the speaking machine of wolfgang yon kempelen

Upload: damibard

Post on 04-Jun-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    1/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 151

    noise calculations re given in column 12, and valuesof Zr from Eq. (41) orZ r = B,-]- ZXB,-]--t-K-/0' (41)

    are given in the last column.APPENDIX 2: HEARING LOSS FOR SPEECH

    In this appendix he relationshipbetween he hear-ing loss or speech'4 nd the hearing oss udiogramwillbe considered. et/s be the hearing oss at the fre-quency f for a pure tone. It is the ordinate in theaudiogram. f we consider/shas the sameeffect uponthe threshold level as an attenuation --R from theflat response ystem, then by analogy to Eq. (23) thehearing oss or speech/, s given by

    10-'/'=0G'lO-s/'df'99),4H. Fletcher,"3, methodof calculating earing oss or speechfrom an audiogram," J. Acous. Soc. Am. 22, 1 (1950).

    If we consider nly the octave requencies 25, 250,500, 1000, 2000, 4000, and 8000, then the followingequation s approximatelycorrect./5,= - 10 og{Y'.W10-kn0} (00)

    where k takes the successivealues of 125, 250, 500,1000,2000,4000, and 8000. The weightsare givenby/ 1.4kWk=o.7kflf'

    For 125 c.p.s. W=0.000, 250 c.p.s. W=0.003, 500c.p.s. W-0.104, 1000 c.p.s. W-0.388, 2000 c.p.s.W=0.395, 4000 c.p.s. W--0.106, 8000 c.p.s. W-0.004.So for most purposesone needs to consideronly thefour frequencies 00, 1000, 2000, and 4000 and useweights0.1, 0.4, 0.4 and 0.1.For a fairly flat audiogram t is approximatelycor-rect to take an average of the hearing loss at 500,1000, and 2000 c.p.s.

    THE JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA VOLUME 22, NUMBER 2 MARCH, 1950

    The Speaking Machine of Wolfgang yon Kempelen*HOaE DUDLEY

    Bell Telephone aboratories, urray Hill, New JerseyAND

    T. H. TARNOCZYBiophysical aboratory f the nstitute or Anthropologyf theMuseumof Natural History,Budapest, ungary(ReceivedOctober17, 1949).

    The physiological otions nvolved n speaking an be indicated o the eye or to the ear. For the eyesuitably chosen ymbolsmay be written to indicate the physiological ositions ssumedn forming eachsound; for the ear synthetic soundsmay be producedby motions n a mechanismbuilt to simulate thespeech rgans. he degree f phonetic uccess ay be estimatedn the case f the visiblesymbols y listeningto soundsormedwhen he indicated hysiologicalrocessesre carriedout, and n the caseof the speech-simulatingmechanism y comparinghe synthetic peech roducedo normallyspoken peech. ignificantadvances longboth the visualand the aural inesare describedrom earliest imesdown o the present.Wolfgang von Kempelenproduced he first speakingmachineworthy of the name around 1780. Thispapergives isbackground,descriptionf the apparatuse built,anda discussionf the methods sed nproducing he varioussounds,itting his work into the over-allpicture of speech-imitating evices rom thespeaking f idolsof ancient imesdown o the automaticelectrical econstructing f speechn the vocoder.For portraying to the eye the physiological haracteristics f speech here are discussedhe more out-standingmethodsrom claimed ymbolic lphabets f ancient anguagesown o the recentspectrographicvisible speech.

    OWARD he endof the 18thCentury Hun-garian, Wolfgang Ritter yon Kempden, or, inHungarian, Kempden Farkas Lovag, first built a com-plete and, on the whole, a surprisinglysuccessfulspeakingmachine.Speechwas ormedby manipulation* Orally presentedbefore the AcousticalSociety of America,May 5, 1949, New York, by Dudley with original draft byTarnoczy. The paper here, in general, ollows he oral presenta-

    tion including a set of figuresand also other material not in theoriginal draft.

    of mechanical lementssimulating he essential artsof the humanvocalsystem. n 1791he published 456-page book,' illustrated with 25 plates, describing isobservationson human speech production and hisexperimentsduring the two decades e had been work-ing on his speakingmachine.The appearanceof hisbook was a great social event. Introductory to thexMechanismuser menschlichenprache ebst er Beschreibung

    seiner sprechenden aschine. Also published n French at thesame time (1791).

