservicesf()r
dailylife;market(shop),office,
amusementplace,school,park,li-
braryandsoforth.Thevirtualhu-
manecityisdividedintoth3sub-
spacescorrespondingtovarioushu-
JIQNo.101,1995 6
ひ り ら ら らmanaCtlVltleS;COmmUmCat10nzOne,
marketzone,businesszone,school
zone,amenityzone,publicservice
zone,transportationzone,andso
{brth.
(4)Regulationbypubliclawandprinci-
plesofeconomicswhicharesimilar
tothoseinarealcity.
(5)SupPortsystemsrelatingtosecurity
(ex.thepolicesystemtoguarantee
safety;self-defensesystem),guidein
virtualhumanecity,andsofbrth.
Inadditiontoabove,oneimportant
thingtobekeptinmindistheneed
f()rfirstrankarchitects,whocan
designanattractiveandcomfort-
ablevirtualhumanecity.
Competitiveprinciplesofeconomics
shouldbeintroducedinavirtual
humanecity,sothatthecitywill
haveafreemarketandchangedy-
namicallyaccordingtoitsactivities.
Here,twotypesofcompetitioncan
beconsidered.Oneisthecompeti-
tionamongdifferentvirtualhumane
cities.Theymaycompeteagainst
oneanothertoincreaseresidents
andvirtualshops(serviceproviders)
ineachcity.Conseqμentlythewin-
nerwillexpandtheactivitiesand
cityarea,butadefeatedcitywill
decline.Theotheristhecompetition
betweenavirtualcityandarealcity.
Auserwillchooseavirtualhumane
cityorarealcityaccordingtowhat
onecandoineachcity.Twokindsof
citieswmsharetherole.Thatis,a
userwillliveinbothcities,andthis
willcausetheintegrationofliving
space・
Byemployingthisconcept,itiseasy
toputinordervariousproblemsin-
cludingcharging,competitionin
marketsandsoforth.Usuallycharg-
ingisaverydifficqltanddelicate
problem.Applicationofthecharg-ingmechanisminarealworldtoa
virtualhumanecitywillhelptode-
cidethechargingprincipleintheuse
ofmultimedia.Residentscanwalk
aroundinpublicspacefreeofcharge.
Theycanenjoywindowshopping.
Whenonecarriesouteconomlcac-
tivitiessuchastobuysomething,to
sellsomething,toopenone,sshoP
andsoforth,onewillbecharged.
Severaltypesoftaxsystemsmaybe
introduced.Thisconceptisquite
differentfromthechargingmethod
employedinthecurrenttelecommu-
nlCat10nServ1CeS.
5.FromMultimediatO
Human(e)Media
Asarealcitycannotexistwithoutres-
idents,avirtualhumanecityiscon-
structedontheassumptionthatthere
areresidentsoruserswholiveinit.
TheresidentswiUcommunicatetoone
anotherovervariouskindsofactual
gapssuchasdistance,physicalhandi-cap,language,age,cultureandsoforth.
Moreover,theywillenjoythehuman
friendlyinterfacetocommunicatewith
computersystems.
Multimediashouldsupportthehuman
lifeandactivities.Thenewinforma-
tionenvironmentobtainedbymultiー
7 JIQNo.lo1,1995
mediawillleadtotheframeworkof"humancommunicationtechnology
."
Toputemphasisonthis,wewouldlike
tointroduceanewterm"human(e)
media(human-centeredmedia,human-
orientedmedia)."Multimedia"dis-
cusseduntilnowisratheratechnical
term.Ontheotherhand,"human(e)
media"impliesaclearsenseofpurpose
todiscussmultimediawithrelationto
humans.Sofarmultimediahasbeen
discussedfromtheviewpointofwhat
wecandobyusingit.However,now
wehavetothinkoverofwhatweshould
dobymultimedia(humanemedia)and
whyweuseit.
thermore,human-friendlyinterface
betweenhumansandcomputers.The
newconceptof"virtualhumanecity"is
presentedtoputinordervariousprob-
Iemsinvolvedinmultimediatechnolo-
gy.Finallywehaveintroducedthe
newterm"human(e)media"withthe
clearsenseofpurposethatmediatech-
nologyshouldalwaysdirecttohumans,
andshoUldalwaysbehumanetohu-
mans.
References
[1]M.BenedictEd.:"Cyberspace:first
steps,"MITPress(1992).
6.Conclusion
Inthispaper,wehavediscussedabout
"multimedia"fromtheviewpointofits
relationtohumans.Theinf()rmation
environmentwhichisindispensablefor
uswillbeconstructedbymeansofmul-
timediaanditwillsupportdailyhu-
manlifeandactivities.Thisnewenvi-
ronmentleadstotheframeworkof"hu-
mancommunicationtechnology,"which
supPortabetterandω αrmheαrte(icom-
municationamonghumans,andfur-
[2]C.E.Loeffler:"VirtualPolice:anet-
workedvirtualrealityapplication,"
Proc.ofIEEEandIEICENR,94
(NetworkedReality),3-1(May1994).
[3]S.Sugawara,G.Suzuki,Y.Nagash-
ima,M.Matsuura,H.Tanigawaand
M.Moriuchi:"InterSpace:networked
virtualworldforvisualcommunica-
tion,"Proc.ofIEEEandIEICENR'94(NetworkedReality)
,3-2(May
1994).
JIQNo.101,1995 8
CurrentActivitiesfor
MultimediaCommunications
SadamiKURIHARA
NTTMultimediaPlanningandPromotionOf資ce
1.Introduction
CurrentNTTnetworkprovidesavari-
etyofservicesincludingbasictele-
phones,leasedcircuits,packet-
switcheddatacommunications,pag-
ersandISDNservices.Ithasevolved
fromtheoriginalanalogPublic
switchedtelephonenetwork(PSTN)
intoasetofsubnetworkswhichnow
suiteachdemandofspecificservice.
Thenumberoftelephonesubscribers
inJapanhasgrownrapidlyinthelast
30yearsandithasnowreachedas
manyas58millions.Amajorshiftin
telecommunicationsisnowinprogress,
however,asthedemandfbrthecom-
montelephoneserviceisbecomingsat-
urated,andwidespreadinterestisbe-
9inningtofocusonnewcategoriesofservicessuchashigh-speeddatatrans-
missionamongcomputers,multime-
diacommunications,mobilecommuni-
cationsandinteractivevideoonde-
mand(VOD).
Withthosetransformationsofcommu-
nicationsnetworksinmind,NTThas
releasedastrategytopromotethede-
velopmentofmultimediacommunica-
tionsservicesasofthisJanuary.This
presentationrevealsourcurrentactiv-
itiesasthefirststepaimingfbrthe
earlyarrivalofthecomingmultimedia
age.First,R&DactivitiesfOrthefu-
turemultimediacommuhicationsare
listed.Secondly,thepilottestsofmul-
timediaservicesusinganationwide
highspeedbackbonenetworkbasedon
ATMtechnologyandopticalfiberdata
transmissiontechnologyareoutlined,
whichaimforpromotingthedevelop-
mentofnewinformationprovidingser-
vicesandhighspeedcomputercommu-
nications.
2.ExamplesofR&DActivi・
ties
Therearetwotypesofmultimediaap-
plications,Overlayingadvancedfunc-
tionstoexistingtoolsisthefirsttype.
Offeringnewtoolswhichcreatenew
workstylesistheother.PersonalMul-
timedia-multipointTeleconferencesys-
tem(PMTC),shownasanexampleof
thefirsttype,addsthevideotelecon-
ferencefunctionstoPCsorWSswhich
arealreadyusedformanypurposeson
thedesk.TeamWorkStation(TWS)
andClearBoard(CB)areexamplesof
thesecondtype.Theyofferusersa
9 JIQNo.101,1995
cooperativeworkspaceonthedesktoP
byapplyingvideocornmunicationfunc-
tions.
Inthefuture,wewiUusecomputers
andcommunicationsystemsmoreof二
tenasinformationaccesstools.Inthis
case,multimediapresentationtech-
niquesmustbeeffectivelyapplied.The
ISDNvisualinformationsystem(IVIS)
hasbeendevelopedtooff>rthevisual
informationserviceforvideophones.
Thedigitalvideoresponsesystem(D-
VRS),whichstoresanddeliversdigital
videoinfbrmation,hasbeendeveloped
tooffervideoinformationservicessuch
asthevideoondemandserviceswhich
cantransmitanyvideoprogramwhen-
everrequested.
Thosesystemsaredescribedbelow.
2.1Personalmultimedia・multi
pointteleconferencing
system[1]
ThePMTChasbeendevelopedwith
fourdesignconcepts,whichareoffer-
ingnewconferencelocations,creation
ofvirtualconferencespacesondisplay
screensofWSsorPCs,apPlicationof
multimediainformationtechnologiesto
makedocumentsforconferences,and
offeringnewenvironmentsforgroup
works.
Therearethreemainfeaturesofthe
PMTCmultirnediauserinterface.The
firstfeatureisthevirtualconference
spacescomposedwithcommonspace,
closedspace,andlocalspace.Thesec-
ondpointisthatnotonlytheconferees
butanymaterialcanbeeasilycap-
turedanddisplayedthroughoutthecon-
ferenceusingthemulti-windowformat.
PMTCisuser-friendlybecauseital-
lowsustocontroltheconferencewin-
dowswiththesametechniquesweuse
tomanipulateoutregularfiles.The
thirdisuseofaudiowindows.These
helpustorecognizewhichparticipants
arespeaking.Theyalsoheightenthe
senseofattendingtheconference.If
theimageofacertainpersonapPears
ontheright-handsizeoftheworksta-
tiondisplay,thenhisvoiceshouldcome
fromtheright-handside.Intheseways,
thissystemprovidesenhancedmulti-
mediacommunicationservices,andwe
areabletoparticipateinateleconfer-
encewhileremainingatoutowndesk.
2.2TeamWorkStation[2]
TeamWorkStationisamultimediaap-
plicationsystemthatallowsgeograph-
icallyseparatedpeopletoshareeach
other'sdesktopsbymeansofvideocom-
municationfunctions.Whenwelook
atthetasksweperfbrmatourdesks,
wefindthatwewriteanddrawoften.
Wealsofindalotofcooperativetasks,
forinstance,designmeetingsandin-
structionalmeetings.Thissharingof
desktopshasseveralimportantimpli-
cationsthathavepassedunnoticedso
far.Forinstance,ourdesktopbe-
comesextended.Itcontainsnotonly
ourmaterialsandfiles,butalsothe
materials,suchasprintedbooks,and
負lesofourpartner.
Majorcomponentsareasfbllows.Each
userhastwovideocamerastocapture
JIQNo.101,1995 10
thedesktopandtheuser.Onlyone
monitorisneededtodisplayTeam
WorkStationimagesandthoseofthe
user'scomputer.Theuseralsoneeds
codingequipment,avideoCODEC,arid
aTeamWorkcontrolboxwhichcom-
binesdifferentimagesusingtranslu-
centoverlays.AnISDNnetworkor
LANmightbetheconnectingbridge.
Thewaysoftheoverlayfunctionin
thissystemarerealizedasthefollow-
ing.First,videoirnagesofthefaceand
thedesktoparecombinedusingthepic-
ture-inpicturetec}mique.Eachuser
thentransmitshiscombinedimage.
Uponreceivingthecombinedimage
fromhispartner,theuserconvertsit
intoatranslucentimagewhichisthen
overlaidonhisscreen.Ineffect,both
sidessharethesamescreenimage.
2.3ClearBoard[3]
isusedtodisplaytheremoteimageon
thebackofthescreen.Bothpartici-
pantscanseetheirpartner'sfaceand
workasitlaysundertheirownwork.
Face-to-facediscussionsbecomevery
easyandtheworkflowisnotinter-
ruptedbytheneedto.Switchthegaze
direction.Inthisway,ClearBoardal-
lpwstwopeopletohaveaconversation
whilevirtuallyannotatingtheirpart-
ner'sdrawing.
TheoriginalClearBoard,ca11edClear-
BoardI,usesmarkingpenssotheim-
agesaremademanuallyandaredi笛 -
culttomanipulateandrecord.Acom-
puter-baseddrawingsystemhasbeen
addedtoClearBoardII.Theinclined
screenisactuallyadigitizerthatover-
laysalargeLCDdisplayscreen.Clear-
Boardllallowspersonalcomputerfiles
tobeplacedonthescreenandtobe
sharedandmodifiedeasily.
TeamWorkStationisquiteadvanced
butwehavealreadycreatedanalter-
nativecommunicationtoolnamed
ClearBoardinsomesuperior.Thecon-
ceptofClearBoardistorealizeseam-
lesscommunicationbetweenconversa-
tion,writinganddrawinginthemeet-
ing.Whenconferenceparticipantslook
attheblackboardtheydonotseetheir
partner'sface.Lookingateachother
makeswritingimpossible.Thisprob-
lemiscompletelysolvedbyClearBoard.
Thehalf-mirroraUowstheparticipant
tobephotographedbytheoverhead
camera.Thiscameraalsorecordsthe
informationdrawnontheboard.The
videosignal,withaudio,istransmit.
tedtotheirpartner.Avideoprojector
2.4Visualinformationsystem
[4]
TheISDNvisualinformationsystem
havebeendevelopedtooffervisualin-
f()rmationservicesbyusingvisualter-
minalsthroughISDN.Thissystemis
composedofavisualinformationcen-
tersystemandvisualterminalssuch
asvideophonesorteleconferencingter-
minals.Examplesoftheservicesin-
cludebulletinboardservices,message
services,mailservices,andgathering
boardservices.Thevideoinfbrmation
centeriscomposedofageneralpur-
poseworkstation,videoinformation
storageequipment,andvideoinfbrma-
tiontransfercontrolequipment.The
11 JIQNo.lo1,1995
maximumnumberofchannelsfbrsi.
multaneousconnectionsis50.Termi-
nalspeedsrangefrom64Kto1.5M.
Thecentercanholdaudiovisu .alinfor-
mationequivalentto500hoursonvid-
eo.Thissystemisactuallyaimedat
videophonesandteleconferencingsys-
tems.
3.MultimediaUtilization
PilotTests
NTTispromotingtwotypesofspecific
experimentalplans.Thoseplanswill
beconductedwiththecooperationof
customers,manufacturesandinforma-
tionserviceproviders.
2.5Digitalvideoresponse
system[5]
3.1Pilottestofhighspeed
backbonenetwork
Amoreadvancedsystem,thedigital
videoresponsesystem,hasbeendevel-
oped.Thedigitalvideoresponsesys-
temprovidesretrieval-basedvideowith
broadcastorvideoquality.D-VRSis
composedofavideoinformationcen-
ter,B-ISDNandvisualterminals.The
functionsofthecenterincludesvideo
informationstorageanddeliverycon-
troltorealizethesameservicesasvid-
eo-ondernandserviceswhichisavery
pullernow.
Thecontrolequipmentinstalledinthe
centeracceptstherequestthroughthe
communicationnetworkanddelivers
thedigitalvideodatathroughB-ISDN.
Itcancontrolvariousvideohandling
functionssuchasstart-stop,slow-quick
andforward-reverseforeachuser.
Hierarchicalmernorycontroltechnolo-
giesperformedwithopticaldisklibrar-iesandarraydiskshavebeenadopted
inthestoragesystemtorealizemulti-
mediaserviceseconomically.Forex-
ample,pub}icnewsortowninforrna-
tionwillbeholdonarraydisks,where-
asvideolibrarieswillbestoredasopti-
caldiskfiles.
Respondingtouserneedsforcomputer
communications,NTTiscloselyin-
volvedintheATMtechnologydevelop-
mentandisnowatthestageofcom-
mercializingit.Utilizingacombina-
tionofATMandopticalfibertechnolo-
gies,NTTwillconstructahighspeednationwidebackbonenetworkaspart
ofitsintracompanynetworksoperat-
ingatagigabitlevel.Itisbuilton
SynchronousDigitalHierarchy(SDH)
transmissionsystemsatthespeedsof
2.4Gb/s,10Gb/s,etc.,andprovides
varioustypesofUserNetworkInter-
face(UNI)upto156Mb/stousers.
Throughthisproject,NTTwillestab-
lishconstructionandmanagementtech-
nologiesaswellasnewapPlicationsfor
thebroadbandnetwork.
About90projectsareplanningtojoin
thepilottest.Apartofthepilottests
willstartfromthisSeptember.
3.2Pilottestofmultimedia
serVlces
Duetoadvancesintechnology,inter-
activevideocommunicationssuchas
videoondemand(VOD)havebeende-
JIQNo.101,1995 12
velopedandareexpectedtoresultina
greatmarketdemand.Underthisback-
ground,NTTwillconductpilottestson
integratedsystemsthatprovidecom-
municationscircuitstocabletelevision
companiesandhighlysophisticatedser-
vicessuchasVODalongwithitsexist-
ingtelephoneservices.Theexperiment
targetingordinaryhomeusers,which
reachthenumberofseveralhundred
inafewareas,willbelaunchedin1995.
Thetestprogramincludestheprovi-
sioningcircuitsforcabletelevisioncom-
panies,atrialtestofVOD,andtele-
phoneservices.
References
[1]T.Nishimura,S.Masaki,H.Yamagu-
chi,H.Ichihara,T.Sakatani,andS.
Ishibashi:"SystemDesignofPer-
sonalMultimedia-multipointTele-
conferenceSystemforN-ISDN",
Proc.19931EICEFallCon£,D-220
(1993)
[2]H.Ishii,andN.Miyake:`『rowardAn
OpenSharedWorkspace:Computer
andVideoFusionApproachofTeam
WorkStation",Communicationsof
theACM(CACM),SpecialIssueon
CSCW,ACM,Vol.34No.12,pp.37-
50(Dec.1991).
[3]H.Ishii,M.Kobayashi,andJ.Gur-
din:``IntegrationofInter-personal
SpaceandSharedWorkspace:Clear-
BoardDesignandExperiments",
ProceedingsofCSCW'92,ACM,
PP.33-42(Nov.1992).
[4]KTamura,C.Matsuda,H.Tsuchiya
andM.Maruyama:"AStudyofthe
DataChannelControlMethodinVi-
sualMessageHandlingSystem",
Proc.ofthe19921EICEspringConf.,
B-696,(1992).
[5]H.Sakamoto,andH.Nakano:"Vid-
eo-on-DemandServiceSystem",To-
kai-SectionJointConventionRecord
ofthe61nstitutesofElectricaland
RelatedEngineers(1993).
■
13 JIQNo.101,1995
DevelopmentofMultimedia
ServicesforCurrentand
FutureBroadcastingMedia
AkioYANAGIMACHI
NHKScienceandTechnicalResearchLaboratories,Japan
、
1.Introduction
Tobecompetitivewithotherareasof
theelectronicinformationmediain-
dustryinthemultimediaera,broad-
casterswiUneedtointegrateadvanced
featuresofmultimediaintotheirser-
vices.
ISDB(lntegratedServicesDigital
Broadcasting)hasbeenproposedas
thebroadcastinginfrastructureforthe
comingyears,whichistotaldigital
broadcastingandintegratesmany
kindsofmultimediaservicesinto
broadcastingchannels.Meanwhile,as
animportantsteptowardtherealiza-
tionoffullscalefuturemultimediaser-
viceswithISDB,broadcasterscanim-
plementbasicmultimediabroadcast-
ingservicesbyusingthedatachannel
capacityofcurrentbroadcastingchan-
nelS.
2.EnhancementofBroad・
castingServicesina
MultimediaContext
Itispredictedthatinformationtrans-
missionandexchangetechniquesin
thebroadcast,communications,and
computerareas,supportedbydigital
technology,wiUdevelopdramatically
overthenextcoupleofdecades.The
computerarea,inparticular,willsee
veryrapidimprovementinperformance
andfusionwithmultimediaservices,
anditwillalsohaveastrongimpacton
otherareas.Inthecommunications
area,plansareunderconsiderationto
connectallhouseholdstoahigh-speed
digitalnetworkbyopticalfibers
throughwhichvariousserviceswillbe
provided.
Iftheabove-statedpredictionsand
plansarerealized,allinformationme-
diaindustriesrelatedtobroadcasting
willしmdergoradicalchanges.Theprop-
erroleofbroadcastingofofferingaccu-
rateinfbrmationtoalargenumberof
peoplesimultaneouslyatreasonable
cost,andthesocialandculturalneeds
fbrsucharolewillcontinueunchanged.
Theaudience,meanwhile,willgradu-
allygrowaccustomedtotheeveryday
practiceoftakingtheinitiativeinac-
cessingPersonallynecessaryorpre-
ferredinfbrmationbymeansofcom- ヨputers,communicationsservices,and
LlQNo.101,1995 14
videopackages,allhavingmultimedia
fUnctiona】ity.Itshouldbenotedthat
theaudiencewillbemoreactivethan
atpresentinselectinginformation.
Thisimpliesthenecessityofreviewing
broadcastingservicesfromthe皿bi-
asedstandpointofactiveusersalong
withthatoftraditionalpassiveaudi-
ences,torenewtheconceptionofbroad-
CaStingServiCeS.
Undersuchcircumstances,broadcast-
erswillneedtointegrateadvancedfea-
turesintotheirservices,suchasinter-
activefunctionsandreinforceduser-
orientedattributes.Thiscanbeimple-
mentedbybroadcastersremodeling
theirindustryintoISDBthatalsoac-
commodateswirednetworksinaddi-
tiontothebasalradiotransmission
media.
3.ConceptsofISDBMulti・
mediaServices
Describedbelowaretheprimarycon-
ceptsofbroadcastingservicesusing
ISDBwhencombinedwithwirednet-
works.
(1)Accessibletoanyone,anytime,any-
where
(2)Userfriendly
(3)Dynamicrepresentation
(4)Real-timeadvantages
(5)Interactivewithusers
4.LayerModelforBroad・
castingServices
Figure1(a)showsthecompositionof
currentbroadcastingservicespresent-
edinalayerstructure.Thisfigure
indicatesthatatpresentone-waytype
programsarebroadcastedusingthe
SDTV(NTSC,PALorSECAM)or
HDTVsystemasrepresentationmedia
bymeansofwirelessanalogtransmis-
sion.
Incontrast,thestructurefbrfuture
multimediabroadcastingservicesisil-
lustratedinFigure1(b).Transmission
willbedigitalizedtofacilitatethefunc-
tionalityofupPerlayers,aswellasto
giveflexibilityandexpandabilitytobroadcastingservices.Anewlayerwill
beintroducedthatwillfunctiontosup-
porttheselectionofprogramsandin-formationbyusers.Serviceswillbe
reinforcedwithdynamicrepresentation
functionsincludedintherepresenta-
tionlayer.Intotheprogramlayer,func-
tionswhichofferinteractiveprograms
willbeintroduced.Interactivepro-
gramswillbeavailableintwotypes:
pseudotwo-wayprogramsusingtermi-nalfunctionsandactualtwo-waypro-
gramsusingnetworks.FigUre2illus-tratesthetwotypesofinteractivityin
thebroadcastingservice.Inthefbrmer
type,userswillinteractivelyaccessa
varietyofinformationbroadcastand
StOreitinthereCeiVermemOrieS.In
thelattercase,userswillbeconnected
toabroadcastingstationthroughatwo-
waynetworkandwillinteractivelyac-
cessinformationorparticipateinpro-
9ramspreparedbythestation.