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    2/16

    152 HOMER DUDLEY AND T. H. TARNOCZY'book are listed,assubscribers,22namesof outstandingpeopleof Austria, Russia, taly, and Hungary.Von Kempden was born on January 23, 1734, inPozsony,a city in westernHungary, now Bratislava,Czechoslovakia, nd died in.Vienna on.March 26, 1804.He reached igh government osition n the Habsburgmonarchybecoming,n. 1767, Aulic Counselor f theChamberof the Domain.of Empress nd QueenMariaTheresa.He traveledwidelyoverEurope.He made he

    application.o moderncivilization.s indicated y theindustries uilt uponspeech ince on Kempelen's ay.Man's unaided voice carries ess han a mile at a speedof only 12 milesper minute,a speed ctuallysomewhatless than. that of a fast jet plane of today. But thetelephonendustries arry he spoken. ord housandsfmiles and at speeds f millionsof milesper minute;radio industries rovide or audiences umberingmil-lions of people at a time; phonographndustriespla_n.sf the fountainn SchSnbrunnnd later the. preservepeechor unborn enerationso hear.Trans-designs f the Royal Castleof Buda. He organizedhewoolmanufacturen.South-Hungary. e wasa skillfulengineer nd a genius n organization.The question arises as to what motivated yonKempelen o attempt the building of any speakingmachine, or no one had, so far as known.,attemptedthis difficult ask n. he centuries recedinghe day in1769he starteda job that, off and on.,would take overtwo decadeso complete. he answern part, no doubt,is that he had some nterest n the problemof speechby deaf-mutes.But, more fundamentally, here wereseveralactors avorableo the developmentf a speak-ing machine y a man.of his ability and inclination tthat particular ime. Pervading ll timeshasbeen.hebasic mportance nd significancef speech.But, inaddition o this general rge, herewasstirring broadin yon Kempelen'sayan. rousedcientificuriosityn.the wakeof the Reformation, nd from this there hadarisengradually he faint beginningsf physiologicalphonetics s a science. hese hree actorswill be dis-cussed riefly before eturning o his book.

    Speechs of suchbasic mportancehat civilization.asknown.odaycouldnot existwithout t, yet like theair we breathe, t so envelops s that we fake it forgranted.The continuingmportance f speechn. its

    FO. 1. Automat6nchessplayerf von Kempden. CourtesyofOxford University Press.

    missionof the written word has been accomplished ythe telegraph ndmore ecentlyhefacsimilendustries.And even n yon Kempelen'simes herewere ndustrieshandling the printed word which have expandedenormouslyn. the intervening earsto flood us withbooks,magazines, ewspapers, aps, folders,pam-phlets, advertisements,tc. Religion,education.,heprofessions,n. actallorganizedociety,ependedhenas now upon he spoken nd written.word.By speechman.raised himself to a positionabove and distinctfrom the lower animals.Accordingly,n ancient imes,man. ook speech s a symbolof his divineoriginandassumedhis gods were, somehow, peakinggods.Naturally then., he priests ried to make their idolsappearo speak irectly o thepeople. or thispurposespeech aspiped n. rom a concealedriest o makewords issue forth from the mouth of the Oracle ofOrpheus n the Isle of Lesbos.' n one evelation.,hisoracleaccurately redictedhe violentdeathof Cyrusthe Great n hisexpeditiongainsthe Scythians.n theMiddle Ages, thousand ears ater, RogerBacon ndothersbuilt small alkingheadsof bronzeand woodasmodels f ingenuitywith concealedubesbringingn.aspeaker's oice but without intent of superstitiousdeception..Von Kempelenwasborn.n a time of aroused cien-tific curiosity.Galileohad passed n. ess han.a centuryearlier.Therewasstirringa healthful kepticism hichdemandedhat, ruth be sought ot in a blind faith butby experimentalethodsf cut-and-try. hisscientificvigormanifestedtselfstronglyn the design f auto-mata o produce otionsfvarious orts hen.nergizedas by windinga spring.Thus a Frenchman amedVaucan.onabuilt a man-like figure, or android as itwascalled, hat playeda flute with all the complicatedmotions eededor the lips and fingers. e later madean automaton.or simultaneouslylayinga shepherd'spipeheld n onehandwhilebeating tambourwiththe otherhand.He alsodeveloped miraculous uck,perhapsn someways the most extraordinary uto-maton ever constructed.t moved ts wingsand walkedin. a natural manher. t drank water, muddling n. theact. It would take corn from one's hand and swallow itwith a completesimulation. f the digestion rocess