5.ServiceandProgram
Images
Consideringtheconceptsandstructure
15 JIQNo.lo1,1995
Future
Present
ProgramsOne・waylineartype
RepresentationSDTV(NTSC,PAL,SECAM),HDTV
TransmissionAnalog,Wireless
Programs
One-way
lineartype
Pseudotwo・way
interactive
RepresentationSDTV,HDTV,3D・TV,Data(Text,Tele-
fax,Computergraphicsimages,etc.)
UseraccesssupPortProgramguides,lndex,Content
attributes,Hyperlink,etc.
TransmissionDigital,Wireless,Cable
咽
FigurelCompositionofBroadcastingServices
Pseudotuハ0-U/a凹interactive
\
t‡ ‡⑧ 翻 ㊥
下山0一 山ayinteractiVe
0∫9'tal
/bro∂dcasting\
・鷲
Tei・minal
仁 ・f・ '-User/翻 Audience
↓ ↓㊥ ④ ◎
↓⑧
Figure21nteractivityinBroadcastingServices
JIQNo.101,1995 16
describedabove,wehaveimaginedthe
servicesandprogramsofISDBand
theirutilizationbyusers,f()rthetime
around20yearsfromnowwhenISDB
issupPosedtobefiユllyoperationaland
theimageswillreachfruition.Even
afterthistimehaselapsed,theway
peoplegenerallywatchbroadcastpro-
gramsmaybeunchanged,i.e.,mostly
theso-called`couchpotato'style.One
ofthegoalsoflSDBistomakeiteasier
andmorecomfortableforthe`couch
potato,userstoaccessprogramsand
information.Meanwhile,ISDBwi}1re-
spondtotherequestsofusersoncethey
becomeactive.
6.Requirementsforthe
ISDBSystem
Theconsiderat{onsdescribedabovelead
tothefollowingrequirementsforan
ISDBsystem.
(1)Reliabletransrnissionthroughvar-
ioustypesoftransmissionchan-
nelsincludingsatellite,terrestrial
andcable.Figure3illustratesthe
relationbetweenthetransmission
channelsandpossibleservicesto
becarried.
(2)Flexibleintegrationofvarioustypes
ofservices.
(3)Commonaltyinthetransporting
f()rmatforvarioustypesofservices.
(4)Programsofdifferenttypes:real-
time,down-loadedtype;one-way,
pseudotwo-wayandtwo-waytype.
(5)Programsandinformationwhich
canbepresentedeitherbyasingle
mediumoracombinationofrepre-
sentationmedia(programcompo-
nents).
(6)Variousdefinitionsorqualitylev-
elsandscalabilityintroducedin
videoandaudiorepresentation.
(7)Va1うousrepresentationpossibilities
bydata:text,still-pictures,graph-
ics,facsimileprint,PCsoftwareand
real-timecomputergraphics,etc.
(8)Synchronizationamongthecom-
binedprogramcomponents,and
thosewithanabsolutetimeifnec-
essa「y・
(9)SupPortingschemestoimprove
humaninterface.
(10)Guidesforusers:menus,previews,
summariesandindexing.
(11)Identificationofservicesandinf()r-
mation:Selectionbycategoriesand
attributes.
(12)Descriptionschemesassociated
withabreakdownofprogramcon-
tentintoparts.
(13)Addressingofindividualsorgroups
ofusersincludingconditionalac-
ceSS.
(14)Interfacewithnetworkandpack-
agese「vlces・
17 JIQNo.101,1995
,'一
,'
TransmissionChanngls
醐
Te「rθstrial \Satθllitg2.5GHz'「s"〈
Possibl6Sgrvices
;;JMobileReceptio,■6"
\へ ・〉
Program
HighQuality
、、
、
も
1
`、、
、 ・ ,'
'
'
Figure31SDBTransmissionChannelsandPossibleServices
(15)Expandabilityforfutureextension
ofservices.
Toconfirmthefeasibi|ityoftheISDB
system,anexperimentalsystemhas
beenbuilt.Theexperimentalsystem
isbasedonthesyntaxofMPEG-2Sys-
temforitssignalmultiplexingfunc-
tions.
7.MultimediaService
Implementationwith
CurrentBroadcasting
Media
Japanesesatellitebroadcastingsys-
tems,includingtheMUSE-HDTVsys-
tem,haveadatabroadcastingcapabil-
itycalledthedatachannel.Thetech-
nicalstandardsfortransmissionand
theprotocolsforbasicmultimediaser-
viceshavebeenestablished.Itisim-
portanttograduallystarttheintro-
ductionofservicesusingcurrentbroad-
castingmediasuchasthesatellitedata
broadcastingchannel,sincethecom.
pleteimplementationoffuturemulti-
mediaserviceswillneednotonlyhigh
technicalcompetencebutalsotheex-
perienceofthoseprovidingmultime-
diaprogramsandtheadaptationof
userswhowillbeutilizingsuchservic-
esintheirdai】ylives.Commercial
servicestodeliverelectronicallythe
videogamesoftwaredirectlytothe
homebyusingthesatellitedatachan-
nelareplannedandthemultimedia
servicesinwhichaudiencescanaccess
tohyperlinkedinformationcontentare
beingdevelopedfortheMUSE-HDTV
system.
JIQNo.101,1995 18
8.Conclusion
ISDBwillbetheinfrastructureforfu-
turebroadcastingservicesinthe`mul-
timediaera'thatwillarrivewiththe
newcentury.Basiccharacteristicssuch
asthemultiplexingofvariouskindsof
programcomponentsignalsandthe
transmissionofmultipleprogramTV
signalshavebeentestedbyusingthe
MPEG-2basedexperimentalsystem.
SinceISDBhasaveryhighpotential,
studieshavetocontinuetoexamine
thefeasibilityofmoresophisticated
functions,whileatthesametimethere
shouldbeagradualintroductionof
basicmultimediaservicesusingcur-
rentbroadcastingchanpelstogainex,
perienceinprOgramproductionaswell
asintheutilizationofthesenewser-
vicesbyusers.
19 JIQNo.101,1995
CollaborationandCommunication
-Avideoconferencesystemwith
facilitiesformultipleeye・contact-
YutakaMATSUSHITA
DepartmentofInstrumentationEngineering
FacultyofScience&Technology
KeioUniversity,Japan
1.HistoricalTurningPoint
ofCommunicationInfra・
structure
Morethan100yearshavepassedsince
thetelephoneservicestarted.Thetele-
phonenetworkisspreadingallover
theworldandhasmaturedasasocial
infrastructure.Wecannotconduct
businessandalsocannotpassoutdai-
lylifewithoutsuchfacilities.Howev-
er,socalledcivilizeddiseasesstealup
unconsciouslytothefacilities.
Itissaidthattheprobabilityonwhich
yωcancatchanaveragebusinessman
throughaphonebyonetrialisroughly
25%.Theotherpeopleofthesame
sectiontowhichthebusinessmanbe-
longshavetousuallyrespondtoyou
threetimesoutoffourtimes.Th輌s
showsthattheyareihterruptedvery
oftenbytelephoneringingsdespitethe
factthattheyhavetoworkhard.
Itcanbesaidwithoutexaggeration
thatatelephonedeterioratesthepro-
ductivityinanofflce.
Itisexpectedthatnewcommunication
facilitiessupplementingthetelephone
willberealized.
Itissaidthatwereceivealmostall
informationunconsciouslyinourdaily
lifethroughimagesorpictures.
Wefaceahistoricalturningpointof
communicationinfrastructure.Anew
infrastructurecalledvisualcommuni-
cationisexpectedinthe21stdentury.
Wetherefore,areexistinginthedays
befbrethetelecommunicationrevolu-
tion.
2.lmportanceofNon・
VerbalInformation
Apsychologistindicatestheimportance
ofnon-verbalinformationwhendeli-
catenuancebetweentwocorrespon-
dentshastobetransmittedasshown
inTable1.
JIQNo.101,1995 20
Table1
Verbal
Voice
Non-verbal
7%
38%
55%
Thecallingpartycansendonly7%of
theinformationtobetransmittedver-
bally.Ifhecanuseatelephone,hecan
send38900ffurtheraddedinformation
sincethepartner(calledparty)can
understandwhetherheisangry,sad,
glad,etc.,ornot.
Furthermore,iftheca}lingpartycan
transmitmotionpictures,hecansend
55%additionalnon-verbalinformation
suchasfacialexpressionsandgestures
sothatthepartnercanreadhisinmost
thoughts.
Fromtheseconsiderations,multime-
dianetworkstransmittinghighquali-
tymotionpicturesareessentionally
requiredfbrtwocollaboratingpeople
beingatdifferentsites.
3.LayerdArchitecturefor
HumanRelationship
Twopeoplearesittingonabenchina
platfbrmoccasionally.Copresence .be-
tweenthetwoisestablished.Oneof
themcallsout"Excuseme,couldyou
tellmethewaytothetemple."The
calledpeopleawaretheexistenceof
thecallingpeople.Thenawareness
(secondlayerofFigUre1)isachieved
betweenthetwo.Dowehavetodistin-
guishcopresence,awarenessfromcom-
municationwhentwopeoplewantto
communicateeachother?
Whentwopeopleareinthesameplace,
itisnotnecessarytodistinguishco-
presence,awarenessfromcommunica-
tionsincetheyareautomaticallyin-
volvedincommunication.
Inordertotransmitdelicatenuance
betweentwocommunicatingPeople,it
isextremelyimportantthateye-con-
tactbetweenthetwoisachieved.When
IwasinvitedtoanNTT,ssystemsshow
afewyearsago,Icametoastopfbra
whileatthecornerofTVtelephone.I
andabeautifulwoman(instructor)
wereparticipatedinasmallexperi-
mentusingthecompactTVtelephone
withabuilt-insmallcamera.Iwas
strainedfortalkingwiththeyoung
beautifulwomanthroughaTVtele-
phone.However,herpictureonthe
screenwasnomorethanabroadcast-
ingpictureonaTVsetbecauseshedid
notgazeatme.Ididnotfeelhereyes
Collaboration
Communication
Awareness
Copresence
verbal,hon-verbal(eye-contact)
gazeawareness,aura,nimbus
FigurelALayeredArchitectureforHumanRelationship
21 JIQNo.lo1,1995
sinceoureyesnevermet.Asthephrase
says,"Eyesaremoreeloquentthanthe
tongue,,'itisessentialthateye-contact
betweentwopeopleisachieved.
Itiscalledgazeawarenesstobeableto
understandwhatthecorrespondentis
gazingon.Wecanfeelakindofauraof
apassionatespeakerandfurtherrec-
ognizetheatmosphere(nimbus)ofthe
meetingsitewhenweareinthesite.
Suchauraandnimbuscanberegarded
asakindofawareness.
Whentwocorrespondentsareinthe
samelocation,gazeawarenessandeye-
contactareautomaticallyachieved,in
addition,theycanfeelauraandnim.
bus.However,thosearenotachieved
automaticallythroughcommunication
circuits.Suchaninteractionofwhich
delicatenuancehastobetransmitted
iscalledcollaboration.Suchaninter-
action,suchasteachingthewaytothe
templeiscalledcommunication.Itis
expectedtodistinguishcommunication
fromcollaborationclearlysothatwe
canunderstandeasilywhatkindsof
functionsareinstalledinaterminal
andnetwork.Collaborationisapro-
cessforcreatingnewvaluesbysupPle-
mentingeachother.Someexamples
areshowninFigure2.
Suchabrainstormingthateachpar-
ticipantprovideonlyhisownideato
theissuesiscommunicationbutnot
collaboration.However,thecasewhere
newideasarec了eatedbysupplement-
ingeachotheriscollaboration.
Inordertosupporthighlevelcollabo-
rationformanymembersbeingindif-
ferentsites,highqualityawareness
mustbeinstalledineitherterminals
ornetworks.
(
co(
冨
三'○
留 §自 廿
日.巨
編 皇:
3
0・upl・R・1・U・n・hi>
← 卿1・ ⌒ 〉()
¢・・ryi・g・De89
COLLABORATION
一
〕
)
(
C・a・hi・gth・W・yω …〕
唾 ・・,㎞ 恥 ⊂ 輌9⊂ …)
)∈eri・di・alMeet・ ・D
l
COMMUNICAT[ON
Figure2Collaboration/Communication
JIQNo.101,1995 22
4.AVideoconferenceSys・
temwithFacilitiesfOr
MultipleEyeContact
Here,wedescribethedesignandim-
plementationofamultipartyvideo-
conferencingsystem"MAJIC"that
projectslife-sizevideoimagesofpar-
ticipantsontoalargecurvedscreenas
ifusersinvariouslocationsareattend-
ingameetingtogetherandsitting
aro皿datable.MAJICalsosupports
multipleeye-contactamongthepartic-
ipantsandawarenessofthedirection
oftheparticipants'gaze.Hence,users
cancarryonadiscussionin・amanner
comparabletoface-to-facemeetings.
Wemadevideo-taperecordingsofabout
twentyvisitorswhousedtheprototype
ofMAJICattheNikkeiCollaboration
FairinTokyo.Ourinitialobservations
basedonthisexperimentarealsore-
portedinthispaper.
4.1DesignofMAJIC
Weidentifiedthefollowingfourdesign
requirementstoimplement"Multi-At-
tendantJointInterfaceforCollabora-
tion"(MAJIC):
(3)Life-sizevideopicturesofthepartic-
ipantsareshownwithoutboundaries
toachieveasenseofreality,and
(4),Asharedworkspacemustbeprovid-
edatthecenteroftheparticipants.
4.21mplementationofMAJIC
Figure3showsaprototypeofMAJIC;
thereisadeskequippedwithawork-
stationandacurvedscreen,f()urfeet
longbyeightfeetwide,forminganarc
witharadiusoffourfeet.Videoimag-
esofotherparticipantsarepr()jected
onthescreenbyvideopr()jectors.When
theusersitsatthe6enterofthearc,
thedistancebetweenhimandother
participantsisaroundf()urfeet.Wehaveconcludedthataroundf()urfeet
maybethebestdistanceforface-to-
facemeetingswith30r4colleagues,
sincethedistancefrompeoplewhowork
togethertendstobeshorteratasocial
distanceoritissometimesataperson-
aldistance.Figure4isapictureofthe
backofthescreeninwhichyoucansee
thedeskthroughthescreen.
HowtosupPorteye-contact
(1)Multi-way(morethantwo-way)
round-tablemeetingsandmultiple
eyecontactbetweenparticipants
mustbesupPorted,
(2)Userscanbecomeawareofthegaze
ofoneparticipanttowardanother
participant,inotherwords,users
maintainawarenessofwhoisvisu-
allyattendingwhom,
TheMAJICscreenisathintranspar-
entfilmwithalargenumberofvery
smaUhexagonsprintedonbothfaces.
Whitehexagonsareprintedontheback
side(faceA)andblackhexagonsare
printedonthebackside(faceB)as
showninFigure5.Sincethewhite
hexagonsreflectlightandisdarkbe-
hindthescreen,faceAcanbeusedasa
normalscreen(seeFigure3).Theblack
23 JIQNo.101,1995
Figure3APrototypeofMAJIC(frontview)
Figure4APrototypeofMAJIC(backview)
JIQNo.101,1995 24
refiection
passmg
transparentfilm
passmg
abSorption .
FaccAFaceB
Figure5MAJICScreen
電 竺i蜘\ 一 一
‥ 、こ・
\
sαeen
dirθctional
mbrophon
一 一ρ一 一/
/一,','
○\
\ ,'
,'バ,'\
↑ 一'一 、
vidgocamera
/
/、、
ク\ 、lW、、p輌 図 ・
…>elhgmgt
/、 、work
v… 司 ■ ・ ・
Figure6SystemArchitectureofMAJIC
25 JIQNo.101,1995
hexagons,ontheotherhand,donot
reflectlightatallanditisrelatively
lightinfrontofthescreen,andconse-
quentlyonecanseetheotherside
throughthescreenfromfaceB(see
Figure4).Thetransmissibilityofthe
screendependsonthesizeandthe
numberofthehexagons.Wehaveused
a40%transm{ssibilityscreenf()rthe
MAJICprototype.
Figure6illustrateshowMAJICsup-
portseyecontact,AvideoprojectorlocatedaboveuserAprojectsalife-size
videoimageofuserB'onthescreen
andavideocameralocatedbehindthe
screentakesapictureofuserAand
viceversa.Inordertotakeclearpic-
turesofusers,thevideoprojectorsare
setupsothattheydonotenterthe
fieldofvisionofthevideocameras.
Sinceeachvideocameraissetupatthe
centerofthepartner'sfaceonthevideo
image,userAandBcanmakeeyecon-
tact.Moreover,sincevideoimagesof
participantsarelife-size,userscanread
theirpartner'sfacialexpressionsand
observethedirectionoftheirpartner's
gaze・
JIQNo.101,1995 26
Multi・Media・PC'Communications
HiroshiFUJIWARA
GraphicsComm皿icatiqnLaboratories
1.Summary
PersonalComputer(PC)environment
ischangingrapidlyinthe「1990sto-
wardsthe21stcentury.Multi-Media
technologies,whicharethe'greatest
inventionsinthelatterhalfofthe20th
century,haveprovidedPCenvironment
wit})thebiggestinfluenceinanyother
enVlrOnmentS.
Thispaperanalyzestheessenceof
Multi-Mediafromthetechnological
pointofview,andpredictsthefuture
ofPCenvironmentthroughourrelat-
edtrials.
2.InformationInfrastruc・
ture
Thereforeinthispaper,"lnformation
Infrastructure"consistsofcomputers,
communication,broadcastingandstor-
ages(seeFigure1).
3.WhatisMulti・Media?
DefinitionofMulti-Mediaisstillun-
certain.Therearevariouskindsof
definitionsaccordingeachground.In
thispaper,de6ningMulti-Mediahas
beentriedforfbcusofdiscussion.For
example,Mediaisagenericnameof
representationmedia(text,voice,vid-
eo,graphics,etc.)andtransmission
media(cable,satellite,ATM,etc.).
Multi-Mediaisdefinedasintegration
ofseveralrepresentationmediabyone
transmissionmedium.
"lnformationInfrastructure,"which
willdefinitelygrowfromanationalto
aglobalscale,andcouldplayakey
roletowardsthe21stcenturyasthe``lndustrialInfrastructure"hasp}ayed
akeyroleinthe20thcentury.
Anetworkofhighwaysoranetworkof
railwaysareincludedin"lndustrial
Infrastructure."Ontheotherhand,"lnformationInfr'astructure"should
includeanykindofinformationmedia
ratherthananetworkingmedia.
Basedonthisviewpoint,specialfea-
turesofMulti-Mediaareasfollows.
(1)Allinformationaredigitized.
(2)Humanbeingscanusemulti-media
informationinteractivelybythehelp
ofcomputers.
Currentinformationmediaexistde-
・pendingonthefrequencyandthe
numberofusersasshowninFigure
2.However,Multi-Mediatechnolo-
27 JIQNo.101,1995
lnfrastructure
lndustrial
lnfra.lnformation
lnfra.
■lnformationlnfra.
(Medialnfra.)
Computer
Communication
Broadcasting
Storage
FigurelInformationInfrastructure
9ieSmightChangetheexiStinginfOr-mationmediaintoMulti-Media.
4.WhatareMulti・Media
Technologies?
InvariouskindsofMulti-Mediatech-
nologies,themostimportanttechnolo-
giesarethoughttoincludecomputertechnologies,compressiontechnologies,
VLSItechno}ogies,andtransmission
technologies.
softwaretechnologiesand3Dgraphics
aremoreimportantthananyother
ones.Incompressiontechnologies,
whethertheymightbevideo,audioor
still-images,theyshouldbeoptimized
totransmissionmediaandrequired
quality.VLSIarchitectureandimple-mentationarealsoimportantespecial-
lyforMulti-Mediacoding.Finally
transmissiontechnologiesareusedfor
personaldigitalcommunicationbe-tweenusersandinformationprovid-
ers.
Incomputertechnologies,interactive
JIQNo.101,1995 28
コ889×
Year
Week
Month
Day
Real・
Time
Commerdal
Letters
lnternet
Entertainment
Monthly
agaZln
Gameof
VideoAudio
E'Mail
*
Week呼
Magazin
Broadcast
E・Mail
*
Newspaper
・・a・c・・1⑧ …e・al
TECHNOしOG
・Computer
・VLSl
・CompressiOA
MULTI・M∈DIA!
Telecom
*卓 **
T・1・ph。nec儒=,
Broadcasting
*:Digital
**:Partiallydigital
Personal
Figure2
GroupMassNumberofusers
PotentialInformationMediaforMulti-Media
5.WhatisPCinMulti・Me・
dia?
PC'shavechangedoverthepast20
yearssincethe1970salongwiththe
progressofmicroprocessortechnologiesandsoon.Basedonthesediscussions,
IwouldliketopredictPCinaperiodof
Multi-Media.
Firstly,PC'scouldhaveadditionalrep-
resentationfiユnctionsratherthandis-
playoftextandgraphics.Secondly,
PC'scouldhavetheadvancednetwork-
ingfunctionsinchldingtelecommuni-
cationandbroadcasting.
However,improvementofthenetwork
environmentshouldbepromotedin
advance,orinotherwords,Multi-Me-
diaPC'smightbethedrivingfbrceto
promotenetworkinfrastructure.
Inthispaper,theconceptandopinion
f()rPC'sintheMulti-Mediaperiodis
describedclearlyandconcretely.
29 JIQNo.101,1995.
SuperHighDefinitionImage
ApplicationforMedicalUse
SadayasuONO
・N'1"rOpticalN.etworkSystemsLaboratories
1.Overview
Assuringthatpeoplelslivesarecom-
fortableandworry-freerequiresthe
establishmentofasoc{almedicalsys-
teminwhichawiderangeofadvanced
medicaltechnologiesisaVailableand
thatcanbeeffectivelyimplemented
withoutmajoreconomicburdens.The
diverserangeofmedicalequipment
developedovertheyears,whilemainly
helpingtoraisethelevelofmedical
technology,hasnotbeeneffectively
implemented.
Ofcourseitispossibletoattainthe
requiredlevelofimagequalityformed-
icaldiagnosticsusingfilmandpaper,
butuseofsuchmediahasamajor
drawback.Itisdiffricult,usingsuch
media,fordoctorstoobtaindiagnostic
informationwheneverandwherever
theywant...
Therootofthisproblemuntilnowhas
notbeendealtwitheffectively.The
medicalprofessioninvariouscountries
hasbeendoingwhatitcantoprovide
diagnosticinformation,butthisre-
quiresthatimageson61mandpaper
bephysicallytransported,makingit
diffricultfordoρtgrstogettheneces-
saryinformationattherequiredloca-
tionandintherequiredtimeframe.
Becausepeople,slivessometimerest
onmedicaldiagnosticinformation,the
medicalfieldhasbeendoingwhatever
itcantoobtain,totheextentcurrent
teChnOlOgyallOwS,neCeSSaryinfOrma-
tion.Thisoften,however,leadstothe
samemedicalexamsandprescriptions
beinggivenoverandoverandisoneof
thefactorsthatpreventsthemedical
systemfromoperatingeffectively.More
thaneconomicconsiderations,howev-
er,istheconcernthatsuchrepeated
prescriptionsdamagethepreciseness
ofmedicaldiagnosesandinworsecase
scenariosleadtothedeathofpatients.
SHDimagingandrecenthigh-speed
LANandB-ISDNhaveenabledthecre-
ationofpaperless/filmlesshospitals,
makingitpossibletoobtainmedical
diagnosticinformationanytime,any-
wherethroughthesharedusbofelec-
tronicclinicalchartsonanation-wide
basis、Itisbecomingclearthatthis
methodofmedicalinformationex.
changewillbothadvancemedicaltech-
nologyandhelpthemedicalsystem
JIQNo.101,1995 30
operatemoreeconomically.
2.MedicalImageQuality
Figurelshowsthewidevarietyofim-
agingtechniquesusedinthemedical
field.Itisapparentfromthisfigure
thatformonochromeimages,X-rays
demandthehighestlevelofquality
whileforcolorimages,thesameistrue
forpathologicalexamimages.Note
thatFigure'1showsonlythemostfre-
quentlyusedrepresentativeformsofmedicalimagingtechniques.Theau-
thorsdonotmeantosuggestthatthere
arenotanyotherapPlicableimaging
technologies,justthattheyarefewand
farbetweenandtherearemanymeth-
odsfordealingwithsuchimages.
Iftherewasane}ectronicimagingtech-
nologythatcouldadequatelyhandleX-
raysandpathologicalexamimages,we
couldsatisf>oneoftheconditionsfor
establishingthefilmless,paperlesshos.
pitalswedisωssedearlier.Theau-
thorsofthispaperdetermined,through
thecooperationofanumberofclini-
ciansandpathologists,thatSHDim-
ageshaveahighenoughlevelofimage
qualitytobeusedforX-raysandpatho.
logicalexamimages.SincetheSHD
imagingtechnologyusedfbrmedical
purposescorrespondstoCDs(Compact
Discs)usedforrecordingaudio,alarge
amountofdatacanbemaintained,al-
lowingthereproductionofhighquality
1mages・
3.HighSpeedCommunica.
tionNetworks
ThedatavohlmeofasinglestillSHD
imageistypically12Mbwhilemoving
SHDimagesrequireabandwidthof
6Gbps.Inthepastithasbeennextto
impossibletorapidlyprocess,trans-
mit,andstoresuchlargevolumesof
digitaldata.RecentadVancesinopti-
calf]berandULSItechnology,howev-
er,ismakingitpossible,bothtechno-
logicallyandeconomically,torapidly
transmitandstoreIargeamountsof
data.
Concreteexamplesincludeopticalfi-
bertechnologythathasallowedthe
developmentofFDDI,ATM-LAN,high-
speedEthernet,andB-ISDN,whichall
utilize100to156Mbpstransmission
lines,andULSItechnologythathas
supportedthedevelopmentofhigh
speedprocessorsandlargevolume!high
speedmemory.Inaddition,recentad-
vancesintheamountofdatathatmag-
neticstoragedevices,suchasmagnetic
disksandmagnet-opticaldisks(MOs)
canstorehasmadesystemsthatrapid-
lyprocesslargeamountsofdatamore
economical.
Thespreadofhighspeednetworkswill
continuealongwiththeincreaseduse
ofimagedata。Theauthorsbelieve
thatmedicalimagingwillbethemoti-
vatingforcebehindthisspread.
4.CurrentIssues
Thetwomainareasunderdevelopment
forimplementingfilmless!paperless
hospitalsarelistedbelow.
1)LANinterfacesforX-rayequipment,
31 JIQNo.lo1,1995
(1)X・rayImage
PixelNumber
▲ 硯 、・,ii.20482
皇__Io241亘
.,512・
コ穿
ξ 鴫蒜 α。。,.,i夕 占ゴ::≒
㌫ ばi=:t'16:)i:=:,:"、8bit256
,`,e・illl、、t16M
Gradatien/DynamicRangcColorNumbcr
(3)cr/MRI
PixelNumber
裟:1:;
麟 認_ミiミ:)t-'16
8bit256
16bitl6M
Gradatlon〆Dyna加cRangeColorNumber
(5)Cintigram
PixelNumber
鰺:.=:s
詩lo24'
一 ・type翻:l
onoChrome
16bit16M
Gradaticu/DynamicRangeColorNumber
JIQNo.101,1995
(2)PathologicalExamImage
PixelNumber
羅:
繁
≠ぺ㌻Gradation/DyTiarnicRangeColorNumbcr
(4)Ultra-SonicEchoImage
PixctNumber
40961
20482
10242
5122
256i
ono(≡hrome
頴,、16bit16M
Gradation/DynamicRangeColorNumber
Colorir:9type
(6)Endoscope
PixelNumber
叢:1雲
凌籔ごぽ 恩 蝿iii§・・、、。
16螢へGradatien1DynarnicRangeColo了Number
・加 ≠
FigurelCharacteristicsofMedicalImages
(performanceofopticalsystemisconsidered)
32
endoscopes,MRI,andCTscanners,
etc.
2)Utilityprogramsfbrcreatingclini-
calcharts.
wouldberelativelyeasytoraisethe
imagequalitytotheHDTVlevel,which
infactisthemaximumqualityimage
anendoscopecanproduceduetoits
opticallimitations.
Bothoftheabovearevita】fbrcreating
anenVironmentwheremedicaldatacan
beprocessed,transmitted,andstored
electronically.Numberoneofcon-
cernedwiththedatageneratedbyim-
agingdevices,whilenumbertwoiscon-
cernedwiththedatagatheredbydoc-
torsandexaminationspecialists.
Usingexistinghardwareandsoftware
suchasscanners,wordprocessors,and
graphiceditors,itisofcoursepossibletodigitizetheinfbrmationrecordedon
filmandpaper(clinicalchartsandex-
aminationresultsheets),butthiscan
leadtovariousproblemsintermsof
cost,timeprecision,andeveninput
mistakes.Itisthereforenecessaryfor
informationtobeimmediatelydigitized
atthesource.
Therearefewdif刊cultiesindeveloping
interfacesforthedeviceslistedinnum-
beroneabove.Inparticular,MRIand
CTscannerimagesareobtainedusing
computers,andthesystemsforcon-
trollingthesecomputershavealready
beenputintouse.DevelopingLAN
interfacesisthereforerelativelyeasy.
X-rayequipmentusingtheIP(imaging
plate)method,whichdirectlydigitizes
datawithoutUsingfilm,alreadyex-
ists.Suchimagesrivalthe.qualityof
film-basedX-rays.Anendoscopethat
canproduceNTSCclassimageshas
alreadybeencommercialized,andit
Theutilityprogramsmentionedin
number2abovefbrcreatingclinical
chartsisanareainwhichagreatdeal
ofresearchisrequired.Thereare,of
course,databaseapplicationsalready
availablesuchasFileMakerProand
4thDimensionthatcanhandleimage
data.Themainsubjectthatneedsto
beresearchediswhatsortofuserin-
terfacedoctorsfindeasiesttouse.The
numberonegoalshouldbemakingit
easiertoenterdataintoclinicalcharts.
5.PresentStateofMedical
InformationSystems
Establishinganelectronicmedicalin-
fOrmatiOnSyStemOnanatiOnalSCale
willgreatlybenef]tthe`population.
Thesebenefitsinclude,asstatedinthe
openingparagraph,thespreadofawide
rangeofadvancedmedicaltechnolo-
giesandtheeconomicallyef猛cientop-
erationofthemedicalsystem,which
willmakepossiblesuchthingsasthe
shareduseofclinicalchartsandre-
motediagnoses.Ifthereisnostrong
motivatingforce,however,topromote
thetransitiontoanation-wideelec-
tronicmedicaldatasystem,thistopic
mightjustendupbeinganidealistic
proposal.
Inreality,thecomputerizationofmed-
icalinformationsystemsstillfaces
manyunresolvedtechnologicalbarri-
33 JIQNo.101,1995
ersanddevelopm'entinthisareaisata
nearstandstill.Theuseofsuchequip-
mentasMRIandCTscannersthat
directlyhelpincreasetheprecisionof
directdiagnosesisincreasingatamod-
estpace,whilethatofequipmentwith
amoreindirectrolesuchasPACS(pic-
turearchivingcommunicationsystem)
isnot.Asaresult,therearealmostno
hospitalsinJapantodaythatareusing
100MbpsclassLANsintheirclinical
departments.Thisismainlybecause
theydonotseeanyapparentbenefits
tobeobtainedthroughintroducingsuch
networks.
Thissituation,however,isinthepro-
cessofbeingdramaticaUyaltered
throughrecentadvancesinmedical
technologiesandtheincreasingversa,
tilityandwide-scalespreadofmedical
equipment.Theendresultwillmean
aninCreaSeintheamOUntOfCliniCal
chartdataperpatient,asweUasthe
costofmaintainingthatdata.Theprop-
ermaintenanceofchartsisvitalfor
accuratediagnosesandtreatments.It
isthereforecriticaltofindsomewayof
properlymaintainingPat{ents'charts;
computerizedmedicalinformationsys-
temsareextremelyeffectiveinthisre-
spect.
beds,forexample,haveastaffoflittle
lessthanthirtysolelyforthepurpose
ofmaintainingcharts.Throughcom-
puterizedchartmaintenancethisfig-
urecanbereducedtoone-halftoone-
fifthofthecurrentlevel.Theresulting
costreductionisnotsolelyrelatedto
cuttingbacklaborcosts.Italsoisa
factorofreducingchartstoragespace,
whichisespeciallyexpensiveatlarge
hospitalsinmajorcities.Obviously,
thelargerthehospitalis,thegreater
thelevelofsavingswillbe.
Directcostreductionsarenearlynegli-
giblefbrfamilydoctorsandsmallhos-
pitals.Thebenefitsforsmallhospitals
istheincreasedshareduseofclinical
charts,whichwillresultinagreater
rolef()rfamilydoctors.Thisrolewill
includemedicalexamsthatdonotuse
large-sizedequipment,healthconsul-
tations,andmakingtheroundstopa-
tientshomes-allwellsuitedtothe
grayingPopulationinJapan.
Thebenefitsofcomputerizationwillin
thefuturebemostapparentinthearea
ofdiagnosesandtreatments.Inpar-
ticu}ar,thiswillleadtocostreductions
ingeneraltaskssuchasprocessingin-
surancebills.
6.BenefitsofComputeriza.
tion
Thereductionthroughcomputerization
ofchartmaintenancecostswillbemost
apParentinrelativelylargehospitals
(between200to5000rmorebeds)that
havefull-timestaffforma{ntaining
clinicalcharts.Hospltalswith1000
Progressinthefieldofmedicalscience
isbasedonexperience,andstatistical
analysisofmedicaldataisvitalfor
gainingthisexperience.Medicalcharts
containtherawdatatobestatistically
ana】yzed.Computerizingthisdatawill
playanindispensable.roleinensuring
thatstatisticalanalysisofmedicaldata
isbothswiftandaccurate.Itwillthus
」【QNo.lo1,1995 34
bepossibletoanalyzetheside-effects
ofmedications,quicklyrespondtothe
treatmentofcommunicablediseases,
andobtainquickofficialapprovalfor
effeCtiVeneWtreatmentS.
7.Conclusion
Theauthorsofthispaperarepropos.
ingtheimplementationof佃mless/pa-
perlesshospitalsandbuildingofa
"MedicalInternet."ThepresentInter-
netmainlytransfersrelativelysmall
textfilesandisnotcapableofhandling
thelargemedicaldatafilescontaining
highqualitySHDimages.Problems
laynotonlywiththeissueoffilesize,
butinadditionotherconditions,differ-
ingfromthoseontheInternet,mustbe
satisfied,eveninsuchareasasreli-
abilityandtheprotectionofconfidenti-
ality.Thisobviouslyincludesthestan-
dardizationoffileformatsformedical
diagnosticdata(primarilyclinical
charts),andtheestablishmentofafile
tnanagementSyStem.
Theseconditionsnecessitatebuilding
adedicatedmediealnetwork,complete-
lydifferentfromthepresentInternet.
UsingB-ISDNwillbethefastestway
toeconomicallybuildanetworkofthe
requiredscale.Theinitialphaseof
buildingthisnetwork,however,wm
requireratherstrongpromotionand
leadershipformallcorners.
Thefundsrequiredforpromotingthis
policyalsomustcomefromfinancial
supportbyallareasofthepublicsec-
tor.Suffricientpublicapprova1,howev-
er,forthisfinancialsupportmustbe
obtainedbefbreadedicatedmedical
networkofaneffectivescalecanbe
built.
35 JIQNo.101,1995
PersonalInformationTerminals
MasafumiMINAMI
DevelopmentDepartment,Business&IndustrialSystemsCompany
SonyCorporation
1.Introduction
Recently,severaltypesofPersonalIn-
formationTerminalshavebeenintro-
ducedintothemarket.Theapplica-
tiontrendforthesedevicesisshifting
fromjustdatamanagementtools(e.9.
scheduling,addresscards,etc.)toward
richercommunicationcapabilities.In
thistalk,theauthorwillpresentMag-
icCapldeviceandthesupportingtech-
nologies,whicharedevelopedbyGen-
eralMagicInc.,asanexampleofthe
latterapproach.Inthefollowingsec-
tions,thekeytechnologiesarefirstpre-
sented,thensomeapPlicationexam-
plesofthesetechnologiesarediscussed.
2.KeyTechnologies
Thebasicconceptofthistechnologyis
thatinordertorealizenewcommuni-
cationfeatureswhichdonotcurrently
exist,itisnotsufficienttoonlydefine
user-friendlyPersonalInformation
Terminals.Supportfornetworkinfra-
structuresisalsoneeded.Thekey
technologies,therefore,consistofthe
followingS:
i.TelescriptasasupPortfornetwork
infrastructures
ii.MagicCapasanOSfbrPersonal
InformationTerminals
iii.MagicmailasanTelescript-enabled
messagingapPlication
2.1Telescript
Telescriptisadistributedobject-ori-
entedinterpretiveprogramminglan-
guagefordevelopingcommunicatingapPlicationsoverthenetwork.
Themainconceptofthelanguageis"Agent"and``Place"(asinFigure1)
.
Thesearebothrunningprograms,i.e.
process.A``Place"isastationarypro-
cesswhereasan"Agent"ismobilein
theTelescriptnetwork.
ATelescriptnetworkisanetworkcon-
sistingofmachineswithTelescriptlan-
guageinterpreters(TelescriptEngines
orTSE'sinshort).Agents,therefore,
cangotoanyPlaceonthenetworkand
dothespecifiedtasksthereonthebe-
iMagicCapandTelescriptaretrademarksofGeneralMagic ,Inc.
JIQNo.101,1995 36
User
Place
σぷθr
A8ent
、
TelescripEngine
80
erVtcePlace
TelescrlptEnglne
、
σぷθr
8ent
ServicA8ent
TelescriptEngine
FigurelAgentsandPlacesinTelescript
haveoftheendusers.
Fromtheprogrammers,pointofview,
therearethefollowingadvantagesover
theexistingtechnologies.
i."Go"methodinvocationtomove
anotherPlace
TovisitseveralPlaces,thepro-
grammerneedonlyinvoke"Go"
methodswithdestmatlonID'sin
theAgentprogram
ii.IndependencefromNetworkPro-
tocolS
Allconceptsinthelanguageare
protocol-independent.Thepro-
grammerscanfocusonthebehav.
iorofapPlications,i.e.whereto
visitnextandwhattodothere.
Asconsequence,theprogrammer
candevelopAgentprogramswith
moreease.
2.2MagicCapdevice
MagicCapisaTelescripレenabled
(thereforeTSEinside)OperatingSys-
temwiththeuserinterfaceasinFig-
ure2.
Theuserscansend
iMagicmailmessages
1i.OtherAgentsdevelopedby3rd
softwaredeveloperstonetworks
toaccomplishhis/herobjectives.
Fromtheuser'spointofview,Mag-
icmail"Telecard"issimulatingthe
physicalpostcards.Ithasaddi-
tionalfunctionality,utilizingTele-
scriptcapabilities.Forexample,
asshowninFigure3,theTelecard
37 JIQNo.101,1995
(fbrarrangingameetingwithre-
cipients)includesabuttonwhich,
inturn,includesaspecificagent.
Theagentisprogrammedas:
IfYesbuttonispressed,
Thenfillinthemeetingdate&time
inthelocalschedulerandgotothe
nextdestination,
Elsejustgobackandinformthe
senderthatdateisnotgood.
◎D・ ・kT…d・y,(kt。ber18-4仁 【管H・11w・y
.12.
くゲ 互 』,_'6魯I
n=10u1-一一
/薫 孝
違 憲㌧ ・
、
ロ ー 一己 園
Rl山 嘗 白 Wl圏k輌
Figure2MagicCapStart・upGUI
?cardtoKojiWed ,.Mayll 【蚕戸Generalappoint川ent
DearKoji.
lwouldliketoseeyouon
Wednesday,Mayl1.199AatlO
am..Canyoumakeit?Press「yes'or'no'tOletrneknOW
.
Sincerely、
MasafurniMinami
…IL・
#醤
t●:KojiInai
fr●■:MasafumiMinami
"dilmeetingWed.5!11
図tapfordetails
口send
巨コ.address
図discard
脚●ras◆
口ex7tetld
R ◇ め .乙 繭 ◎
Figure3TelecardforMeetingArrangement
,JIQNo.101,1995 38
蝿轟
・・騨一 藁謬熟懇謬く 古
参s
Figure4TheConceptualizationofElectronicMarketplace
inTelescript
3.ApplicationExamPlesinElectronicMarketplace
BesidesMagicmail,anotherTelescript-
enabledapplicationsistherealization
ofthe"ElectronicMarketplace."Using
Telescript,itisnotonlyverystraight
f()rwardtoimplementitoveranetwork
(asinFigure4),butitisalsopossible
toprovidenewfeatureswhichdonot
existinthecurrentnetworkservices.
Wewillpresentthethreeexamples.
3.1Watchingtheevents
Itisoftenthecasethatusersdon't
wantto"busy-wait"forexpectedevents.
Forexample,waitingfordesiredstock
values,lookingforusedcars,knowing
thechangetoflightschedulesandso
forth.Intoday'sinfrastructure,a|lwe
havetodoistocontactperiodicallyand
tocheckiftheeventsoccur.Byutiliz-
ingAgents,usercanspecifythede-
siredeventsandsendittothenet-
work.Theagentwaitsfbrtheeventto
occurandifitoccurs,itinformsthe
user(asinFigure5).
3.2Navigatingthroughthe
hugeamountofinforma・
tioninthenetwork
Therearehugeamountsofinfbrma.
tioninnetworks_toomuchinforma-
tiontosearchfordesiredthings.Users
havetoseekforthedesiredinforma-
tionmanually,e.9.net-surfrmgonWWW
servers.ByutilizingTelescript,users
canspecifytoanAgentthedesired
itemsandtheagentcanseekforthem
inplaceoftheuser(asinFigure6).
39 JIQNo.101,1995
OnPersonalCommunlcator
'♂ぽ
OnTeiescnptNetWork
FiguresWatchingEvents
OnPersonalCommunlcator
OnTelescnptNetwork
Figure6NavigatingthroughtheHugeAmountofInformation
3.3Coordinatingmultiple
eVentSatOnetime
Itisalsooftenthecasethatendusers
wouldliketomakemultipleappoint,
mentsforoneday.Forexample,to
makereservationsforaconcertanda
restaurantseatingforValentine'sday.
Inthiscase,eachappointmenthassome
mutualconstraints,i.e.avoidingtime
conflicts.UnderTelescript,thiscon-
straintisexpressedbyvariablebind-
ingsintheAgentprogram.Allthe
userhastodoisjusttoparametarize
theagentandsendittothemarket-
place(asinFigure7).
JIQNo.lo1,1995 40
OnTelescrlptNetwork
Figure7CoordinatingMultipleEvents
4.Summary
WepresentedTelescriptandMaglc
Captechnology,asanexampleofPer-
sonalInformationTerminalsandsup-
portinginfrastructuretechnologyand
showedthe{rpossibleapphcationex-
amples.
Allapplicationexamplesmightbere-
alizedbythecombinationofotherex-
istingtechnologiesTheimportant
pointis,however,thatwithoutthe
schemepresentedinthistalk,creat-
ingsuchapPlicationswlllbeavery
time-consumingtask,i.e.wehaveto
definetheinterfaceand/orprotocols
possiblyforeveryapplicat】onThe
Telescript/MagicCapsolutionisone
ofthefirsttosolvetheaboveproblem
forprovidingtheendusersthenew
convenienceofcommunications.
5.Acknowledgment
WethankJimWhiteandGeorgeFan
atGeneralMagicInc.forprovidingus
thetechnicalsuggestionsandthesome
ofthepresentationmaterials.
6.References
L
2.
White,J.,TelescriptTechnology:The
FoundationfortheElectronicMar-
ketplace,GeneralMagicInc.1994
White,J.,TelescriptTechnology:
ScenesfromtheElectronicMarket-
place,GeneralMagicInc.1994
41 JIQNo.101,1995
PortableMultimediaMachinefor
TheBusinessPerson
KensukeADACHI
MultimediaEngineeringLaboratory
ToshibaCorporation
1.Introduction
ThePChasbeenwidelypopularinthe
o笛ceandnowthereisastrongre-
quirementthatitbeusedanywhere
outoftheof猜ce. ,LSIandintegration
technologymakesthePChigherinper-
formanceandsmallerinsize.Popular
PCtypesareshiftingfromdesk-topsto
lap-topsandportables;theNote-type
PCandSub-Note-typePCshaveap-
peared.MMPC(Multi-mediaPC)inte-
gratedwithCD-ROMislikelytospread
widely.Thenumberofsoftwaretitles
storedinaCD-ROMisincreasingf()1-
lowingtheincreaseofMMPCs.The
MMPCisexpectedtobeverypopular
attheof6ceandhomeinthenearfu-
ture.GenerallytheMMPCdealswith
videoinformationcompressedusually
byMPEG.Judgingbythesetrends,a
portableMMPCisdesiredstronglyand
willdowellinthemarket.ThefoUow-
ingrequirementsfortheMMPCsare
considered.