    The Scientific apersof Sir CharlesWheatstone1879). SirDavid Brewster, etters n NaturalMagic (1832).aSci. Am. 24, 32 (1871). An Account f theMechanism f anAutomaton,tc. (translation y J. T. Desaguliers)1742).edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    3/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 153aided by chemical means. Another Frenchman, LeDroz, made up a writing child while his son made abullfinch hat would ump up from a snuffbox, wag itstail, spread ts wings,pour forth a melodious ong,andthen dart down into the box as the lid closed.

    Now yon Kempden was also a skilled mechanician.In the spirit of mechanicalngenuit2 f the times hebuilt4 in 1769 his famous"chessautomaton" (Fig. 1)with a turbaned Turk playing an almost unbeatablegame of chessseated at a desk on which there wasmounted a chess board. The device was exhibited overEurope in the next four years by yon Kempden andafter his death by Maelzel in Europe and later inAmerica until burned in a fire in 1836. Before theexhibition the cloak at the back of the Turk was liftedand, severalat a time, various compartmentsof thedeskwereopenedup'to show heir apparentemptiness.When everythingwas closed,with a lot of noise rominsidemachinery, he Turk laid down a long-stemmedpipe and started a chessmove. Von Kempden himselfremarked that this chessplayer was not a true auto-ma.ton, ts only outstanding eature being the skill ofthe deception.The deceptionapparently consistednhaving a skilled player concealhimself n the cabinetand play the game from information received n theupward displacement f a small iron ball under eachsquareon which there was placedone of the chessmen,each of which containeda strong magnet. The c.hessplayer is said to have beaten Napoleon n one of itsgames.n 1821,when the playerwasbeingexhibited nLondon, Robert Willis, noted for his later researches nsyntheticvowel production wrote a 40-page bookleton the chessplayer. His attempted ar/alysisdoesnotagree entirely with Murray's explanationof the auto-maton mechanism.

    The difficulty of the problem, the mechanical n-genuity, the breadth of view and to some extent thefeel and daring for showmanship anifestedby yonKempden in the chess layer are characteristicshownin his development f the speakingmachinestarted nthe same year, 1769. But the chessautomaton wascompleted n six months while the speakingmachineoccupied on Kempden for much of the time for overtwenty ears efore epublishedis esultsn 1791.This aroused nd growingspirit of experimentaln-vestigation n sciencen generalnaturally led to ques-tions as to the physiology f speech roduction, huslaying the foundationof experimental honetics s ascience.A century before yon Kempelen's studies,Baron FranciscusMercurius ab Helmont publishedabook in Latin contending hat the Hebrew alphabet4H. J. R. Murray, A Historyof ChessOxfordUniversityPress,London, 1913), pp. 876-7.5 Robert Willis, Trans. Camb. Phil. Soc. $, 231 (1829).6Robert Willis, An Attempt to Analyze. he Automaton ChessPlayer of M. de Kempden with an Easy Method of Imitating theMovements f that Calebrated igure, with 10 Plates (1821).7Baron FranciscusMercurius ab Helmont, Alphabeti vereNaturalis ttebraici BrevissimaDelineatio (1667).

    Fro. 2. Hebrew etter M as a tongue osition .according o Helmont.was a "natural" alphabet n that the letter symbolsrepresentedhe actual tonguepositions nd so shouldbe useful n teachingspeech o the deaf-dumb.VonKempelenreproducedour of Helmont's illustrations,'oneof which sshown ereasFig. 2. The tongue ositionis for M, the 13th etter of the Hebrewalphabet, ro-nounced em, as ndicatedn Hebrewbelow hepictureof the head. The figures n the borderof the headbandare related forms for the letter M, those numbered"4," for instance,being copied from ancient coins.Von Kempden criticizedHelmont, pointing out thatfor the soundof M, the tongueposition s of little im-portancebut that the significantpositions re in theopennasalpassagesnd the closedips, although helatter are shownopen by Helmont. The small elementaheadof the distorted ongue s part of the Hebrewletter M but seemso haveno phonetic ignificance.A more serious epresentation f alphabeticchar-actersby the useof phonetic ymbols ortrayingvocaltract positions aspublishedn a book a year ater byBishop ohnWilkins n England.Figure3 is a copyofpage378 of his book.Here are illustrated he picturesof the he/d with' vocal positions or 34 soundscon-sisting f eightvowels nd 13 pairsof voiced onsonantswith unvoiced ounterparts. omparinghe consonantsin the list with thosegivenby Fletcher onenotes hatboth lists contain our voicedand four unvoiced topconsonants,our voiced and four unvoiced ricatives,and five voicedsemivowels. ilkins differs n showing