1)Portability;
MMPCshouldbesmallandlight
enoughtobetakenouteasily.
2)Capabilityofmulti-mediaprocess-
ing;
MMPCshouldhavethefollowing
functions.Character,voice,image
andvideoprocessing.
3)Compatibility;
Alotofapplicationsoftwareshould
beavailablefortheMMPC
2.PenComputer
Torealizeportability,apenbasedcom-
puterismoresuitablethanakeyboardbasedcomputer.Butthereareseveral
problemsinproducingapenbasedcom-
puter.Thefirstproblemisthehand-
writingrecognitionneededtoenter
characters,isnotyetmatureenough
andhassomelimitationsinaccuracy;
usuallytowriteacharacterinaboxon
anelectronicfbrmwhichpopsupon
thescreen,andtokeepthenumberof
strokesconsistingofacharacter.Es-
peciallyJapaneselanguage`Kanji'con-
sistsofsomanystrokesthatwehave
troublewritingacharacterofmany
strokes.Therefore,Kanamixedby
JIQNo.101,1995 42
KanjitoKanjiconversionisnecessary
toenteranyKanjieasily.Wehave
developedthiselementarytechnology,
andLaveinstaUedtheminthefollow-
ingPenComputers.
1)Dyxlanote
DynanoteisaPenComputerwhichis
optionallyconnectedtoakeyboard.It
isdesignedasanordinaryPCwitha
keyboardintheofficebuttobeaPen
basedPCwithoutakeyboardoutof
theof6ce.Touchingamemlonthe
screenwithapen,salespeopleinthe
fieldcanenterseveralcustomerorders
intothePenPC.Beforesalespeople
returntotheiroff]ce,theycansendthe
entereddatatoahostcomputer
throughthetelephonelines,andwhen
theyreturn,theycandealwiththe
datausingakeyboardandsendthe
datatothehostcomputerviaaLAN.
Dynanoteisexpectedtobeusedeffec-
tivelyinthesalesfield.Thesuccessor
totheDynanote,theT200,thefirst
colorpenPC,isdevelopedandshipped.
Itsshapeandmainspecsareshownin
FigurelandTable1.
2)XTEND
XTENDisanorganizerbutithasa
largedisplay(640dotY400dot)and
3.5tFDD.Itisdesignedtobeaporta-
bleinformationtoolwiththeconcept
tobringtheinformation,references,
makenotesandtoutilizeit.Different
fromtheothermaker'sorganizers,in
additiontoPIM(personalinfbrmation
management),ithasaWP(WordPro-
cessor)softwarecompatiblewithToshi-
ba'sRuposoftwareandreads/writes
thedocumentfileofRupothroughFD.
Moreover,ithastheabilitytorefer-
enceadocumentfiledinMS-DOSfor-
mat.Itssizeis225(W)¥175(D)¥
30(H)mmandtheweightisabout900g.
Itsshapeandoneof・itsscreensare
showninFigure2andFigure3.
3.SystemExamplesofPen
Computer
PenPC(Dynanote)isbeingusedinthe
businessandmanufacturingfieldsbe-
causeofitsportabilityandeasyma-
nipulation.Twoexamplesarede-
scribedinthefollowing.
1)Tomakedailyandmonthly
businessreports
Salespeoplewhoaretradingcommuni-
cationequipmentwillinquireabout
informationoftheircustomerfromthe
databaseofthehostcomputerand
downloaditontoaPenPC.Before
visitingthecustomer,theyhavethe
knowledgeofthestatusofexisting
equipment.Theycantaketheinfor-
mationwiththem,tradewiththecus-
tomerreferringtotheinformation,and
entertheorderintotheirPenPC.Af-
terthemeeting,detailedrequirements
ofthecustomeraremadeontheirPen
PC.Whenthesalespeoplecomeback
totheiro笛ce,theysendthedatafrom
theirPenPCtothedesk-topPCcon-
nectedtothehostcomputer.Thehost
computerupgradesitsdatabaseusing
thedataandmakesdailyandmonthly
reports.Figure4showsthesystem
configuration.
43 JIQNo.101,1995
「,、/へ
Figure1
Table1
ModelItem T200CS(color)T200(Black&White)
Software WindowsfbrPenComputing
Processor i486D×2-40MHz32bit
8KBcacheinCPU
Memory Standard:4MB,Max:20MB
Display 9.5"640×480dotColorSTNLCD(T200CS)
9.5"640×480dotMonochromeSTNLCD(T200)
SimultaneousDisplayonExternalMonitor
Input CordlessStylusPen(ElectromagneticConductiveDigitizer)Keyboard(Option)
DiskDriVes 1.8"80MBHDD3.5"ExternalFDD(Option)
Interfごces User-installableMemorycard,RS-232C,Printer,
FDDKeyboard,PCMCIA(Type2×1Type3×1)
Power Lithium-ionBattery(3000mAh),ACAdapter
Dimensions 27.OW×21.OD×4.1cmH
1.95Kg(T200),2.Okg(T200CS)
JIQNo.101,1995 44
Figure2
タイトル:[地 図1]
評価サン■
%5・ 任
榊駆」 驚z轟]1艦
Zl吻二1[]
1『
L;
弦(こ il
一1一
「ヨ⌒ !
!'4媛 澱へ
惨=郵 遍`
1|
1口細6
冨 下疏 〒「
メモ1■[=コ
石間 ツール
テキス ト 手 書き サウン ド
メ モ 日?モ 眺0 0の
Figure3
45 JIQNo.101,1995
,
DaIabaseCenter
= \r、
一ノ
ノ
.
、
一 ■輪一ミ
、
一O【herSv51em
'
一
D訓aba駝Server
r
\
\・、bli、D・、il。,、、,、N,㌔。、∨) _
」.
「
、\
_Braocb
Dahba父
C凶om臼'P⑩du"ioolo∫or1かUoo
DJIy胴 龍山lyR¢1蛤n
1 Roo任r
/
圃」
RS232C
D醐 。,pcl
PenPC
%Mク
F
'
イ 一U∩Load 刊 門
1>一慢 フ
・,、。c・ どwか ㎞1
..
Figure4
Hcudoffice
Upd"te-datai
-:
ltasedLine
-
OrderD・ 佃
BusinessOt'ficcOP
Ulxbltddala
-
RS・232C
-'lhtDatnSent
toSever
Business
Offi㏄Strver
BranchlBusinessOfficeBusinessFom
TheFi!tforEuehSalcsnian-
RS・232C
-
TheD;taScnttoSevcr
PenPC(Busin△s
SupportTerm}nu|)
OrdCTDatn
BusinessServi㏄(>r
MCAOperationStation
MCATnasmission-
OrderDlt8
PenPC(BusinessSupportTerminal
Figure5
JIQNo.101,1995 46
2)Toutilizeportabledatabase
storedinPenPCs
Salespeopletradingmedicinetoadoc-
torneedcorrectinformationabouta
lotofmedicine.Therefore,theyhadto
takeathickpamph}etinwhichmedi-
cal.infbrmationofmorethan15,000
medicinesaredescribed,oraskabout
theinfbrmationtotheiroffricethrough
telephone.PenPCenablesthemto
taketheinformation,toretrievemedi-
calinformationeasilyandtogivecor-
rectinformationtoadoctor.General-
ly,originalmedicalinformationis
storedinthehostcomputerandthe
latestinformationisdownloadedonto
thePenPCbeforethesalespeopleleave
theof6ce.Thissystemconfigurationis
showninFigure5.
5.MM・Dynanote
Wethinkthatinthenearfuture,por-
tableMMPCsw川appearonthemar.
ket.WearedevelopingaMMPenPC,
videoPCcardandMPEGIdecoderPC
cardtostrengthenvideoprocessing.
MMPenPChascolorVGALCD,PC
cardslot,CD-ROMandVideoforWin-
dowsastheOSdoes,sothetitlesoft-
ware・canbedisplayedonscreen.Ifa
MPEGIdecoderPCcardissetinthe
PCcardslot,moviescanbedisplayed
attherateof30framepersec.Ifa
videoPCcardisconnectedtoMMPen
PCviathePCcardslot,videodata,for
exampleTV,VCRandVideoCamera,
areenteredintotheMMPenPC.
6.Conclusion
ItisconsideredthatportableMMPCs
willbepopularinthenearfuture,and
theywillbeusedfbrmulti-mediapre-
sentationandvideoconferencing.Vid-
eoprocessingandintegrationtechnol-
ogyisneededtorealizeportable
MMPCs.Wehavebeendevelopingel-
ementarytechnologyandhavemade
severalproductsincludingsomeofthis
technology.Weareaimingatproduc-
ingaportableMMPenPCtobeused
byanyperson,atanyplaceandany
time.
47 JIQNo.101,1995
MultimediaHomeAppliance
anditsT「end
SukeichiMIKI
AudioVideoInformationTechnologyLaboratory
CorporateProductDevelopmentDivision
MatsushitaElectricIndustrialCo.,Ltd.
Abstracts
AsaresultoftheJapaneseGovern-
ment'sNationalBroadbandISDN(B-
ISDN)PlansaswellastheInfbrma-
tionSuperhighway,awidevarietyof
experimentsonmultimediasystemhas
beenstarting.Thisyearmaybethe
secondyearforexperimentsonmulti-
mediasystems.
Thispapergivesaforecastforhome
inf()rmationandcommunicationser-
vicesonemergingmultimediainfra-
structure.Information-orientedhome
apPliancerequiredbytheserviceswill
bealsodiscussed.
1.ANewInformationand
CommunicationInfra・
structureandaCulture
Change
IntheKansaiScienceCity,theBroad-
bandISDN(B-ISDN)experimentof
B-ISDNBusiness-chance&Culture
Creation(BBCC)startedlastyear.
VariousInfbrmationSuperhighway-re-
latedexperimentshavebeguninthe
UnitedStatessincelastyear.Ship-
mentsofCD-ROM,whichstoresmulti-
mediaonanopticaldisk,haverapidly
increased.Personalinformationter-
minalswithhandwritinginput(ink
media)havebeenintroducedtothe
market.ExperimentsforaPersonal
HandyPhoneSystem,whichisawire-
lessmultimediasystem,havealsobe-
gun・
Then,thisyearmaybethesecondyear
ofexperimentsofmultimediasystems.
Whenwethinkaboutthefutureof
multimediainformation-orientedhome
apPliances,itisimportanttoconsider
notonlythenewservicemadepossible
bytechnologicaladvancesbutalsothe
newlifestylesandculturewhichwill
becreated.Weareabletoforecastthe
specificationsofinformation-oriented
homeapPlianceproducts.
Inmultimediasystems,thereisacycle
wherein:
Changesintheregulatoryinfrastruc-
ture,suchasthosepertainingtobroad-
JIQNo.101,1995 48
castingandtelecommunication,and
developinginfrastructure,suchasafi-
ber-opticnetwork,willlendthemselves
tothecreationofnewmultimediahome
apPliances;thesenewproductswillin
tUrnCreateneWinf()rmatiOnServiCeS
and/oranewmultimediainformation-
orientedculture;and,thiswillacceler-
atethedevelopmentand/orexpansion
ofnewmultimediainfrastructure.
Thereisalsoareversecycle.Amulti-
mediasystemwillbeestablishedinthe
next10to20years.Thestartingpoint
forthisispreviously-mentionedcycles
basedonanexistingsystem.
2.MultimediaandExisting
Systems
Multimediasystemscombineavariety
ofdigitalcodedmediasuchascharac-
terdataandvideodata.
(2)Aninformationservicefordailylife
(aserVicethroughwhichwecanhave
spare-timemuchmorethannow,
100timesmore)
(3)Aninfbrmationservicefbreduca-
tion(aservicethroughwhichwecan
educateourselvesmuchmorethan
now,100timesmore)
(4)Aninf()rmationserviceforsocialre-
lationship(aservicethroughwhich
wecanenlargeacquaintancesmuch
morethannow,100timesmore)
4.ACIient・serverModel
withSeamlessAccess
Asmentionedabove,multimediasys-
temswillinitia}lycomeaboutthere-
sultofthefusionofthefollowingsys-
tems;
However,therecanexistasystemthat
combinesbothdigitalandanalogme-
dia.Inreality,itismostlikelythata
digitalsystemwillbeintegratedinto
anexiStinganalOgSyStemSOaStOCre-
ateanewmultimediasystem.
-thefusionofbroadcasting,telecom-
m皿ication,publishingandcomput-
ersystems,
-thefusionofpackagedmedia,wired
mediaandwirelessmedia,
3.NewMultimediaServices
andTheirImpacton
HomeAppliances
Newmultimediahome-informationser-
vicescanbeclassifiedintothefollow-
ingfourcategories;
(1)Aninformationserviceforentertain-
ment(aservicethroughwhichwe
canenjoyandbenefitfromTVmuch
morethannow,100timesmore)
-thefusionofexistingsystemsand
newmultimediasystems,and
-thefusionofavarietyofcodeme-
dia.
Then,thefusionoftheabovesystems
willresultinverysophisticatedand
complicatedsystems.However,we
anticipatethedevelopmentofamulti-
mediainfrastructureconsistingofa
deliveryand/orswitchingsystemcom-
49 JIQNo.lo1,1995
prisedofdistributedmu】t輌mediadata-
baseserversstoringmoviesandother
infbrmation,andthousandsofclient
terminals.Thesystemshouldprovide
fOraSeamleSSaCCeSS.
ThePersonalHandyPhoneSystemin
Japanhasadatatransmissionspeed
of32Kbps,andisaseamlessaccess
allowingwiredandwirelesssystemsto
belinked.TheAsymmetricDigital
SubscriberLine(ADSL)systeminthe
UnitedStateshasadatatransmission
speedof1.5Mbps,andcanalsobe
expectedtobecomeaseamlesssystem
capableofbothphone-lineswitching
andmoviedelivery.
5.MultimediaHomeAppli・
ances
Whentheanticipatedandrequisitein-
frastructureandmultimediasystems
comeintobeing,multimediahomeap-
plianceswillneedtopossessthefo1-
lowingfeaturesinordertomeetcon-
sumerneeds;
socialandconsumerneeds features
highqualityandlongproductlife → lowtroublerate
easyuse → human.machineinterface
suitabilityforsmaUroomsand
personaluse
→ smallsizeandenergysave
easeofrecordinginformation
forfutureuse
→ preciseandintelligentcontrol
communicationfunction
tocomm皿ities
→も コ
interaCtlvity
Consideringthefeaturesabove,thefol-
lowingaretypicalexamplesofmulti-
mediahomeapPliances;
一settopfbrinteractiveTV
一electronicbook
一WireleSSCOmmUniCatiOnterminal
一portableinf()rmationterminal
Abroadarrayofnewinformation-and-
communicationmultimediaapPlianc-
esastheaboveiscertaintobecome
availab}eintheinternationalmarket-
placewiththeincreasingestablish-
mentofinfrastructureintheyearsto
come.
JIQNo.101,1995 50
TheRealWorldComputingProgram
RyuichiOKA
TheoryandNovelFunctionsDepartment
RealWorldComputiingPartnership
TheRealWorldComputingProgram
(RWC)isanationalresearchprogram
fundedbyJapan,sMinistryoflnterna-
tionalTradeandIndustry(MITI)with
abudgetof$500millionfor10years
(1992-2001).TheRWChasasitsover-
alltechnicalobjectiveonthedevelop-
mentofflexibleandadvancedinfor-
mationtechnologies,thatarecapable
ofprocessingavarietyofdiversified
information(suchasimages,speech,
texts,andsoforth).TheRWCempha-
sizestechnologiesthatmatchthefiex-
ibilityofhumaninformationprocess,
ingcapabilitiessuchaspatternreco9-
nition,handlingofincompleteinfor-
mation,learningandself-organization,
allofwhicharemanifestedintheway
peoplesolveproblemsintherealworld.
TheapproachtotheobjectivesofRWC
researchprogrammaybestructurally
explainedasfollows(seeFigure1).The
flexibleinformationprocessingwillbe
basedonthreetypesofresearchcate-
gories:novelfunctions,parallelsys-
tems,andopticalsystems.Sincepat-
ternsandsymbolsarebasicandcom-
moninformationfoundinthereal
world,thenovelfunctionswillbere-
quiredtocarryouttheintegrationof
symbolsandpatternsfbrobtainingin-
tuition-likefunctions.Theintegration
willberealizedbyexploringbothprob-
lemsofrepresentationandalgorithms
whichoccurintherealworld.Repre-
sentationconcernstheframeworkfor
modelingtherealworldandisalso
stronglyrelatedtoparallelsystemsin-
cludingopticalsystems.Algorithms
concernthedynamicsforsolvingprob-
lemsformulatedintermsoftherepre-
sentation.Neuralmodelsareseenas
promisingcandidatestorealizeboth
therepresentationandalgorithm.
Fivemajorresearchthemes,described
here,willgiveamoreconcreteimageof
thetechnicalobjective.
Theoreticalfoundations:Thetheo-
reticalfoundationsforflexibleinfor-
mationprocessingwillprovidenew
methodsforsolvingill-posedproblems
bytheintegrationofsymbolsandpat-
tefnsinconnectionwithlearningand
self-organization.Inordertopromote
integrationofconventionalmethods,
theoriesforpatternrecognition,mu}ti-
variatedataanalysis,probabi}isticand
statisticinference,neuralcomputing,
andmachinelearningwiUbedeepened
andunified.
51 JIQNo.101,1995
辛ResearchCateg・ries
Rl⊥IC
FlexibleProcessingSystems(speechandgesturerecognition,etc}
蒙難 灘凝・
離離 難
Figure1
Novelfunctions:Thegoalistodevel-
opthetechnologiesfbrflexibleinfor-
mationprocessingthatconventional
technologylacksintermsofrobust-
ness,openness,andreal-timecapabili-
ty.Specificresearchgoalsinclude:
(i)Establishingschemesforflexible
recognitionandunderstandingof
multi-modalinformation,includ-
ingmovingimages,speech,texts,
andgestures,anddevelopingin-
teractive.informationsystemswithwhichhu'manscancommunicate
throughsuchinformationmedia
suchas
・Gracefulman-machinedialogue
systemsusingspeech,facialex-
pressionsandgestureswithatleast
a3000vocabularyunit.
・Informationretrievalsystemswith
learningandinferencecapabilities
forunderstandingusers'intentions
fornewspaperarticlesofseveral
yea「s・
・Speechunderstandingsystems
whichcandetectandunderstand
specialtopicsinatleast2hours
conversationspeechusing20r3
thousandkeywords.
(ii)Developingflexiblesystemsthat
areabletoautonomouslysenseand
understand,andalsocontroltheir
realworldenvironment.
JIQNo.101,1995 52
・Robotswhichimitatehumanac-
tionsfbrperformingcooperatlve
works,likewrappinganddeliver-
ingobjectstohumans・
Massivelyparallelcomputing:
Amassivelyparallelsystem(called
RWC-1)whichwiUbedevelopedto
supPortthe`realworldcomputa-
tion.'Researchgoalsinclude:
(i)Providingmultiplecomputation
paradigmssuchassharedmemo-
ry,messagepassing,dataparallel,
multi-threadingandneuralnet-
works.
(ii)Establishingflexibleadaptationto
theexecutionstatusofproblems
andcomputationresources,i.e.
suchparallelsystemswillhavethe
mechanismsfbrbeingadaptedto
thesystemloadcondition,variety
ospropertiesofinformation,re-
quirementsofreal-timeness,ro-
bustnessforbothhardwareand
software,etc.
(iii)Providingcomputationandmemo-
rycapacitytopermitanunpredict-
ableamountofcomputationwhich
■illal}ow .programmerstogive
primitivedescriptionofthejrprob-
lems,and
(iv)Providingextremelyfastintercon-
nectionnetworksandsynchroni-
zationmechanisms,bothofwhich
areessentialformassivelyparal-
lelprocessingandtheimplemen-
tationofneuronsandelementsof
othernovelmodels.
Thesoftwarearchitectureofthemas-.
sivelyparallelsystemconsistsofthe
kernel,implementationlanguage,base
language,operatingsystem,andpro-
grammingenvironmentlayers・The
baselanguagesupportsthenotionof
object-orientation,reflection,andtime-
dependentprogramming.Usingrefiec-
tion,theprogramminglanguageseman-
ticsaswellastheresourcemanage-
mentstrategyismodifiedbythepro-
gramminglanguageitsel£Thenew
languagefunctionalitiesmightbeeasi-
lyconstructedbythebaselanguage.
Sincewehavetoconsider``time"inthe
realworld,thebaselanguageprovides
thefacilitytodescribethereal-time
system.
.Neuralnetworksystems:Herethe
goalistoestablishanewtypeofneural
modelorcomputation,whichisdiffer-
entfromconventionalmodelssuchas
back-propagation,Boltzmannmachine,
andsoforth.Connectionistmodelswill
bepursuedascandidatesthatwillpro-
videanewschemeforrepresentation
oftherealworld.Thedomainof`neu-
ral'researchshouldbeextendedtodeal
withnoisy,uncertainandincomplete
informationencounteredinthereal
world.
Opticalcomputingsystems:the
RWCaimsatestablishingbasictech-
nologyfor;
(i)Opticalinterconnectiondevices
andnetworks.
(ii)Opticalneuralmodels,devices,and
systems.
53 JIQNo.101,1995
(iii)Opticallogicdevices,circuits,and
digitalsystems,and
(iv)Advancedopto-electronicintegrat-
edcircuitsdevelopmentenviron-
ments.
Summary:TheobjectivesoftheRWC
meetthechallengingproblemsinpat-
ternrecognitionandartificialintelli-
genceinbothconceptualandcomputa-tionalaspects.
First,theapproachtowardstheinte-
grationofsymbolsandpatternsshould
leadtonewmodelsandalgorithmssuit.
abletorealizefunctionswithrobust-
nessandopennessbeyondthatofcon-
ventionalpatternrecognitionandarti-
ficialintelligence.Scalablealgorithms
withsimplicityandtransparencywill
alsobeexpectedtoprocessalarge
amωntofinfbrmationsuchasimages,
speech,textandsoforth.
Second,real-timerealizationofthese
modelsandalgorithmswiUrequirea
newarchitectureofmassivelyparallel
computationwhichpossessesawide
spectrumofrealtimeInput/Output
channels,thatmakeclosecontactwith
therealworld.
TypicalapplicationsofRWCwillin-
cludetherealizationofinformationsys-
temscapableofsupportinghumanac.
tivities,suchas:i)anewviewpointfor
humansintheircreativeactivitiesby
meansofautomaticsearchandinfer-
enceofinformationinlargedata-base
includingimagesandspeech,ii)anin.
terfacebetweenhumansandtheiren,
vironment,e.g.supportingdisabled
peoplebymeansofintelligentmoni-
torlng.
Inconclusion,theRWCprogramisnot
asimplesuccessoroftheFifthGenera-
tionComputerProject,butwillpro-
motechallengingresearchtosolvein-
trinsicproblemsrelatedtoartificial
intelligenceandpatternrecognition.
ThekeyconceptoftheRWCisthusthe
integrationofsymbolsandpatternsfor
realizingintuition-likefunctions,which
allowsthecomputertocommunicate
intermsofpatternlikeimagesand
speech,whileusingsymbolicrepresen-
tationfortheinformationprocessing.
Appendix
TheresearchthemesoftheProgram
currentlypursuedbythePartnership
areasf()UOWS:
[TsukubaResearchCenter(TRC)
andDomesticDistributedLabora-
tories]
・EcologicalandEvolutionModelsfor
MassivelyParallel/DistributedSys-
tems(TheoryFujitsuLaboratory)
●StatisticalInferenceasaTheoretical
FoundationofGeneticAlgorithms
(TheoryGMDLaboratory)
・AVisionProcessorinNeuralArchi-
tecture(TheoryMitsubishiLaborato-
ry)
・ComputationalLearningTheoryof
ProbabilisticKnowledgeRepresenta-
JIQNo.101,1995 54
tions(TheoryNECLaboratory)
●InformationIntegratingInteractive
Systems(TRCInformationIntegra-
tionLaboratory)
・RealWorldAutonomousSystems(TRC
ActiveIntelligenceILaboratory)
●LearningandGrowthFunctionsfor
AutonomousMobileRobot(Novel
FunctionsFujitsuLaboratory)
・InfbrmationIntegrationTechnology
forApplyingSignLanguageRecogni-
tion(NovelFunctionsHitachiLabo-
ratory1)
.Desk-workSupport-basedonEpisod-
icMemory(NovelFunctionsHitachi
Laboratory2)
・Self-OrganizingInfbrmationBases
(NovelFunctionsMitsubishiLabora-
tory)
・Multi-ModalHumanInterfacewith
SecretaryAgents(NovelFunctions
SharpLaboratory)
・ProgrammingInteractiveRea1-Time
AutonomousIntelligentAgents(Nov-
elFunctionsSICSLaboratory)
●ActiveperceptionCognition(Novel
FunctionsSNNLaboratory)
.FlexibleStorageandRetrievalof
Multi-mediaInformation(NovelFunc-
tionsISSLaboratory)
・AdaptiveEvolutionComputers(TRC
NeuralSystemLaboratory)
・ANewModelofNeuralNetworks
CalledNeuralLogicNetworks(Neuro
ISSLaboratory)
・PatternRecognitionBasedonStruc-
turedNeuralNetworks(NeuroToshi-
baLaboratory)
・GenericTasksforSymbolInforma-
tionProcessingandPatternInfbrma-
tionProcessing(NovelFunctionsMRI
Laboratory)
・ParallelInformationProcessingMech-
anismsandAttentionMechanismsin
theBrain(NovelFunctionsNTTLab-
oratory)
・CooperativeProblemSolvingBased
onHeterogeneousKnowledge(Novel
FunctionsOkiLaboratory)
・VisionBasedAutonomousSystems
(NovelFunctionsSanyoLaboratory)
・MassivelyParallelComputationMod-
el,OS,ProgrammingLanguageand
En'vironment(TRCMassivelyParal-
lelSoftwareLaboratory)
・MassivelyParallelExecutionModels
andArchitectures(TRCMassively
ParallelArchitectureLaboratory)
・Development,Implementation,and
EvaluationofaProgrammingModel
forMassivelyParallelSystems(Mas-
sivelyParallelSystemsGMDLabora-
tory)
・MassivelyParallelObject-Oriented
55 JIQNo.101,1995
Models(MassivelyParallelSystems
MitsubishiLaboratory)
・MassivelyParallelProgrammingEn-
vironment(MassivelyParallelSys-
temsMRILaboratory)
・AdaptiveMassivelyParallelSystems
(MassivelyParallelSystemsNEC
Laboratory)
・ResourceManagementintheMas-
sively・ParallelComputing(Massive-
lyParallelSystemsSanyoLaborato-
ry)
・AMassivelyParallelMachinewith
OpticalInterconnection(Massively
ParallelSystemsToshibaLaborato-
ry)
tronicsMitsubishiLaboratory)
・Electro-PhotonicProcessorNetworks
(OptoelectronicsNECLaboratory)
・OpticalBusInterconnectionSystems
(OptoelectronicsNSGLaboratory)
・3-DimensionalOptoelectronicInter.
connection(OptoelectronicsOkiLab-
oratory)
●Researchof3D-IntegratedStacked
OpticalDevicesforOpticalComput-
ingSystemsandtheirApplications
(OptoelectronicsSanyoLaboratory)
●ParallelOpticalInterconnectionby
OpticalFibers(OptoelectronicsSum-
idenLaboratory)
●SpecialLightDeflectors(Opto-elec-
tronicsFujikuraLaboratory)
・OpticalInterconnectionbyWave-
lengthDomainAddressing(Optoelec-
tronicsFujitsuLaboratory)
・WavelengthTunableSurfaceEmit-
tingLDArray(OptoelectronicsFu-
rukawaLaboratory)
.OpticalInterconnectionandsignal
ProcessingExploitingthroughOpti-
calFrequencyAddressing(Optoe|ec-
tronicsHitachiLaboratory)
・StackedOpticalComputingSystems
(OptoelectronicsMatsushitaLabora-
tory)
●OpticalNeurocomputing(Optoelec-
・Multi-FunctionalSurfacleOptical
DevicesforOpticalInterconnection
(OptoelectronicsToshibaLaboratory)
[OverseasSubcontractors]
・LearningaMapwithaStereo-Vision
BasedSystem(IstitutoperlaRicerca
ScientificaeTecnologica)
・HybridEvolutionaryProgramming:
DevelopmentofMethodologyand
ApPlicationsinSoftautomationfor
HighAutonomousandIntelligent
RoboticSystems(GoedelSchoolSoft-
wareDevelopmentCompany)
・IntegratedInformationProcessing
forPatternRecognitionbySelf-Orga-
nization,PrototypeOptimization
andFuzzyModeling(Universityof
JIQNo.101,1995 56
Sydney)
[DomesticSubcontractors]
BasedonMassivelyParallelProcess-
ing(TsunenoriMine,CollegeofGen-
eralEducation,KyushuUniversity)
Theory/NovelFunctions
・Bi-DirectionalTranslationofPattern
andSymbolInformation(Takashi
Omori,DepartmentofElectronicsand
InformationScience,TokyoUniversi-
tyofAgricultureandTechnology)
・ModelingofVisualRecognitionBased
ontheInteractionbetweenPattern
InfbrmationandSymbohcInforma-
tion(MasumiIshikawa,Department
ofControlEngineeringandScience,
KyushuInstituteofTechnology)
・DiscriminationofSpokenLanguages
andDialects(ShuichiItahashi,Insti-
tuteoflnformationSciencesandElec-
tronics,UniversityofTsukuba)
・SpontaneousSpeec}1Understanding
(TetsunoriKobayashi,Departmentof
ElectricalEngineering,WasedaUni-
versity)
・Example-BasedTranslationonMas-
sivelyParallelcomputers(Satoshi
Sato,SchoolofInformationScience,
HokurikuAdvancedInstituteofSci-
enceandTechnology)
・AcquisitionofLinguisticKnowledge
forNaturalLanguageProcessingfrom
TextCorpora(TakenobuTokunaga,
DepartmentofComputerScience,To-
kyoInstituteofTechnology)
・NaturalLanguageUnderstanding
・UnifiedPlanningofRecognitionand
ActioninChangingEnvironments
(YoshiakiShirai,DepartmentofMe-
chanicalEngineeringforComputer-
ControlledMachinery,OsakaUniver-
sity)
・Description/UnderstandingofImag・
esandPerformanceEvaluationoflm-
ageprocessingAlgorithms(Keiichi
Abe,DepartmentofComputerSci-
ence,ShizuokaUniversity)
・ParallelCooperativeImageUnder-
standingSystems(TakashiMatsuya-
ma,Departmentoflnf()rmationTech-
nology,OkayamaUniversity)
・AdaptiveObjectModelforChanging
Environment(MichihikoMinoh,In-
tegratedMediaEnvironmentExperi-
mentalLaboratory,KyotoUniversi-
ty)
・ImageUnderstandingofPresentation
MediabyIntegrationofMultipleIn-
f()rmationsSources(YuichiOhta,In.
stituteofInformationSciencesand
Electronics,UniversityofTsukuba)
● ComputerVisionAlgorithmBasedon
CooperativeComputation(Naokazu
Yokoya,InformationTechnologyCen-
ter,NaraInstituteofScienceand
Technology)
・Active-ControlRobotHeadforVision-
soundInformationIntegration(Hiro-
57 JIQNo.101,1995
chikaInoue,DepartmentofMechano-
Infbrmatics,UniversityofTokyo)
・TaskExecutionSystemwithMulti-
SensorFusioninHumanRobotCoop-
erativeWork(TomomasaSato,Re-
searchCenterforAdvancedScience
andTechnology,UniversityofTokyo)
noruTsukada,DepartmentofInfor-
mationandCommunication,Tama-
gaWaUniVerSity)
・ComputationalDynamicsofChaotic
NeuralNetworks(KazuyukiAihara,
DepartmentofMathematicalEngi-
neeringandInformationPhysics,Uni.
versityofTokyo)
NeuralSystems
MassivelyParallelSystems
・DynamicsofNeuralNetworks(Shuji
Yoshizawa,DepartmentofMechano-
Infbrmatics,UniversityofTokyo)
●FeedbackLearningonNeuralNet-
work(YoichiOkabe,ResearchCenter
forAdvancedScienceandTechnolo-
gy,UniversityofTokyo)
・Modeling3-DVisualInformationby
NeuralProcessingSystems(YuzoHi-
rai,InstituteofInf()rmationSciences
andElectronics,UniversityofTsuku-
ba)
・NeuralMechanismsandInformation
RepresentationforColorVision(Shi-
roUsui,DepartmentofInfbrmation
andComputerSciences,Toyohashi
UniversityofTechnology)
・ConstraintsSatisfactionSystemsUs-
ingHopfieldNeuralNetworkMod-
ules(YutakaAkiyama,Institutefor
ChemicalResearch,KyotoUniversi-
ty)
・Temporal-PatternDependentLearn-
ingRuleandaModeloftheHippoc-
ampalCorticalMemorySystem(Mi-
・RealTimeMusicInformationProcess.
ingBasedonParallelProcessing(Yoi-
chiMuraoka,Departmentoflnforma.
tionandComputerScience,Waseda
University)
・FlexibleComputerassistedAnalysis
SystemsforNgn-LinerProblems
(ShinichiOishi,DepartmentofInfbr-
mationandComputerScience,Wase-
daUniversity)
・DistributedShared-MemorySystems
forMassivelyParallelProcessingSys-
tems(KeiHiraki,DepartmentofIn-
f()rmationScience,UniversityofTo-
kyo)
●SuperParallelArchitectureBasedon
F皿ctionalModel(ToshioShimada,
DepartmentofInformationElectron-
ics,NagoyaUniversity)
●Object-OrientedConcurrentDescrip-
tionFrameworksforMassivelyPar-
allelComputing(AkinoriYonezawa,
DepartmentofInformationScience,
UniversityofTokyo)
JIQNo.101,1995 58
OpticalSystems
・SemiconductorLaserDiodesforOpti-
callyTriggeredDigitalIC's(Takeshi
Kamiya,DepartmentofElectronicEn-
gineering,UniversityofTokyo)
・OpticalPropertiesofQuantumwell
StructuresinaMicroCavityandthose
ApplicationfbrSemiconductorLaser
Switches(MasahiroTsuchiya,Depart-
mentofElectronicEngineering,Uni-
versityofTokyo)
・OpticalParallelDigitalComputers
(YoshikiIchioka,DepartmentofAp-
pliedPhysics,OsakaUniversity)
●ArchitectureandPackagingTech-
niqueofOpticalComputer(JunTani-
da,DepartmentofAppliedPhysics,
OsakaUniversity)
・DevelopmentofanlntegratedOptical
NeuralNetworkModule(Toyohiko
Yatagai,InstituteofAppliedPhysics,
UniversityofTsukuba)
・LearningCapabilitiesandMassively
ParallelProcessing(MasatoshiIsh-
ikawa,DepartmentofMathematical
EngineeringandInformationPhys-
ics,UniversityofTokyo)
59 JIQNo.101,1995
MassivelyParallel
RWC・1
Computer
ShuichiSAKAI
MassivelyParallelArchitectureLaboratory
RealWorldComputingPartnership*
Abstract
Thispresentsamassivelyparallelcom-
puterwhichisbeingdevelopedbyReal
WorldComputing(RWC)Programin
Japan.Thepurposesofthisresearch
anddevelopmentaretoeffricientlysup-
portflexibleandintegratedinfbrma-
tionprocessingwhichareresearchtar-
getsinRWC,andtopursueageneral
purposestand-alonemassivelyparal-
1elsysteme笛cientlysupPortingmul-
tipleprogrammingparadigms.Forthe
purposes,anewmassivelyparallel
computerRWC-1isnowunderdevel-
opmentwithastrongcollaboration
amonghardwarepeople,softwarepeo-
pleandapPlicationpeople.
ThepresentationshowstheRWC-1
architecturalfeatures,RWC-1software
featuresandtheplanfbrmachinede-
velopment.
Thearchitecturalfeaturesofthissys-
temare:(1)ReducedInterprocessor-
Comm皿icationArchitecture(RICA)
wherecommunication,schedulingand
instructionexecutionaretightlyinte-
grated,(2)supPortforamassivelypar-
alleloperatingsystem,(3)acubecon-
nectedcircularbanyaninterconnection
network,and(4)anindependentI/O
network.
Figurelillustratesanorganizationof
RICA.RICAmeansanarchitectural
fusionofcomputationandcommunica-
tionandsimplificationofthewhole
structure.
Amassivelyparalleloperatingsystem
isnecessaryfbranef6cientandsafe
stand-alonecomputer.Aglobalvirtu.
almemoryisoneofthemostimportant
technologiesforachievingit.Wepro-
posetheglobalvirtualaddressing,wherePEaddressisvirtualizedaswell
aslocalmemoryaddress.
*TsukubaMitsuiBuilding16F ,Phone=+81-298-53-1650Fax:+81-298-53-1652
EMAIL:[email protected]
1・6-1Takezono,Tsukuba,Ibaraki305,Japan
JIQNo.101,1995 60
Message lnterconnectionNetwork
HIIMIIMMessageQueue
MessageHandling
Pipeline
(instruction_address,data_address,d1,d2,,
睡
R|SC{Superscalar}Pipe|ine
「Message
GenerationPipeline
ThreadEx㏄utor
FigurelReducedInterprocessorCommunicationArchitecture
Asaninterconnectionnetwork,we
adoptedCubeConnectedCircularBan-
yan(CCCB)(Figure2),since(1)itneeds
onlyasmallhardware,(2)operates
withahighband-widthandsmallde-
lay,(3)performsselfrouting,(4)store-
and-forwarddeadlockiseasilyprevent-
edwiththisnetwork,andsoon.
Figure3illustratestheRWCsoftware.
Themainsoftwarebeingdevelopedin
TRCisakernelSCore,adescription
languageMPC++,andabaselanguage
OCorε.
TRCisnowdevelopingamassivelypar-
allelprototypecalledRWC-1.RWC-1
willhave1,024PEsandwillbeassem-
bledinMarch1996.RWC-1isanex-
perimentalmachineforprovingthe
concepts,mechanicalfeaturesandeffi-
ciency.Theabsolutepeakperformance
isnotthepurposeofthissystem.
61 JIQNo.101,1995
Figure2CubeConnectedCircularBanyanNetwork
JIQNo.101,1995 62
ApPlications
Language
Extensions
BaseLanguage
(OCore)
Programming
Environment
Operating
System
ImplementationLanguage(MPC++)&Kemel(SCore)
Hardware
Figure3RWCSoftware
RWC-1containsoriginalprocessor
chips,originalinterconnectionswitch-
esandoriginalmaintenancehardware.
Asfbrthefirststep,wehavebeende-
velopingthefirsttestedsince1993.
63 JIQNo.101,1995
CooperativeInteractionsinFuzzy
DecisionSupPortSystems
KoichiYAMADA
AdvancedTechnologyCenter
Yamatake-HoneywellCo.,Ltd.
Introduction
Human-computerinteraction(HCI),or
thehumaninte㎡ace,isoneofthemost
vigorousresearchtopicsincomputer
scienceandtherelatedareas.Howev-
er,therearemanyconflictingimages
ofHCIresearch,andsometimesthe
fieldseemstobechaotic.Whatever
theviewpoint,thefinalgoalisacom-
monone;realizationofeasy,intelli-
gentandfriendlysystems-systems
withcooperativeinteractions.
Therearetwomajorstreamsofresearch
thatdealdirectlywithcooperativein-
teractions.Oneiscooperativeanswer-
ingininformationretrieval[1].How-
ever,mostresearchinthefieldisde-
pendentonaprioridomainknowledge,
oronheuristicsinthedomain.The
otherisanapproachthatdealswith
theuSer'SintentiOnSinmoregeneral
dialogues[2][3].Theyare,however,
tooprecisefbrUseinpracticalsystems.
neeringResearch(LIFE).First,we
demonstrateanapProachforcoopera-
tiVeinteraCtiOnSininfOrmatiOnretrieV-
alsystemswhichrequiresnoapriori
knowledgeorheuristicsdependenton
thedomain.Then,wediscussanother
approachthatutilizestheuser'spref-
erences.
CooperativeAnswersto
Queries[4]
Informationretrieval(IR)isoneofthe
mostbasicdecision-makingtasks.
However,thetaskisnotsosimplefor
mostdecisionmakers,whooftenhave
toaccessunfamiliardatabasesusing
unfamiliarIRsystems.Insuchcases;
queriesfrequentlyfail(retrievenodata)orretrievetoomuchdatatoexamine
allofthem.Then,theyhavetocon-
StrUCtaneWqUerytOretrieVeaman-
ageablenumberofdata.Butthereis
stillnoguaranteethatthisnexttrial
willbesuccessful.
Inthispaper,weintroducesomealter-
nativeapProachesemployingfuzzylogic
whichwehaveinvestigatedattheLab-
oratoryforInternationalFuzzyEngi-
Cooperativeansweringisonetechnique
tosupPortsuchunhapPyusers.CooP-
erativeanswersarethosethatgivethe
usersomehelpfulhintsforconstruc-
JIQNo.101,1995 64
tionofthenextquery.Thesearegiven
when1)thequeryfails,orwhen2)too
muchdataisretrieved.Unlikeap-
proachesproposedsofar,ourapproach
generatescooperativeanswerswithoutdomaindependentheuristicsand
knowledgegivenapriori.Instead,it
createsknowledgeaboutthedistribu-
tionofdatainthedatabasefroritself,
andutilizesit.Thecreatedknowledge
iscalledamacrodatabase,becauseit
describe3thecontentsofthedatabase
globally.
nrtionofhemacrdatabase
Themacrodatabaseisasetoflinguis-
ticexpressionseachofwhichdescribes
afuzzyclusterofdatainthedataspace.
Itisgeneratedbythefollowinga}go.
rithm:
fromthedata.Inthefirsttrial,set
thenumberofclusterstotwo.
(4)Trytoexpressthegeneratedfuzzy
clustersbythelinguisticlabelsde-
finedbefore.IfthefUzzyclusters
areexpressedwellbythelinguistic
labels-thatis,theirprojectionsto
attributeaxesarealmostincluded
inmembershipfunctionsdefinedon
theattributes,gotothenextstep.
Otherwise,increasethemlmberof
clusterbyone,andgoto(3).
(5)Expressallfuzzyclustersbythe
combinationsoflingUisticlabels.
Then,eachclusterisafactorofthe
macrodatabase.
Cooerativeanswerinwhenauer
fails
(1)Expressnominalattributesbynu-
mericalones;e.9.inthecaseofan
apartmentdatabase,theattribute
"floortype,,(seeFigure2)isexpressed
byitsaveragefloorspace,and``the
neareststation"byapairofmap
coord輌nates.
Theuser,squeryisgiveninaconjunc-
tionoffuzzyconditionsexpressedin
fuzzysetsonattributes.Ifthematch-
ingdegreeofdatatothequeryisgreat-
erthanacertainvalue,thatdatais
retrievedfromthedatabase.Ifthere
isnosuchdata,thequeryfails.
(2)Definelinguisticlabelsbymember-
shipfunctionsonthesenumerical
attributes(SeeFigure1).Eachlin-
guisticlabelexpressesacertaincon-
ceptontheattribute.Thesemem-
bershipfunctionsmustcoverthe
entireuniverseofdiscourseofthe
attributes.
(3)Applyafuzzyclusteringmethod
calledFuzzyC-meanstodatainthe
database,andcreatefUzzyclusters
Whenaqueryfails,thesystemgives
theuserthefollowinginformationin
order;1)thenearestclustertotheque-
ry,asasetofdatathattheusermight
accept,2)thenearestdatatothequery
asanalternative,then3)compromises
thattheusermustacceptifnotsatis-
fiedwiththealternative(Figure2).
Thelinguisticexpressionsofthemac-
rodatabasedescribedbeforeareused
toprovidetheinfbrmation.
65 JIQNo.101,1995
X2
L11 Li2 Li3
X1
×:datainfuzzyclusters
x1,x2:axcsofattributes
k=1,2,3:fu7.zYclusters
Lk':linguisticlabclsdefinedlbymcmbershipfunctions
FigurelFuzzyClustersandTheirLinguisticExpressions
U:Showmcsomcapartmcntswith
由encarcs【stalation=Tama-Plaza,
noortypc=IR(Oncroom).
S:Thereisnosuchanapartmcnt.
But,therearesomeapartmentsyoumightwant;
theneareststata1ion=aroundKajigaya,
floortype=aroundIR.
Howaboutthenextone,forexample:
theneareststation=Miyamae-daira,
tlmetos田uon=7mtn.,
howold=Oyear,
floortype=IK(1bedroomandsmallkitchen),
Rent=73,000yen.
IfyouwantanapartmentaroundTama-Plaza,
theavailableapartmentsarelimitedto
floor-type:morethan3LDK(3bcdrooms).
Figure2AnExampleofCooperativeAnswersofRealEstate
Advisor[4]
」lQNo.101,1995 66
OOerativeanSwerinwhentOOmUCh Usermodelinindecisionsu r
dataisretrieved tems
Queriessometimesreturntoomany
datafromthedatabasefortheuserto
examineaUofthedataprecisely.In
thatcase,theusermightwanttore-
finetheretrieveddatasetbygiving
additionalconditions.However,thisis
notsoeasy,becausethereisnoextra
informationtoshrinkittoaproper
size.So,thesystemgivestheuserthe
linguisticexpressionoftheretrieved
datasetandthemlmberofdata,then,
aftercheckingthedistributionofthe
retrieveddatasetforeachattribute,it
urgestheusertoprovideanextracon-
ditionfortheattributesforwhichthe
distributionisthelargest.