    8 BishopJohn Wilkins, An Essay Towards Real Character nda Philosophical anguage 1668).9 t, FletchehSpeech nd Hearing 1929), p. 6.

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    4/16

    154 HOMER DUDLEY AND T. H. TARNOczY

    FIO. 3. Illustrated physio-logical lphabet ymbols fWilkins.

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    5/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 155

    a completeset of unvoicedsemivowels l, hm, hn, hng,and hr, and omitting he three ransitionals , w, and h.The unvoicedsemivowelsre found n some anguageswith Welsh containingall five. He lists y as a vowel,probablyconsidered as a vowel (we= r) and reallyprovided a multiplicity of h soundswith his five un-voiced semivowels.He showedonly eight vowels asagainst 11 in the Fletcher ist. In the upper right-handcorner for each of the 34 sound illustrations are Wilkins'proposed honeticsymbols. n the upper left portionof Fig. 3, nine headsare shownwith no sectioningn athree-by-three arrangement for those soundshavingsignificantip positions.n eachof the other 25 illustra-tions, a section s shown o reveal the positionof vocalparts within the mouth and throat. Some physicallysignificantsymbolsused are a circle for the roundedlips of the o sound,a pair of closed ips for p, closedlips with a semicircle r small wave to representvocalcord action for b, a straight horizontal ine through themouth to represent he flow of air through. he mouthfor the consonants here here is a relatively free flow,and a slightly curved ine at the top to represent heflow of air through the noseas in the caseof m, hm, n,and hn.

    Wilkins, in the samebook, page376, shows sort ofsyllabary eproduced ere as Fig. 4. Consonantsollow-ing a vowel in a syllableare shown n small letters aslisted n column1; thosepreceding he vowel are shownby capital letters as listed n column9. Column 2 givesthe consonantsymbols or indicating the consonantsalone. The row numbered gives the six vowels heusedand row 2, their symbols.Columns -8 givevowel-consonant combinations-and columns 10-15, the con-sonant-vowel ombinations.He employsa small circleplaced high, middle, or low for the first three of hisvowelsand a semicircle imilarly placed for the otherthree. His consonantist of Fig. 4 includes , w, and y,though these are not in the illustrated list of Fig. 3.The reason or including them here is that they com-bine with the vowels. At the bottom of the figure isthe Lord's Prayer written with thesesymbols.To round out the picture of phoneticportrayal, wenote that since yon Kempelen's time, considerablefurther advancehasbeenmade. n particular,AlexanderMelville Bell worked out a set of symbolshe termed"visible speech," howing n minute detail the completevocal action in producingnot only the speechsoundsbut also whispers,whistles,sobs,grunts, clicks,hisses,sighs,coughs, neezes, isses, nd all other sounds ro-ducible by the human vocal mechanism.A detailedexplanationof the symbols s also given by his son,Alexander Graham Bell.n Figure 5 2 showshis basic"universal alphabet," the sounds rom it used in the

    x0Alexander Melville Bell, Visible Speech--TheScience f Uni-versalAlph. betics D. Van Nostrand Company, nc., New York,1867).n Alexander Graham Bell, Mechanismof Speech 1907), secondedition.' From ;Explanatory ecture n VisibleSpeech1870).