DecisionSupPortwithaUser
Model
Inconsultingsystemssuchasapart-
ment,hunting[5]andbook-selection[6]
advisors,theuseroftendoesnothave
anyclearpriorimageabouthis/herde-
cision.Whatthesystemmustpropose
intheseapplicationsis``afavorable
decision"ratherthan``thebestone."A
usermodelexpressingtheuser'spref-
erencesiseffectivefbrthispurpose.
Preferences,fbrexample,canbeused
asadditionalconditionsfbrqueriesin
dataretrievalsystems,whentoomuch
dataisretrieved,andcanalsobeused
asinformationforderivationoffuzzy
goalsforinteractivefuzzyprogramming
[7],whichisafuzzyversionofmulti.
objectiveprogramming.
Ingenera1,therearetwowaystobuild
ausermodel.Oneisexplicitmodel-
ing,whichrequirestheusertoinform
thesystemofhis/herpreferenees.The
otherisimplicitmodeling,wherethe
system.gue .ssestheuser,spreferences
duringinteractions.Fortheaimof
cooperativeinteractions,implicitmod-
elingispreferable.
Stereotyping[6]isoneofthemajor
approachesfbrimplicitmodeling.A
stereotypeisknowledgethatrepresents
typicaltraitsofatypicaluser.Italso
hasitsactivatingconditions,whichare
usuallybasedonthebasicprofileof
theuser,suchasoccupation,gender,
age,etc.Usingstereotypesasknowl-
edge,thesystemcanguessmanyas-
pectsoftheuser'spreferencesinthefirststageofinteractions.
However,stereotypinghasinherent
shortcomings-theinitiallyobtained
usermodelusuallyincludessome
wrongguesses,so,thesystemalways
hastomonitortheinteractions,infer
theuser'spreferencesfromchlesin
interactionsandcorrectthewrong
guesses・
Theuser'spreferences,however,are
noteasytoinferfromtheinteractions,
becausetheusertalksaboutdifferent
levelsofpreferences.Sometimeshe!
shemaysaydirectly;"Ipreferacheap-
erapartment."Inothercases,theuser
maymentionahigherlevelofprefer-
encesuchas"Iwantsafety,"which
67 JIQNo.101,1995
letsthesystemdeducethathe/she
wantsanapartmentwithacaretaker
and/orremotelocksystem.Further-
more,whentheusertalksaboutcon-
creterequirementssuchasthedesire
foranapartmentwithtiledwalls,itis
possibletohypothesizethathe/shepre.
fersa"good-lookinピ'apartment.From
thehypothesis,youcanalsoassume
thattheuserprefersanewapartment。
in[o,1],andafuzzysetQonQ(theset
ofQLs)aregiven,fuzzyabductionisthe
procedurebywhichtoobtainafuzzy
set(fuzzyexplanation)RonP(theset
ofPis)thatderivesΩwithR.Here,"2
derivesQ"meansthatthefollowing
equationholds:
qj=max(Pi+rij-1)・
i・P・ † ・・j≧1 (1)
Thoughthesystemwhichweareim-
plementingatLIFEemploysonlythe
directandthedeductiveapproaches,
wediscussthethirdapproachofhy-
pothesizing,whichinvolvestheinfer-
encingprocesscalledabduction.
-Abductionistheinferencingprocessby
whichtoderiveasetofhypothesesthat
explainsagivensetofeventswithaset
ofrules.Aderivedsetofhypothesesis
calledanexplanation.Asasimpleex-
ampleofabduction,supposethatyou
lovetogototheraces,andthatafriend
saysyoumustbeagambler.Inthis
case,probably,hisinferenceisNOTa
deduction,becausehedoesnotseemto
uSearUleSaying"yOuareagambler,
becauseyouliketogototheraces."A
moreprobableruleis"youliketogoto
theraces,becauseyouareagambler."
So,whathedidwasfindingahypothe-
sisthatexplainsthefactthatyoulike
togototheraces.Thatisanabduc-
tion.
wherepiandqjaretruthvaluesofP,
andQj,andarealsomembershipval-
uesofPiinPandQjinQ・respectively・
Theoretically,thereisnoguarantee
thatfuzzyexplanationsalwaysexist.
However,ifanyfuzzyexplanationsdo
exist,thereare,ingeneral,onlythe
largestandmultipleminimalfuzzyex-
planations.Thelargestfuzzyexplana-
tion,2maxisgivenbythefollowingequa-
tions:
Rmax=◇(PijA/Pi),
i,j
・・」・ヤ+1。蕊
where◇isanoperatordefinedas:
(2)
(3)
a・A◇b/B・ぱ3/A:iQ::(・)
Then,Pkmin(k=1,...,Nmin),giveninthe
nextequatiOn,iSaminimalfuzzyex-
planation,ifitisasubsetofPmaxand
doesnotincludeotherEピmin(k'≠k).
Fuzzyabduction[8]isdefinedasf()1-
lows;whenasetofrulesR,whereeach
・ul・"R、j・.Pに ・Q、"h・ ・at・uth・alu・ ・ij
1Lkmin=Σ(△(qj-rij+1!Pi)),
j,qj≠Oi∈ φ(j)
(5)
JIQNo.101,1995 68
Rl:Good-looks->New(1.0)
R2:Good-looks->Tiled-wall(1.O)
R3:Good-looks->Remote-lock-system(O.6)
R4:Safety->Remote-lock-system(1.0)
R5:Safety->Caretaker(0.8)
R6:Safety->Residential-area(O.8)
R7:Comforい>Caretaker(0.8)
R8:Comfort->Basement-garage(0.8)
R9:Silence->Residential-area(1.0)
RIO:Silence->Sound-proof(0.6)
(a)Rulesthatderivelower-levelpreferences
fromhigher4evelpreferences.
Remote-10ck-system(0.8)
Caretakcr(0.6)
Residential-area(1.0)
Sound-proof(0.6)
(b)Theuser'slower-1evel
preferences.
0.8/Safety+ .1.0/Silence
(c)higher-levelpreferences
thatexplain(b).
Figure3 111ferencingHigher-levelofPreferencesby
FuzzyAbduction
wheref(j)meansasetof"i"ssuchthat
㌔ -qj≧Oforthegivenj,and△iisan
operatorwhichchoosesatermfrom
amongstthosewithdifferent"i"s.
easonln referencesbfuzzabduc-
迦
SupPosethatthereisarealestateagent
whohassomeknowledgeabouttheten-
denciesofapartmenthunters.Forex-
ample,customerswhowantgood-look-
ingapartmentsalwaysrequirenew
oneswithtiled・walls,andtendtowant
remotelocksystemsasshowninR1,
R2andR3inFigUre3(a).Inthese
rules,preferencesintheantecedents
arecustomers'basicinclinations,or
higher-levelpreferences.Thoseinthe
COnSequentSaretheirCOnCreterequire-
ments,orlower-levelpreferences.
Now,supPoseagainthatacustomer
hascometotheagency,andthatthe
agentunderstandshim/hertowantan
apartmentshowninFigure3(b),after
somediscussionwithhim.Then,the
agentcanapPlyfuzzyabductiontoin-
ferthecustomer'shigher-levelprefer-
ences.Sincethebestexplanationof
thegivenfactsisusuallygiveninmin-
imalexplanat輌ons[8],weuseeq.(5)to
obtainasolution.Inthiscase,onlya
minimalfuzzyexplanationisobtained,
asshowninFigure3(c).
Conclusions
Afewapproachesrelatedtocoopera-
tiveinteractionsinfUzzydecisionsup-
portsystemsareproposed.Inthese
apProaches,fuzzylogicplaysimpor-
tantrolesindealingwiththefuzziness
whichisessentialinmacroexpression
ofadatabase,andinusers'preferenc-
es.
69 JIQNo.101,1995
References
[1]T.Gaasterland,etal.:AnOverview
ofCooperativeAnswering,Journal
ofIntelligentInformationSystems,
Vol.1,pp.123-157(1992).
[2]J.F.Allen,C.R.Perrault:Analyzing
IntentioninUtterances,Artificial
Intelligence,VoL15,pp.143-178
(1980).
[3]K.Yamada,R.Mizoguchi,etal.:
ModelofUtteranceandItsUsein
CooperativeResponseGeneration,
LectureNotesinComputerScience
753,Human.ComputerInteraction
(Eds.L.J.Bass,etaL),Springer-
Verlag,pp.260-271(1993).
[4]J.Ozawa,K.Yamada:Cooperative
AnsweringwithMacroExpression
ofADatabase,Proc.IPMU,94,pp.
17-22(1994)
[5]N.Harada,K.Yamada,etal.:User
ModelingforCooperativeDecision
SupportwithFuzzyRepresentation,
FirstAsianFuzzySystemSympo-
sium,pp.142-150(1993)
[6]E.Rich:Usersareindividuals:indi-
Vidualizingusermodels,Internation-
alJournalofMan-MachineStudies,
VoL18,pp.199-214(1983)
[7]A.Nukuzuma,K.Yamada,etaL:
DecisionSupPortSystemAimingat
CooperativeInteractions,Proc.2nd
Inter.Con£FuzzyTheory&Tech-
nology,pp.26-28(1993)
[8]K.Yamada,M.Mukaidono:Recog-
nizingIntentionsbyFuzzyAbduc-
tiveReasoning,Proc.FifthIFSA
WorldCongress,pp.187-190(1993)
JIQNo.101,1995 70
Neuro-FuzzyApPlicationfor
IntelligentSystem
ToshioFUKUDAandKojiSHIMOJIMA
DepartmentofMicroSystemEngineering
NagoyaUniversity
1.Introduction
Intelligentautonomousrobotsystems
arerequiredinmany6eldsandplaces.
Theintelligentrobotshavetocarry
outtasksinvariousenvironmentslike
humanbeings.Theyhavetodeter-
minetheirownactionsinuncertainor
changeableenvironmentsbasedon
sensoryinfbrmationandtheirknowl-
edge.Humanoperatorscangivethe
robotstheirknowledgeinadvance
concerningtasksandskillsinatop-
downmanner.However,whenthero-
botsperformtasksinan皿certainen-
vironment,theknowledgemaybe
meaningless.Inthiscase,therobots
havetoadapttheiractionstotheir
environmentsandacquirenewknowl-
edgeorskillsbythemselvesthrough
actions.Thisprocessproceedsina
bottom-upmanner.
Thispaperintroducesahierαrcんicαl
intelligentcontrolschemeforintelli-
gent.robots.Thehierarchicalintelli-
gentcontrolhasconsistsofthreelev-
els:theadaptationlevel,theskilllev-
el,andthelearninglevel.Thisscheme
hastwocharacteristicswithrespectto
thelearningProcess:toP-downap-
proachandbottom-upapproach.To
linkthethree】evelsandhavesuch
characteristicsforknowledgeacquisi-
tion,theschemeusesartificialintelli-
gence(AI),fuzzylog{c,neuralnetworks
(NN)andgeneticalgorithm(GA)[1-3].
Eachtechniquehasadvantagesand
disadvantages.Inordertoovercome
thedisadvantages,thispaperintroduc-
estheirintegrationandsynthesistech-
niques.Thosearekeytechniquesfor
intelligentcontrolofthesystemsinro-
boticsandmechatronics。
2.IntegrationandSynthesis
ofNeuralNetwork,Fuzzy
I.ogicandGeneticAlgo.
rithm
TheAI,fuzzy}ogicandneuralnetwork
havesimilarperfbrmancewithrespect
tosignaltransformation,thoughtheir
methodsaredifferent.Eachmethod
hassomemeritsanddemerits.Tablel
isthecomparisonofthem.Toover-
cometheirdemerits,someintegration
andsynthesistechniquesandGAhave
beenproposed.
71 JIQNo.lo1,1995
Thefuzzylogicandtheneuralnet.
workscanbeusedaspreprocessorsof
theAI.Theytransformnumericaldata
settothesymbolicdataset.Togive
therulesfortransformation,human
operatorseasilydeterminerulesofthe
fuzzylogic.However,whenthenum-
berofinputparametersincreases,de-
terminationoftherulesbecomeslabo-
riousforthehumanoperators.Inthis
case,theneuralnetworksareuseful.
Whileshowingdatasetsofinput/out-
puttotheneuralnetwork,itlearns
themandworksasatransformative
function.Drawbacksoftheneuralnet-
workarethatthehumanoperators
cannotgivetheirknowledgebefbre-
handnorunderstandtheacquired
rules.Moreover,theconvergenceof
thelearningisveryslowandtheneu-
ralnetworkcannotleannewpatterns
incrementa}ly.Tosolvetheseprob-
lems,thestructuredneuralnetworks
wereinvestigated.
T}1efuzzyneuralnetworkisacom.
binedneuralnetworkwiththefuzzy
logic.Humanoperatorsareableto
givetheirknowledgeinthefuzzyneu-
ralnetworkbymeansofmembership
functions.Themembershipfunctions
aremodifiedthroughthelearningpro-
cess.Afterthelearningprocess,the
humanoperatorscanunderstandthe
acquiredrulesinthenetwork.With
respecttotheconvergenceofthelearn-
ingprocess,thefuzzyneuralnetwork
isfasterthantheconventionalneural
network.Formultipleinputpara・me-
ters,thehierarchicalfuzzyneuralnet-
workisavailable.However,itisdif臼 -
culttooptimizethestructureofthe
hierarchicalfuzzyneuralnetwork.On
theotherhand,thefuzzylogicisused
asacriticfbrimprovementofconver.
genceoflearningoftheneuralnet-
work.Inthiscase,thefuzzylogicde-
terminesthelearningstepdepending
onthestateofconvergence.
Theneuralnetworkwithradialbasis
functionsisalsothestructuredone.It
haspotentialtolearnmorequicklyand
iseasierthantheneuralnetworkwith
thesigmoidfunctions.Forincremen-
tallearning,theAdaptiveResonance
Theory(ART)modelhasbeenproposed.
Ithasatwo-layeredstructure.Itlearns
patternsonebyoneincrementally.
Thatis,itcancorrecterrorsbylearn-
ingnewpatternswithoutusingtheold
patterns.However,theARTmodel
hasaproblemofbadclassif〕cationabiL
ity.
TheNeuralnetworkbasedonDistance
betweenPatterns(NDP)hastheabili-
tiesofincrementallearningandclassi-
fication.TheNDPlearnscategoriesof
patternsonebyone.Itincreasesneu-
ronsoftheoutputlayerusingthein-
crementallearningalgorithm.Ituses
theradialbasisfunctionattheoutput
layer.Dependingonitsaim,human
operatorsshouldgivetheneuralnet-
workef資cientstructureiftheyhave
experiences.Orelse,heuristicap-
proachforstructureoptimizationis
necessa「y・
TheGAisapowe㎡ultoolfbrstructure
optimizationofthefuzzylogicandthe
neuralnetworks.Particularly,theGA
isapowerfultoolforoptimizingthe
JIQNo.101,1995 72
hierarchi6alfuzzyneuralnetwork.On
theotherhand,thefuzzylogicandthe
neuralnetworkcanbeaevaluation
functionfbrtheGA.Itisdi笛cultto
defineevaluationfunctionsforcomplex
optimizationproblems.However,while
usingthefUzzylogicortheneuralnet-
work,humanoperatorscantransfer
theircriterion.Thosearethecompli-
catedreinforcelearningtechniquesbe-
causetheydonotuseteachingsignal
butobtaindesirablestateswhilema-
nipulatingalotofparametersatthe
sametime.TheGeneticProgramming
whichisoneoftheapplicationsofthe
GAandmanipulatessymbolscanpro-
.ducenewrulesorknowledgeforthe
AI.
3.HierarchicalIntelligent
Contro1
Thehierarchicalintelligentcontrol
schemecornprisesthreelevels:aleαrn-
inglevel,askilllevel,andanαdαptα -
tionlevelasshowninFigure1[4].The
learninglevelisbasedontheexpert
systemforareasoningmechanismand
hasahierarchicalstructure:recogni-
tionandplanningtodevelopcontrol
strategies.Therecognitionleveluses
neuralnetworksandfuzzyneuralnet-
worksasnodesofadecisiontree.In
thecaseoftheneuralnetwork,inputs
areanumericquantitysensedbysome
sensors,whileoutputsareasymbolic
qualitywhichindicatesprocessstates.
Thestructuredneuralnetworkforin-
crementallearningiseffectiveformem-
orizingnewpatterns.Thefuzzyneu-
ralnetworktransformsnumerical
quantityintosymbolicqualitybyusing
membershipfunctions.Boththeneu-
ralnetworkandthefuzzyneuralnet.
workaretrainedwiththetrainingdata
setsofa-prioriknowledgeobtained
fromhumanexperts.Asaresult,the
neuralnetworkandthefuzzyneural
networkcantransfbrmvarioussensed
datafromnumericalquantitiestosym-
bolicqualities,andperformsensorfu-
sionandproductionofmetα -hnoωledge
atthelearning}evel.Theimportant
informationissensedactivelybyusing
theknowledgebase.Thesensorsof
vision,weight,force,touch,acoustic,
andothercanbeusedasnodesofthe
decisiontreeforrecognitionoftheen-
Vlronment.
Then,theplanninglevelreasonssym-
bolicallyforstrategicplansorsched-
ulesofroboticmotion,suchastask,
path,trajectory,force,andotherplan-
ninginconjunctionwiththeknowl-
edgebase.Thesystemcaninclude
anothercommonsenseforroboticmo-
tion.TheGAoptimizescontrolstrate-
giesforroboticmotionheuristically.
TheGAalsooptimizesstructuresof
neura}networkandfuzzylogiccon-
nectingeachleveLThus,thelearning
levelreasonsunknownfactsfroma-
prioriknowledgeandsensoryinforma-
tion.Then,thelearninglevelproduc-
escontrolstrategiesfbrskillleveland
adaptationlevelinafeed-forwardman-
ner.Followingthecontrolstrategy,
thelearninglevelseldctsinitialdata
setfbraservocontrollerattheadapta-
tionlevelfromadatabasewhichmain-
tainssomegainsandinitialvaluesof
interconnectionweightsoftheneural
networkintheservocontroller.More-
73 JIQNo.101,1995
Table1 ComparisonofNeuralNetwork,FuzzyLogic,AI,
andGeneticAlgorithm
Ma血
Model
Leaming
Data
Operator
Knowledge
Real
Time
Knowledge
・Representanon
Nonlinea亘ty Oplim立ation
Control
Theory × × × ×Neural
Network × × ×Fuzzy × ∧ ×AI × × 〈
` ×GA × × ×
o
.⊆
∈
①」
三の
⊂o揖
△
▽<
Goa|
NeuralServo
Controller
KnowledgeBase
Signal
Transformation
HumanHeuristic
tnstructionLearning
≡E.aluati。n▲● ●
Ali
…
…
⊂i
§i芦i
…:
:
:
:
:
…
▼
Fuzzy
NN
ロ
⊃∈96co
コ
:GA=====:=::::=::::=:::=
FigurelHierarchicalIntelligentControlSystem
JIQNo.lo1,1995 74
over,therecentsensedinformation
fromthe'skilllevelandtheadaptation
levelupdatesthelearninglevelthrough
theIong-termlearningprocessbyhu-
maninstruction.Therefore,knowledge
atthelearning}evelisgivenbyhuman
operatorsinatoP-downmannerand
acquiredbyheuristicsoftheskilllevel
andtheadaptationlevelinabottom-
upmanne「 ・
Inthesametaskanddifferentenviron-
ments,itisnecessarytochangecontrol
referencesdependingontheenviron-
mentfortheservocontrolleratthe
adaptationlevel.Attheskilllevel,the
fuzzyneuralnetworkisusedforspecif二
ictasksfbUowingthecontrolstrategy
producedatthelearninglevelinorder
togenerateapPropriatecontrolrefer-
ences.Inputsignalsintothefiユzzyneu-
ralnetWOrkarenUmeriCalValUeS
sensedbysomespecif]csensorsand
somesymbolswhichindicatethecon.
trolstrategyproducedatthelearning
level.OutputofthefUzzyneuralnet-
workisthecontrolreferenceforthe
servocontrollerattheadaptationlev-
el.Thisoutputisbasedontheskill
extractedfromhumanexpertsthrough
learningandtrainingsetsobtained
fromthem.Atthesamemoment,the
fuzzyneuralnetworkclustersthein-
putsignalsintheshapeofmembership
fUnctions.Thesemembershipfunctions
areusedasthesymbolicinfbrmation
fOrthelearninglevel.
Intheadaptationlevel,aneuralnet-
workintheservocontrolleradjuststhe
controllawtocurrentstatusofdynam-
icprocess.Particularly,compensation
fornon-linearity.ofthesystemandun-
certaintiesincludedihtheenvironment
mustbedealtwithbytheneuralnet-
work.Thus,theneuralnetworkinthe
adaptationprocessworksmorerapidly
thanthatinthelearningprocess.Even-
tuaUy,theneuralnetworksandthe
fuzzyneuralnetworksconnecttheneu-
romorphiccontrolwiththesymbolic
controlfbrhierarchicalintelligentcon-
trolwhilecombininghumanskills.
Thehierarchicalintelligentcontrolis
alsoappliedtothemulti-agentrobot』
system.Ifthereisnotinteractionbe-
tweenrobots,eachrobothastowork
optimallyforitsownpurpose,sothat
thetotaltaskcanbeachievedoptimal-
ly.Thatis,eachrobotshouldwork
selfishly.Orelseconflictsamongthe
robotsmightoccurwhenusingapublic
source.Thecompetitionmaycausecol-
lisionsanddeadlockstatesamongthe
robotsinalocalares.Inordertoavoid
competition,itisnecessaryforthero-
botstocommunicateandtocoordinate
amongthemselves.Coordination
amongtherobotsisasimportantas
selfishness.TheGAsareappliedhier-
archicallytobalanceselfishnesswith
coordinationforefficientmotionplan-
ning.WhenmultiplerobotsWorkinde-
pendentlyasadecentralizedsystem,
thelearningcapabilityoftherobotsis
indispensablefortheevolutionofthe
system.
Astheresults,integrationandsynthe-
sisofAI,FuzzyLogic,NeuralNetwork
andGAareimportantfortheintelli-
gentsystem,dependingontheirchar-
acteristics.Hierarchicalintelligent
75 JIQNo.101,1995
controlusingthesetechniquesiseffec-
tiveforthecontrolinteUigentsystems
inroboticsandmechatronics.
4.Conclusions
andApplicationsfbrNeuralNet-
worksfbrIndustrialControlSys'
tems,IEEETrans.onIndustrial
Electronics,VoL39,No.6,pp.472-
489(1992)
Thispaperdescribedahierarchicalin-
telligentcontrolschemefbrintelligent
robots.Integrationandsynthesistech-
niquesofAI,fuzzyneuralnetworkand
GAmaketherobotsystemtobeinteUi-
gent.Thesystemhasbothtop-dOwn
andbottom-uplearningabilitieswhile
integratingandsynthesizingthose
techniques.
References
[1]T.FukudaandT.Shibata,Theory
[2]L.A.Zadeh,FuzzySets,Information
andControl,Vol.8.pp.228,(1965)
[3]D.E.Goldberg,GeneticAlgorithms
inSearch,Optimization,andMa-
chineLearning,AddisonWelsey
(1989)
[4]T.ShibataandT.Fukuda,SkillBased
ControlbyusingFuzzyNeuralNet-
workforHierarchicalIntelligent
Control,Proc.ofIJCNN'92-Balti-
more,Vol.2,pp.81-86(1992)
JIQNo.101,11995 76
ConsumerDecisionSupPort
SysteminVirtualSpaceUsingKansei
Engineering
JunjiNOMURA
VirtualRealityR&DGroup
InformationSystemCenter
MatsushitaElectricWorks,Ltd.
Abstract 1.Introduction
Virtualreality,anewparadigmforre-
lationshipbetweenhumansandcoM-
puters,hasbeenrecentlywell-known
andcurrentlyinvestigatedfbrpracti-
caluseinvariousindustrial行elds.
Usingthree-dimensionalcomputer
graphics',interactivedevices,andhigh-
resolutiondisplay,avirtualworldcan
berealizedinwhichonecanpickup
imaginaryobjectsasiftheywereina
physicalworld.Usingthistechnology,
MatsushitaElectricWorks,Ltd.has
beendevelopingseveralapPlication
systemsforindustrialusesince1990.
ThispaperdetailsVirtualSpaceDeci-
sionSupPortSystememployingKansei
Engineeringwhichisappliedforpro-
ductionandsalesmainlyinthesystem
kitchenbusiness.
Keywords
Virtualreality,DSS,Kanseiengineer-
ing.
Virtualreality,anewparadigmforre-
lationshipbetweenhumansandcom-
puters,hasbeenrecentlywell-known
andcurrentlyinvestigatedforpracti-
caluseinvariousindustrial行eldssuch
ascomputergraphics(CG),CAD,CAM,
CIM,robotics,medical/healthcare,
multi-media,gamesandsoon.High-
TechcompanieslocatedintheU.S.A.
andU.K.havereleasedseveralcom・ ・
mercialproductswhichcanbeutilized
fordevelopingVRapPlicationsystems.
ThiscausesthebasicVRsystemcon-
sistingofthree-dimensionalcomputer
graphicsengine,interactiveinterface
deViceswithmulti-functionsensorsand
high-resolutiondisplaystobegradual-
lydecreasingincostandsuchasystem
includingperipheraldevicestobeeasi-
lyavailable.
Since1990,MatsushitaElectricWorks,
Ltd.hasbeendevelopingseveralVR
apPlicationsystemsforindustrialuse.
Inthispaper,wedetailtheVirtual
SpaceDecisionSupPortSystem(VS-
77 JIQNo.101,1995
DSS)whichisutilizedforproduction
andsalesmainlyinthesystemkitchen
business[Nomuraetal.,1990].This
systememploysKanseiExpertSub-
system[Nagamachi,1986]whichac-
quiresconsumers'likingforsystem
kitchensbytranslatingtheirimages
onkitchensintotherealdesignfUrni-
ture.Thisenablestheconsumersto
haveaconcreteimageonthekitchens
whichwillbeavailable,andalsoen-
ablesthemtoseeandtouchthekitch-
encomponentsinthevirtualspace.
2.ConsumerDecisionSup・
portSystem(DSS)and
KanseiEngineering
Thechangingneedsandvaluesofto-
day'sconsumerhashadasignificant
impactontllesalesandmanufactur-
ingProcess.Acustomermustgetthe
necessarygoodsintherequiredquan-
titywhenneeded.However,whenthe
goodsdesiredbymanyconsumersare
diversified,themanufacturerhasadif-
ficulttimecopingwiththeincreased
workload.Changingonlytheproduc-
tionsystemcannotdealentirelywith
thesituation:thecorrespondingsales
systemincludingmarketing,distribu-
tionandinformationservicesmustalso
beimproved.
Computertechnologyisadvancingata
rapidrate.Thedevelopmentofatotal
productionsystemincorporatingCAD,
CAMandCAEispossiblenow.This
alsopermitsmovementfrommasspro-
ductiontotheproductionofavariety
ofgoodsinsmallquantities.Atpresent,
however,mostcomputer-aidedmanu-
facturingisgearedtowardsmasspro-
duction,unabletohandleone-of-a-kind
products.Thespecif]cationsofthese
productsshouldbeeasilychangedto
accommodateindividualcustomer's
needs.Torealizethisconcept,Kansei
EngineeringProductionsystemisnec-
essarywithaknowledge-basedexpert
SyStematitSCOre.USingvirtUalreali-
ty(VR)[Rheingold,1991]technology,
thevirtualspacedecisionsupPortsys-
tem(VSDSS)[Imamuraetal.,1991]
letsusersdesignvirtualproductsand
experiencethemwhileinvirtualspace.
VRisdef]nedbythreeelements田DTV,
1990]:three-dimensionalcomputer
graphicstechnology,interactiveinter-
facedeviceswithmulti-functionsen-
sors,andhigh-resolutiondisplays.
Studiesofthiscomputerenvironment
arebeingperformedby・NASAAmes
ResearchCenter[Fisheretal.,・1986a],
theMITMediaLab[Sturmanetal.,
1989],theHITLab[Jacobson,1991],
theUniversityofNorthCarolina[Airey
etaL,1990],theUniversityofTokyo
[Hirose,1989],theUniversityofTsuku-
ba[Iwata,1990],andtheATRCom-
municationSystemInstitute[Kishino,
1990].
2.1KanseiEngineering
KanseiEngineeringisdefinedas``a
translationsystemofaconsumer,sim-
ageorfeelingintotherealdesigricom-
ponents"[Nagamachi,1986].Namely,whenacustomerexpresseshis/herim-
agetowardanobjectusingadjective,
detaildesignitems(forinstance,ob-
jectstyle,color,material,sizeandsoon)areselectedthrOughtheKansei
JIQNo.101,1995 78
Engineeringprocedure,andusingthe
outputs,adesignerorplannercande-
signtheobject.TheKanseiEngineer-
ingProcedureis;
Steρ1.Collectingtheadjectivewords:
Steρ2.
Steρ3.
Collectmanyadjectivewords
whichhaverelationtotheob-
jectdomain.
Assessingslidesorpictureson
SDscales:
Makepairsoftheseadjectives
inagood-badfashionforthe
SD(orgood'sSemanticDiffer-
entials)scales.Nextassess
manyslidesorpictUresrelat-
edtotheobjectdomainon
theseSDscales.
Elicitingeffectiveadjectives
fromco}lectedadjectives:
Calculatetheassesseddataat
Step2byfactoranalysisor
principalcomponentanalysis,andobtainthesemanticfacto-
rialstructureofadjectiveson
therelateddesigndomain.
Andelicitadjectiveswhich
haveacloserelationtotheob-
jectdomainfromthecollected
adjectives.
Step4.Subdividingtheobjectdesign
intothedesigncomponents:
Subdividetheobjectdesignon
theslidesintodetaildesign
components,nextclassifyeach
componentlntocategoryac-
cordingtoit'squality.Forex-
ample,L-style(category)1ay-
out(component),mahogany
(category)cabinet(compo-
nent).
Step5.Gettingtherelevancybetween
anadjectiveandqualitative
data:
AtStep4,analyzebyHayashi's
QuantificationtheoryType1,whichisakindofmultivari-
ateregressionanalysisdeal-
ingwithqualitativedata[Ha-
yashietal.,1976].There-sultsofthisanalysismeans
relevancybetweenanadjective
andeachdesigncomponent.
2.2ProductionandSales
UnificationSystem
(PSrJS)usingtheVirtual
SpaceDSS(VSDSS)
Productionandsalesactivitiesstart
withtheplanningofsalesandstock
production.Afterevaluatingthere-
sultsofmanufacturing,assembly,and
salesoftheseproducts,businessplan-
ningisperfbrmedagain.Theseare
cyclicandsystematicactivities.Con-
versely,theconsumermustmakevar-
iousdecisionsontheproducthewish-
estopurchase.Thetypeandtimeof
deliveryhasasignifficantimpacton
whethertheuserwillpurchasethe
goods:
(1)Theproductisboughtwhenitcanbe
obtainedimrnediately.
79 JIQNo.101,1995
(2)Theproductisboughtonlywhenit
canbeobtainedbyaspecifieddate.
(3)Theproductisboughtregardlessof
itsdateofdelivery.
Acanofsodainavendingmachine
isanexampleof(1).Thesodawhichis
boughtinacooperativestorebyjoint
purchaseisanexampleof(2).Custom-
orderedsodathatisdesperatelyneed-
edisanexampleof(3).Planningon
theassumptionof(3)issufficientwith
simpleinformationprocessing.But
planningontheassumptionof(1)re-
quirescomplexinformationprocessing.
Incaseof(1),itisnecessarytodecide
productionplanningonthebasisof
salesplanninginformationwhichis
uncertain.Moreover,thequalityofthe
planningaccuracyhasaneffectonthe
businessaccounts.Inthiscase,itis
veryimportanthowtodraftthepro-
duction,salesandstockplanningfrom
informationthatisasuncertainafter
processingasbefbre.Forsalesoftype
(1),itisnecessarytodevelopanexpert
systemonproduction,salesandstock
planning[Nomura,1990]withasimu-
lationfunctionandoptimizationfunc-
tionforthevariousdemands.
The(2)patternisaspecialcase.Itis
veryrarethatproductsmanufactured
accotding・to、astandardspecification
aredeliveredontheappointeddateof
delivery.Itiscommonthatsuchprod-
uctsaremanufacturedaccordingtothe
customer'sspecifications.Sincethese
specificationsdifferamongeverycus・
tomer,itisnecessarytodevelopade-
signsupportandperfbrmanceestima-
tionsystemfbrthecustomertogether
withthedevelopmentofaCIMproduc-
tionlineandastandardpartorder/
stocksystem.Inthecaseofstandard-
izedparts,itispossibletodecideto
purchaseafterlookingandtouchingthe
partsattheshowroom.Butwhena
productistobedesignedbyacustom-
er,itisveryimportantfbrhim/herto
beabletoestimateorexperienceit.
Currentdesignsystemsaretailoredfor
thespecialist,andarenotwellsuited
forintentionsoftheordinaryconsum-
er.Theref()re,itisquitepossiblethat
theCuStOmerCannOtimaginetheCOm-
pletedstateofthesystem.Forexam-
ple,manyproblemswithconstructed
housesfrequentlyoccursincethebuy-
ercannotproperlyimaginethecom-
pletedhome.Howwelhstheroomlit
afterchangingthelightingequipment,
orhowisthecarnoiseattenuatedafter
insertingsound-proofingmaterialinto
thewall?Althoughtheseproductscan
beseenintheshowroomorreportscan
state"10decibelsloweraccordingsim-
ulationresults,"itisdifficulttoactual-
lyfeelwhatthismeans.VSDSSisa
decisionsupPortsystemthatensures
theperformanceestimationandthe
suspectedexperienceareinfactwhat
theuserwillexperience.ByusingVS-
DSStoproduceproductsmostsuitable
fortheindividualcustomer,construc-
tionofanewproductionandsalesuni-
ficationsystem(PS/US)becomespossi-
ble.
JIQNo.101,1995 80
2.2.1Limitationsofdesktop
showroom
Theshowroomisthecontactpointbe-
tweenthecustomerandthemanufac-
turer.Itmustcontainelementsofdis-
play,consultation,advisementandso
on.Avastdisplayspaceandmany
salespeoplearerequiredtodisplayour
variousproductsatMatsushita.Since
thenumberofshowroomsarelimited
duetoexpense,thecontactpointsbe-
tweenthemanufacturerandthecon-
sumerarereduced.Moreover,itis
difficulttodisplayour30,000different
productsincombinationwithstandard
partsaswellaslarge-scaleproducts
suchasoursystemkitchen.
Thedesktopshowroomcanbeapplied
withVR,telerobotics,andmulti-me-
diatechnology,andprovidesadisplay
withoutdisplay-spaceorgeographic
limitations.Thedesktopshowroom
wouldeliminatespaceandgeographic
limitationsbyimmersingthecustom-
erinavirtualshowroom,wherehecan
seeactualproductimagesorhearreal
salespeoplespeak.Forexample,the
productsdisplayedinShinjuku,Tokyocanbeseenbytcustomerusingthe
desktopshowroomatTakamatsu,Ka-
gawaPrefectureasifhe/shewereca-suallywalkingthroughtheremote
showroom.Tomaintainthesesitua-
tions,VR,telerobotics,andmulti-me-
diatechnologymustbeapplied.Thus,
thedisplayspaceandexpensecanbe
kepttoaminimum.
2.2.2CurrentPS八JSforSys・
temKitchen
Oneofourstrongproductslinesisthe"SystemKitchen":Acustomplanned
andbuiltkitchenusingover30,000
kitchenproductsandaninfinitenum-
berofpossiblekitchenlayouts.Since
thecustomermustmakemanydetailed
anddifficultdecisionswhenselecting
hisAiernewkitchen,itwasnaturalto
applyVSDSSandKanseiEngineering
technologytothissalesprocess.
Theoverviewofthe"systemkitchen"
productionandsalesunificationsys-
tem[Nomura,1990]isshowninFigure
1.Thekitchenplanningprocessisde-
pictedinFigure2.Wedevelopeda
prototypeVSDSSsystemcalledViVA
(VirtualRealityforVividA&ispace
system)whichallowsourcustomersto
pseudo-experiencetheircustomkitch-
enbeforepurchasingit.Whenanin-
terestedcustomercomestotheshow-
room,thekitchenplannerfirstexplains
thekitchenproducts'descriptionsus-
ingcataloguesandexhibits.
Thekitchenplannernextdrawsarough
layoutaccordingtothedesiresofthe
customer.Thecustomercanexperi-
enceasimilarpre-existingkitchenus-
ingtheViVAsystemifhe/shewishes.
Thenafloorplan,anelevationview,a
perspectivedrawing,andawrittenes-
timatearecreatedonaCADsystem
basedontheroughsketch.Thecus-
tomer'sownkitchenplancanbetrans-
】atedintoaViVAdatabasewithina
week.Thenexttimethecustomer
comestotheshowroom,he/shecanex-
periencemanyaspectsofhis/herown
kitchen.Thecustomercancheckhis/
herownkitchen.anddecideifitmatch-
81 JIQNo.101,1995
Consumer
Showrooms
ComputerAided
DesignSystem
十
Pseudo-Experienc
SystemtOcheck
LivingAmenity
ManufacturingSystem
ComputerAided
ProductionPlanningSyste
lnventoryControlSystem
f()rPartsandProducts
FigurelConsumer・orientedIntegratedManufacturingSystem
productsaredescribedusing
cataloguesandexhibits
シroughsketchofki【chenlayout
← stagel
(ViVA)
drawi皿gmadebyMATISsystem
(floorplan,elevationviews,Perspectiveviews)
・・㏄・・?ぐ コ
Finaldraw㎞gsmade
stage2
(ViVA)
1αd・rt・th・CIMIi・es
Figure2SystemKitchenPlanningProcess
JIQNo.lo1,1995 82
eshis/herownideahowthekitchen
shouldbe.Oncethecustomerapproves
his/herkitchendesign,finalapproval
andappliancedrawingsaremadeand
theorderissenttotheCIMline.
Inacurrentplanningsystemwithout
theViVApseudo-experiencesystem,the
customercanonlyseefloorandeleva-
tionplanswithoutgettinga``feelinピ'
f()rthekitchen.Sometimesthereare
manydiscrepaneiesbetweenthesys-
temkitchenaetuallymanufacturedand
thecustomer'soriginalidea.TheViVA
systemcaneliminatethesemistakes
thatareoftenmade. 、WithViVA,the
followingitemscanbeexperiencedin
thevirtualkitchen:
1)Thearrangementofcabinetsandap-
pliances.
2)Thegeneralfeelingofavailable
space・
3)Overallergonomicdesign:Theuser
canopenandclosecabinetdoors,
turnonfaucets,movegoodsinthe
pantry,etc.
Originally,systemkitchenplanning
wasdoneonaSun-basedCADsystem
calledMATIS(MatsushitaAmenity
TotalInteriorSystem).TheMATIS
databaseincludesapProximately
30,0000fmatsushita,skitchenprod-
uctsaswellasdataonpreviousand
eurrentcustomers.Thiscustomerdata
includefieldsforroomdimensions,cab-
inetplacement,standardizedpartsin-
f()rmation,specialorderinformation,
etc.
AfterdrawingtheplanontheMATIS
system,thetwo-dimensionalpictureis
firsttranslatedtoathree-dimensional
layouttobeexperiencedintheViVA
system.Aninteractiveconversionpro-
gramcalledStarchrunningonaSili-
conGraphicsIriscanbeusedtocon-
vertthismonochrome,wireframedata
totheViVAfbrmat.Ifthedesired
producthasnotyetbeenconvertedto
theViVAformat(remember,thereare
over30,000possibleproducts)oritisa
special-orderedproduct,Starchisused
toedittheproductandaddcoloror
specialconstraints.
Oncetheproductshavebeentranslat-
edtothree-dimensions,theWringer
programisused.Thisprogramcom-
binestheproductsconvertedbyStarch
andtheroomdimensionsdescribedby
theMATISfile.Aseachwalliscreat-
ed,theproductsareplaceduponitone
byone.TheoutputofWringerisa
SwivelfileandanIsaacfileusedbya
VPLRB2system.TheIsaacdatafileis
usedtorenderthevirtualkitchenona
SiliconGraphicsIris,andtheSwivel
fileistransferredtoaMacintosh,where
itisreadbyaVPLprogramcalled
BodyElectric.BodyElectricmanages
thebehaviorofthevirtualworldby
controUingtherendererandlinkingin
realworlddatasuchasheadandhand
positions.Thecustomercannowexpe-
riencetheirownvirtualkitchen.
2.2.3FuturePS八JShaving
KanSei・infOrmatiOninter・
pretationcapabilities
AsthenextversionofViVAsystem,
83 JIQNo.101,1995
KanseiViVAsystemisbeingdeveloped.
Thissystemisvalidtothefollowing
customers,types:
1.Thecustomerswhohavenoideaof
thekitchen
2.Thecustomerswhoareinconfusion
becausetheysawmanycatalogues
andexhibits
3.Thecustomerswhocannotimagine
whatthesizeoftheirkitchenspace
is(usually,kitchenlookssmaller
thanactualsizeinshowroombe.
causetheheightoftheshowroomis
higherthanhousingone).
Usingthissystem,vaguenessandcon-
fusionofcustomer'simagearecleared.
Soweexpectthatthissystemcande-
creasethenumberoftheconsultation
betweenaplannerandacustomer(usu-
allyitrepeat40r6times,sometimes
over10times),anddecreasethecon.
sultationtime(usuaUyittakes2hours,
sometimesover4hours).First,cus-
tomerinputsthefieldforroomdimen-
sionandheightofcustomerwhouse
kitchenasrestrictionconditions.Next,
he/sheinputslifestyleofhis/herfamily
andhis/herimagetowardthekitchen
ina(ljectivewordsasKanseiconditions.
ThentheKanseiViVAsystemidenti-
fiesthekitchenplanindetail(forin-
stance,kitchenlayout,cabinetcolor,
floorcolor,counterheightandsoon)
usingKanseiEngineering.Wegath-
eredover200adjectivesand18items
offeaturedlifestyles.Thenknowledge
basedexpertsystemsearchessome
kitchenplansthatmatchhis/herpref-
erenceandroomcondition,andedits
thekitchenplan'scabinetarrangement
tofithis/herroomdimension,and
changethekitchenplan,scabinetco1-
or.Throughexperience,thecustomer
canchanget}}ewallsize,cabinetar-
rangement,cabinetcolorandsoonof
thekitchenplan、thatisproposedby
computerintohisownkitchenplan.
Figure3showstherelationbetween
Kitchendesignandadjectivewords.
2.3Futureperspectivesto・
wardanadvancedsystem
forthetotalhouse
CurrentViVAsystemandKanseiViVA
systemaredealingwithkitchenspace.
Inthefuture,wewishtodevelopto
modelanentirehouse.Figure4shows
suchasystem.Thisdevelopmentis
joinedwiththeprojectwhichisa7-
yearplansince1989called"Technolo-
gyDevelopmentProjectforNewIndus-trializedHouses"undertheMinistryof
InternationalTradeandIndustry.The
aimofthisprojectistodevelopasys-
temwhichachievesnewhousingfbr
thecoming21stcentury.Fortheim-
plementationoftheproject,researchanddevelopmentisbeingproceededby"TheNewIndustrializedHousePro -
ductionTechnologyandSystemDevel-
opmentTechnologyResearchAssocia-
tion(WISH21)."MEWtakeschargeof
thedevelopmentof"residentparticipa-
tion"amenitysimulationsysteminthis
project.Usingthissystem,residentcanexperienceandevaluatehousing
perfbrmancesuchaslight,sound,vi-
bration,temperature,airandsoon.
JIQNo.101,1995 84
Warm.SΩft
classic
撫遡亘メ
てく
芯1・},亘 頂濱 †
1D
ark,Digtiil"ieci-一 一""一ーー ー ーー 一一ー ー ー ー一
念
.菱,乏
s】
存 1wwr
漕
iiatural
in,◎dern
蒙P.
〃 ン
《嘗 認'ぴ1膓
、1∴㍉`{選
一・薦1
cle謬ant
high-tech
蟹ぷ`簾 諫
、
与繍
、千
T
L-ight
・CaSUal
㍉ ㍉ 湾{∵.'
Co◎1,Hard
Figure3 TheRelationbetweenKitchenDesignalld
AdjectiveWords
onmulatin
needs
Design
support
Livingamenitys㎞ulatuon
●Space●Light●Sound
●TemperatUre●Air
●Safety●Performance
ost
estunatlon
Evaluationof
livingamenity
Figure4AHousePlanningProcess
85 JIQNo.101,1995
3.Conclusions
Inthispaper,VRapPlicationsystems
forindustrialusewhichhavebeende-
velopedinMatsushitaElectricWorks,
Ltd.isdescribed.VRtechnologyhas
recentlybeeninvestigatedforindus-
trialusesincethebasiccomponentsfor
developingtheVRapplicationsystem
aredecreasingincost.Ontheother
hand,theresearchanddevelopment
concerninghigh-costVRsystemswhich
providetheextremelyhigh.resolution
3-Dgraphicsimagesandrealizehigh
perfbrmanceistobecoptinued.Thus,
itcanbeconsideredthatthedevelop-
mentofVRapplicationsystemsdivides
intotwodirections:10w-costVRsystem
andhigh-costVRsystem.Considering
thispointofview,wearegoingtopro-
ducealow-costVRsystemforpractical
usewhiledevelopingahigh-costVR
system.forresearchinthefuture.