    Englishalphabet,and two diagrams ctdefine he chiefconsonantand vowel symbols,respe.tively. In theuniversalalphabet able, consonants re representedncolumns 1-4; the vowels in 6-8; glides in 5; throatsounds and modifiers in 9 and 0. The first six rows ofconsonants re the unvoicedones, with' the voicedcounterparts given in the next six rows with shortinsidebars added to represent oicing.The diagramatthe left shows hree cord positions,bar for voicing,a circle for wide open as in producing he soundh ofcolumn9, row a, i.e., 9a, and an X for the closed lottisin 9c. A partial opening s n whisperings shownby anoval in 9b with a bar added o showhoarse ocality asin 9h. The point of constrictionaperture) s representedin the diagram by the concavequarter of the three-quarter circles.The aperturemay be at the back of thetongue, he front of the tongue, he point of the tongueor the lips producing he consonant olumns1-4 whichare called back, front, point, and lip consonants,e-spectively. n rows b and h of the consonant ymbolsthere s a minoraperture n the quarteroppositeo themain apertureas represented y a pair of small three-quarter circles; the indicated consonants re termedback-mixed, ront-mixed, point-mixed,and lip-mixed.In rows c and i the main quarter of the symbol isindented to represent dividing the air stream as inthe sound f (4c); such consonants re called back-divided, front-divided, etc. Mixture combined withdivisioniveshemixed-dividedonsonantsepresentedin rows d and k. Rows e and 1 have the three-quartercircle closedby a straight line representing "shut"positionas in the stop consonants;hus l e is the back-shut consonant , 11 s the back-shut oiceconsonant ,etc. In the final rows f and m of the consonants, hewavy line from the uvula in the diagram refers toclosing he nasal pa. sageswith the uvula to give thenasal consonantsalledback-nasal, tc. The meanings.of the symbols or the tonguepositions or vowelsaregiven n the rightdiagram. n upright ar ndicatesvowel as in columns6-8 throughout, while the smallhorizontal bars in the lower six rows indicates roundvowels, .e., vowelsspokenwith round lips. The back,front, and mixedpositions f the tongueare representedby dots or hooksat the left, right, and left plus rightof the vertical vowel bar, respectively.The hook indi-cates that the voice channel s openedwide, the dotnot sowide.The elevationof the tongue s indicatedbythe dot or hook positionson the vertical bar, a lowelevationby the bottom position,a high elevationbythe top position,and a mediumor mid-elevation y thecombination of one dot or hook at the bottom and theother at the top. Thus u as in pull listed as 6k would bedefined s a high-back-wide-roundowel.This explana-tion covering84 consonants nd vowels,-manynot tobe found in any language, s considered dequate orshowing ow the physiologicalroduction f the variousspeech oundss indicated n great detail by Mr. Bell.The International Phonetic Association, ounded in

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    6/16

    156 HOMER DUDLEY AND T. H. TARNOCZY

    1

    3

    6789

    16

    1

    z]

    2.'

    o

    FI6. 4. Wilkin's English lan-guage yllabaryor a philosophicallanguage.

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    7/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 1571886, drew up an International PhoneticAlphabet n1888 providing symbols or presumablyall speechsoundsbut without' specialphysiological onnotationas in the caseof symbols hown n the presentpaper.Recently Potter a and his co-workers ave developedvisible speech nalysiswith the soundspectrograph,portraying the sound spectrum automatically, thusreplacingassumedpositionsof the vocal parts withobjectively ecorded hysicalcharacteristicsf actuallyspoken peech. he words"visiblespeech" rom suchaspectrogram re shown n Fig. 6a, while Fig. 6b showsthe set of manualalphabeticsymbols sed n an experi-mental training program to portray the distinguishingcharacteristics f speechsoundsas revealedby spec-trograms.In the century preceding on Kempelen, here wasmuchspeculation y linguists, hysicists, sychologists,and speecheachers n the mechanism f speech. heunderstanding f physicalsciencewas advancing romphilosophical peculation o scientificexperimentationwith physical apparatus. Von Kempelen, as statedearlier, started working on his speakingmachine n1769 and continued to 1791 at least. Others were alsoactive in this period. Thus, in 1779 the ImperialAcademyof St. Petersburgoffered ts annual prize forexplaining hysiological ifferencesn, and makingappa-ratus for, producing he five vowel sounds/A father),E (they), I (machine),O (note),and U (crude).Theseare the long vowel soundsas used on the Continentand 'so, for consistency,will be used here. The prizewas won by Christian Gottlieb Kratzenstein, 4 born inWernigerode,Germany, who became a ProfessorofPhysiology irst at Halle and later at Copenhagen. emade five tubes as shown in crosssection n Fig. 7.These tubes roughly approximated he size and shapeof the vocalpassages hen set to produce he differentsounds.All were energizedby free reedsexcept the Itube which wasblown into directly in the fashionof anorganpipe. t is interesting o note that Robert Willis,5showed hat the shapesof Kratzenstein's ubes werenot impoitant as the required resonances ould beobtained rom a singlepipe the length of which wasadjusted or the different sounds, nd that, in fact, thevowel seriescould be coveredalternately forward andin reverse s the pipe length was ncreased.From the foregoing ackgroundwe return to a reviewof the highlightsof yon Kempelen's ook. This book s