StudyofCustomConsultationSystem
inTermsofKnowledgeEngineering,"
HumαnInterfbce,Vol.3,(1986).
[Nomuraetal.,1990]Nomura,J.,Oha-
ta,H.,Imamura,K.andSchultz,J.R,"VirtualSpaceDecisionSupPortSys -
temandItsApPlicationtoConsumer
Showrooms,"VisualComputing,
Spring-VerlagTokyo,183/196(1990).
[Hayashietal.,1976]Hayashi,C.and
Komazawa,T.,"AStatisticalMethod
forQuantificationofCategoricalData
anditsApplicationtoMedicalScience,"
InDecisionMakingandMedicalCare,
(edF.T.deDombalandF.Gremy)
North-HollandPublishingCompany
(1976)。
[Hirose,1989]Hirose,T."Creationof
ArtificialReality;"System/Control/In-
formation,Vol.33,No.11(1989)(in
Japanese).
Acknowledgment
Theauthorofthispaperwouldliketo
thankMr.AkiraOhishi,Directorof
InformationSystemCenter,Matsushi-
taElectricWorks,Ltd.,forhissupport
andvaluablesuggestions.
References
[Aireyetal.,1990]Airey,J.,Rohlf,J.
andBrooks,F."TowardsImageReal-
ismwithInteractiveUpdateRatesin
ComplexVirtualBuildingEnviron-
ments,"UniversityofNorthCarolina
atChapelHillTR90-001,(1990).
[Nagamachi,1986]Nagamachi,M.,"A
[HDTV,1990」HDTV&TheQuestfor
VirtualRealityACMSIGGRAPH,
(1990).
[lwata,1990]Iwata,H.,"ArtificialRe-
a】itytoWalkinTheLarge-scaleVirtu-
alSpace,"HumαnInterfaceNeωsαnd
Report,Vol.5,No.1,49/52(1990)(in
Japanese).
[Jacobson,1991]Jacobson,R."Televir-
tuality:"BeingThere"inthe21stCen-
tury,"Hamαn(1991).
[Kishino,1990]Kishino,F."Feelin
Communication,"Three-1)imensional
lmαge,Vol.4,No.2,(1990)(inJapa-
JIQNo.101,1995 86
nese).
[Rheingold,1991]Rheingold,H."Vir-
tualReality."SummitBooks,NewYork,
Tokyo(1991).
[Sturmanetal.,1989]Sturman,D.,
Zeltzer,D.andPieper,S."Hands-On
InteractionwithVirtualEnviron-
、ments,"AGMSIGGRAPH/SIGCH
(1989).
87 JIQNo.101,1995
InterSpace:TowardsNetworked
Reality
GenSUZUKI
NTTHumanInterfaceLaboratories
1.Cyberspace
Asharedandinteractivevirtualspace
onanetworkiscalled"Cyberspace"
[1].Asharedandinteractivevirtual
spacef()rmultipleusersonanetwork
isexpectedasapowerfultoolforhu-
mancornmunication.Asharedvirtual
spaceonthenetworkcanbeappliedto
createvarioususefultelecommun{ca-
tionservicessuchasvirtualoffrice,vir-
tualshop,virtualcampus,andsoon.
Habitat[2],SIMNET[3]andDIVE[4]
arethoughttobetheinitialstepsto-
wardsafutureidealCyberspace.These
virtualspaceservicesarebasedonin-
f()rmationgeneratedbycomputers.All
thevisualimagesofthesesystemare
madethroughcomputergraphicsor
animation,andareimaginary.The
virtualworldsofthesesystemshave
norelationtotherealworld.
Ontheother 、hand,theusualcommu-
nicationtools,suchastelephoneor
videophone,arebasedonrealinforma-
tion,thatis,realvoiceorrealvideo
imagesofrealpeople'sfaces.Inorder
tosupporthumancollaborationinbusi-
nessfields,asharedvirtualspacebased
onsuchrealinf()rmationapPearsnec-
essary.Fromthispointofview,we
proposethenewconceptofavisualcommunicationenvironmentforhuman
collaborationusingavirtualspacecre-
atedby3Dcomputergraphicsandvid-
eotexturemapPingtechnologies[5].
Theproposedsystemconsistsofacom-
binationofanimaginarylocationand
realvideoimages.
Inthispresentation,humaninterface
designissuesforaproposedsharedvir-
tualspacearediscussedandevalua-
tionoftheprototypesystemisde-
scribed.
2.ConceptofInterSpace
Weproposeanewconceptofashared
virtualspace,namedas"lnterSpace,"
thatisbasedonthefollowingneces-
saryfunctions[5].
(1)Seamlessnessinplannedandun-
plannedcommunication
HumancommunicationactiVitiesare
classifiedintoplannedcommunica-
tionandunplannedcommunication.
JIQNo.101,1995 88
Inregularbusinessmeetings,date,
place,purposeandmembersare
.agreedandplannedbeforethemeet-
ing.Ontheotherhand,unplanned
andcasualmeetingsoccurunexpect-
edlyinaelevator,onaroad,orata
passageway.Insuchcasualsitua-
tion,peoplerecalltopicsorthinkof
businessafterseeingthepartner's
faceunexpectedly.Bothplanned
and皿plannedcommunicationplay
importantrolesinoursocialcom-
municationactivities.Asharedvir-
tualspaceisexpectedasanewcom-
municationtoolthatisabletoen.
hanceunplannedcommunication
びコopportunities.
(2)Video-basedvirtualspace
3.HumanInte㎡aceDesign
Issues
3.1Realityofsharedvirtual
space
Concerningrealityofvirtualcommu-
nicationenvironment,therearethefbl-
lowingtwokindsofpoliciestodesigna
sharedvirtualspace.
(1)Creationofrealisticsensation
Thispolicyistheapproachtoasim-
ulatorofphysicalrealspaces.Virtu-
alspacesarevisualizedtoresemble
realmeetingspaces.Visualimages
ofvirtualspacearedesignedinor-
dertorealizeafeelingofbeingthere.
Thevisualcommunicationenviron.
mentconsistsofhumanobjectsthat
expresspeopleinthespace,andtoP-
icalobjectsthattransferinf()rma-
tioninthevirtualspace.Byintro-
ducing3Dperspectiveexpressions,
3Dcomputergraphicsisverysuit-
abletocreateanunderstandable
andimaginativevirtualspace.In
ordertorepresenthumanobjects
andtopicalobjectsinthevirtual
space,videoimagesaremoreusefu】
andmoreexpressivethancomputer
animation.Asthevideoimagetak-
enbyacamerarepresentsasubset
oftheinformationoftherealworld,
peoplecanrecognizetheactualsitu-ationofthepartnereasily.So,the
combinationofaCG-basedvirtual
spacestructurewithvideo-based
objectsishighlysuitableasthevir-
tualspacearchitecture.
(2)Creationofstrengthenedcoexistence
Thispolicyistheapproachtoartifi.
cialandresponsiveenvironmentfor
interactivecommunication.Visual
imagesandaudioresponsesofvirtu-
alspacearedesignedinorderto
createartificialeffectivefunctions
thatsurpassbeingthere.
3.2Visualandaudiorepresen・
tationforvirtualspace
Inrealphysicalspace,visualandau-
dioinformationareobeyedbyfixed
physicallaWs.However,invirtual
space,lawsandrulesofvisualand
audioinformationcanbedesignedfree-
ly.Characteristicsofhumanbehav-
iorsinasharedvirtualspacewillbe
stronglyaffectedbysuchartificiallaws.
Onecangethighlyintelligentandsen-
89 JIQNo.101,1995
sitiveearsoreyesinavirtualspacein
ordertoenhancethepossibilityofun-
plannedencounter.
4.PrototypeSystemand
Evaluation
Inordertoevaluatetheproposedcon-
cept,aprototype"InterSpace"system
wasimplementedusingdistributedper-
sonalcomputersandasystemserver・
Visualimagesofeveryterminalare
createdbythepersonalcomputersof
eachterminal.ThePCgeneratesan
interactivevirtualspacestructureus-
ing3-dimensionalgraphics.Videoim-
agesarecapturedbyacamera,and
videoimagesarepastedontoCGmod-
elsusingtexturemapPing.Audiosig-
nalsofindividualuser,svoicesaregath-
ered,mixedanddelivereddepending
oneachlocationinavirtualspace.The
systemservercontrolsdata,audioand
videOcommunicationbetweentheper-
sonalcomputers.
Theprototypesystemhasbeenevalu-
atedinourlaboratory.Humanbehav-
iorsinavirtualspacewereaffectedby
conditionsandlawsofthevirtualspace.
Walkingthroughasharedvirtualspace
offersnewabilitiesandenjoymenttO
manyuse「s・
5.Summary
cationenvironmenthasbeenproposed.
Theproposedsystemisasharedand
interactivemulti-uservirtualspace
thatconsistsofaCG-basedvirtual
spacestructureandvideo-basedob←
jects.HumanInterfaceDesignIssues
arediscussedfromtheviewpointof
thecreationofanewrealityforthe
enhancedcommunicationenvironment.
References
[1]M.Benedikt:"Cybersp'ace:First
Steps,"MITPress(1992)
[2]C.MorningstarandRR.Farmer:"TheLessonsofLucasfilm'sHabi-
tat,"ibid.pp.(1992)
[3]E.A.Alluisi:"TheDevelopmentof
TechnologyforCollectiveTraining:
SIMNET,aCaseHistory,"HUMAN
FACTORS,Vol.33,No.3,pp.343-
362(1991)
[4]L.E.Fahlen,C.G.Brown,'O.Stahl
andC.Carlsson:``ASpaceBased
ModelforUserlnteractioninShared
SyntheticEnvironments,"INTER-
CHI'93ConferenceProceedingspp.
43-48(1993)
[5]G.Suzuki,S.SugawaraandMMo-
riuchi:`へ7isualCommunicationEn-
VironmentUsingVirtualSpaceTech-
nology,"ICAT'93,(1993)
Theconceptofanewvisualcommuni.
.
JIQNo.101,1995 90
CurrentNews
*SeikoStartsFMRadio
PagingServiceinChina.
HattoriSeikoandSeikoEpsonwillstart
aninf()rmatiOntranSmiSSiOnSerViCe
targetingnonbusinessusersforwrist-
watchterminalsusingFMbroadcast-
ing.Informationsentto"SeikoRecep-
tor"terminalswillincludeweatherfore-
castandstockquotationsbroadcastto
wristwatcheswithadigitaldisplay
function.AnFMfrequencyrangenot
forbroadcastingwillbeused.This
typeofservicewasstartedintheU.S.
in1992andisnowavailableinthree
cities,includingLosAngelesandSeat-
tle.
AtrialdemonstrationstartedinSep-
tember1994inJapan,intheKyoto
area.Sincebroadcastingandcommu-
nicationscomeunderseparatelegal
systems,theMinistryofPostsandTele-
communicationsisstudyingalegalre-
visionandconcernedenterprisesare
awaitingtheconclusion.Forthisrea-
son,thetwoSeikoenterprisesselected
China,'whereregulationsarerelative-
lyrelaxed.Wristwatchesusedasin-
f()rmationterminalswillbemanufac-
turedandmarketedbyajointventure
betweenSeikoandaChineseenter-
ぶ
pnse・
Newtransmissionandrelayfacilities
neednotbebuiltfbrstartingtheFM
servicebecauseexistingbroadcasting
stationfacilitiescanbeused,which
meanslowerequipmentinvestments
thanpreviously.
Thedemandfbrradioinfbrmationser-
vicesisexpandingrapidlyinChinabe-
causetelephonelinescoverageisstill
limited.Morethan10millionpagers
areusedinChina,No.2intheworld
followingtheU.S.'Motorolahasbegun
producingmobiletelephonesinChina,
andCasio,NEC,andMatsushitaCom-
municationIndustrialarealreadyman-
ufacturingpagersandmobiletele-
phonesthere.TheSeikogroupplans
toexpandthesalesofwristwatchin-
formationterminalsthroughthisnew
servlce.
*FujiBankWillStart
TransmittingInformationviaInternet.
InFebruary1995,FujiBankwillbegin
providinginformationsuchasthe
bank'sbusinessindicatorsandfinan-
cialcommoditiesandservicesviaIn-
ternet;theservicewillcoverover10
itemsincludingfinancialandbusiness
infbrmation,e.g.,theanalysisofsettle-
mentsforinvestors,domesticandover-
seaso笛cesforclients,specificcom-
moditiesandservices,socialcontribu-
tionactivities,in-houseclubactivities,
9roupenterprises,andmoneycondi-
tionsinJapan.InformationwiUbe
displayedinbothJapaneseandEn-
91 JIQNo.101,1995
glish.FujiBankstartedtheserviceto
meetneedsinanageofinternational,
liberalizedmoneymarkets.Theirpur-
poseistoimprovetheircorporateim-
age 、andincreasebusinessbysupply・
inginformationinstantaneouslytodo-
mesticandoverseasinvestorsandus-
ers.
InOctober1994,FujiBankconnected
personalcomputersatitsheadquar-terstoInternetthroughanInternet
connectioncompany.Theywilldevel-
opsoftwareandconstructthesystem
requiredforsendinginformationinthe
nearfuture.
TheSanwaBankhasalsoconnected
itspersonalcomputerstoInternetand
planstobegininthespringof1995sendinginformationonthesettlement
ofaccounts,business,employment,etc.,
asapartofitsinformationservices
(IR)forinvestors.TheSumitomo,Saku-ra,DaiichiKangyo,andMitsubishi
Banksarealsostudyingsimilarservic-
es.
ResearchorganizationsinJapansuch
asuniversitiesandthinktanksand
enterprisessuchasinfbrmationand
telecommunicationequipmentmanu-
facturershavequicklystartedIRac-
tivitiesusingInternet,withabout100
enterprisessaidtobetransmittingin-
fbrmation.Anexperimentofinvolving
infbrmationtransmissionfromtheof-
ficialresidenceofthePrimeMinister
hasbeenstartedamongnonbusiness
お コorganlzatlons.・
*NECandC&WTieUpin
Telecommunications
NEChasconcludedacontractfora
comprehensivetieupwithCable&
Wireless(C&W),aBritishcommoncar-
rier.Thetwocompanieswillbegin
withacomprehensivecorporaterela-
tionshipandexchangeinformationand
marketstrategies,thenproceedtocon-
tractsinspecificbusinessareas.Be-
causeC&Wisinfluentialinwideareas
offormerBritishterritory,information
fromC&Wandtheirlong-termstrate-
gieswillbeusefulforNEC'smarketingactivities.C&Wwill,atthesametime,
greatlyincreasetheeff]ciencyofequip-mentdevelopmentandselectionwork.
Thetwocompanieshavestarteddivi-
sion.leveldiscussionsonasatellitecom-
municationsystem.Theywillsubse-
quentlyproceedtoareasoftelephone
switchingsystemsandmobiletele-
phones.CombiningC&W'scommuni-
cationservicesknow-howandNEC's
switchingsystemandsatelliteground
stationtechnology,theywillexpand
businessinChina,whereagigantic
demandfortelephonenetworksand
radiocommunicationsisanticipated.
C&Whasthegreatestbusiness
strengthinAsiaamongmajorEurope-
anandAmericancommoncarriers.
Plansaretopromotemarketexpan-
sionbytyingupwithNEC,thelargest
comm皿icationequipmentmanufactur-
erinAsiaandwhichhaseommunica-
tionequipmentplantsinChinaand
Malaysia.Bytyingupwiththisglobal
JIQNo.101,1995 92
commoncarrier,NECcangainanad-
vantageincompetingwithAT&T,Al-
catel,Siemens,andotherAmericanand
Europeancommunicationequipment
mamlfacturers,whicharebecoming
increasinglyaggressiveinAsia.
SinceAsiancountriesarebehindEu-
ropeandtheU.S.insettinguptele-
communicationinfrastructures,the
potentialrequirementsaregreat.Chi-nahasnotopenedcommunicationser-
vicestofbreignfunds,butplanstoin-
creasethenumberoftelephonelinesto
over100million,morethantwiceas
thenumberunddrcurrentcircumstanc-
es.Foreigninvestmentswillthusbe
essentialforaccomplishingthisobjec-
tive.
*NTTWillEstablisha
VirtualCompanyUsing
YoungEmployees.
NTThasdecidedtoestab】ishavirtual
companyconsistingofabout1,000
youngemployeesbyApril1995.Those
whohaveworkedforN'1"Tforlessthan
tenyearswillbecandidates.Theprin-
ciplebehindthisdecisionistoprornote
theexchangeofinformationamong
youngemployeeswhosemindsareflex-
ibleandwhopossessoriginalideas,
andtoutilizetheirthinkingfbrpro-
motingmultimediabusiness.NTThas
about25,000employeeswhomeetthe
under-10-yearrequirement.Thesepeo-
plewillbeinvitedtomakeapplica-
tions.IntheeventoftoomanyapPlica-
tions,thecompanywillselectemploy-
eesbyafairproceduresuchasthe
drawingoflots.
Thevirtualcompanywillusein-house
LANsattheheadofficeandresearch
institutesandregionalin-housenet-
worksatbranchofflces.LANswillbe
connectedasnecessary.Sinceregional
branchesarebehindinin-houseLANs,
thevirtualcompanywillrentaprivate
linenetworkfromNTTPCCommuni-
cations,aNTTsubsidiary.Theserver
ofthenewcompanywillbeinstalledat
theheadofficeandconnecteddirectly
toexistingcorporateservers.Servers
willlaterbeinstalledatninecommu-
nicationsoftwarecenterslocatedindif-
ferentareasofthecountryandoperat-
edundertheauspicesoftheCommuni-
cationSoftwareDevelopmentHead-
quarters.Regionalserverscanbeac-cessedviaISDNlines,forexample.
Investingabouthalfabillionyen,NTT
willinstall500to1,000personalcom-
putersfbrthenewcompany.Thisisthefirstvirtualcompanytohavesuch
alargescale,andwillbeamodelfor
othervirtualcompaniestof()110W.
*NECParticipatesinthe
FrenchInfCrmationHigh・
wayPlan
NECwillparticipateintheinforma-
tionhighwayplanbeingpromotedby
theFrenchGovernment,whichhasin-
troduced491arge-scaleprojectsfbrthe
constructionofaninformationhigh-
way.NEC,Bull,andFranceTelecom
havemadeajointproposalforimple-
mentingamultimediaelectronictele-
phonebookservice.ThisproposalwasselectedbytheFrenchMinistryofIn-
dustry.
93 JIQNo.lo1,1995
NECwillsupplythehardwareandsoft-
warerequiredforprovidinginfbrma-
tionincharacters,videoimages,and
voice.Bullwillconstructasystem
basedonFranceTelecom'sopticalfiber
andothernetworks.
Themanufacturersinvolvedplanto
supplyhigh-qualityinformationbyty-
ingupwithservicecompaniessuchas
publishersinFrance.Theservicecan,
forexa叩ple,providethetelephone
numberofanenterpriseasweUasits
productinformationintheformofvid-
eoimages.Thegroupplanstoconvert
advertisementsintelephonebooksto
high-qualityvideoimagesanddevelop
theservicetoimplementtelephone
shoppinginfuture.TheFrenchGov-
ernmentplanstotestthesystem,in-
vestingabout300millionyenoverthe
nexttwoyears,withcommercialization
plannedapproximatelyfor1997.
NECisthefirstJapaneseenterpriseto
participateinsuchEuropeaninforma-
tionhighwayplans.NECconsiders
thisagreatcatalystf()rdevelopingthe
multimediamarketinEurope.
、
JIQNo.101,1995 94
BackIssuesofJIPI)ECInformatizationQuarterly
(FormerlyJapanComputerQuarterly)areasfollows:
Publishedin1995
No.100:RecommendationsforG7ConferenceandProgramfor
AdvancedIndustrialInformationInfrastructure
Publishedin1994
No.99:
No.98:
No.97:
No.96:
EDIDevelopmentinJapan
ProgramforAdvancedInformationInfrastructure
TheComputerSystemandPatentInformationattheJapanese
PatentOf〔]ce・
InformatizationofHomeLifeinJapan
Publishedin1993
No.95:
94:
93:
92:
InformatizationPolicyinJapanCutting-EdgeNew-MediaTechnologiesinJapanFifthGenerationComputerSystems(FGCS)ProjectinJapanHypermediainJapan
Publishedin1992
No.91:
90:
89:
88:
JapaneseISDN:PresentandFuture
RegionalInf()rmatization・inJapan
RealWorldComputing&RelatedTechnologies
Information-relatedExaminationsinJapan
Publishedin1991
No.87:
86:
85:
84:
WorkstationsinJapan
VANServicesinJapan
CIMinJapan
LaptopComputerinJapan-Market&UserStrategies-
Publishedin1990
No.83:
82:
81:
80:
DisthbutionIn品 ㎜ationSystemsinJapan
ComputerSecurityinJapan
FinancialInf()rmationSystem .sinJapan
EDIinJapan
Publishedin1989
No.79:
78:
77:
76:
NetirocomputersandFuz'zyTheory-R&DTrendsinJapan-
Japan,sApproachtoPrivacyProtection
StateofCAL(CAI)inJapan
Sof、wareIndustryinJapan-StrivingfbrIncreasedProductivity-
Publishedin1988
No.75:
74:
73:
72:
PersonalComputersinJapan-AnUnab亘dgedAccount-
GlobalizationofTelecommunicationServices
TheMicrocomputerIndustry
-TrainingEngineers,CreatingApplications-
Informatization-HandlingTomorrow'sProblemsToday一
●
●
PleasesendtheORDERFORMdirectlyto=
PromotionDivision
JIPDEC
3-5-8Shibakoen,Minato-kuTEL:+81-3-3432-9384
Tokyo105JAPANFAX:+81-3-3432-9389
_________一 一 ー ー 一 一 ー ー 一 一 ー ー 一 一 一 一 ー 一 一 ー 一 一 一 一 ー ー 一 一
β
ORDER FORM
O Please sendmeJIPDEClnformatizationQuarterlyaschecked
below=
口 AnnualSubscription ¥13, 000
(includingairmailcharge)
口 BackCopies ¥3, 500percopy
(includingairmailcharge)No.
Total= ¥
0 lwillmakeapayment asfollOWS:
口 Bankfundstransfer
BankAccount: MitsubishiBank,ToranomonBranch
AccountNumber: FutsuYokinOOOO739
AccountHolder:(ZaidanHojin)JapanInformationProcessing
DevelopmentCenter(JIPDEC)
口 Checkenclosed
きignature:Date:
Name二.
.Position:
Company:
Address:
Tel: Fax:
蜘
Top Related