    15Potter, Green, and Kopp, Visible Speech D. Van NostrandCompany, nc., New York, 1947). Someother writers have de-scribedapparatus or automatic writing of the symbolsby thevoicebut the symbols ack in definiteness nd uniformity so thatthey can hardly be classified s alphabetic n the stageof develop-ment described. ee, for example,J. B. Flower, "phonographicalphabet" n "The true nature of speech,"Trans. A.I.E.E. 35,213-48 (1916)and J. Dreyfus-Graf,"Steno-sonographiclphabet"in "Le sonographe:16mentst principes,"Schweizer rchiv 14,353-62 (1948).a4 Tentamen coronatumde voce," Acta Acad. Petrop. (1780).The completearticle is in "Sur la naissance e la formation desvoyelles," . de Physique 1, 358-80 (1782).

    divided nto five chapters, hus'Chapter , pp. 1-28--Speech n General;Chapter I, pp. 28-56--Origin of Speech;Chapter II, pp. 57-177Vocal Parts and Their Functions;Chapter V, pp. 178-387Sounds of EuropeanLanguages;Chapter V, pp. 388-456--The SpeakingMachine.

    The first two chaptersgive brief discussionsf speechand its origin in quite general ashion. n the thirdchapterhe examinesnto whichbodilyparts contributeto the forming of speech ounds nd how thesepartsfunctionnormallyand under aulty conditions. e thusexplains he voice, considering articularly the func-tioningof the lungs, rachea,glottis,nose,mouth, ongue,teeth, and lips. In the fourth chapterhe lists most ofthe European lphabetic ounds.He then proceedsoreach soundand by groups o considerwhat would bethe simplest and-operatedmechanismo produce airimitations of these sounds. First he observes his ownvocal systemas objectively as possible o determineexperimentallyhe relative mouth and tongue-channelopenings or the vowels a, e, i, o, and u, with theseresults:

    Vowelsound Mouth opening Tongue-channelpeninga 5 3e 4 2i 3 1o 2 4u 1 5.

    The numbersshow elative sizes, ncreasingrom to 5.He extends his observations to fit into this tabulationsevenother vowels or a total of 12. Then he proceedsto the consonants.igure 8, yon Kempelen'sPlate XII,illustrates his method of practical visualization andadaptationas he reduces pplicable ocal parts to thesimplestmanuallyoperablemechanismor a groupofsounds. n this figure, for making the voicedexplosivesoundsB and D, he substitutes or the mouth a woodenbox, for the lips a pair of hingedwoodenshutters, orthe tonguea hingedwooden lap, operableby a string,and for the air supplya tube to whichcouldbe fitted areed for voicing.The two top subfigures re for the Bsoundbeforeand after the start of the explosive missionof air; the two lower ones, for D. These cut-and-trydesignswere useful o yon Kempelen n clarifyinghisunderstanding f the mechanics f soundproductionbythe human mechanism.His final speaking machinesometimes used less idealized methods as even in thiscaseof the B and D sounds, s will be explainedater.Von Kempelen n the fifth and final chapterof hisbook describeshe stepsof buildinghis speakingma-chine, his tests as he went along and finally how toproduce he different soundsand combine hem intospeech.He began with a search or a suitablesoundsource for imitating the tone from the vocal cords.The most natural-sounding ourcehe found after ex-amining many musical nstrumentswas a drone reedfrom a bagpipe.He first tried to produce he vowelswith a bell-shapedmouth attached to such a reed as

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    8/16

    158 HOMER DUDLEY AND T. H. TARNOCZYCon$onanl to$ition$

    C"

    THE UNIVERSAL ALPHABET.

    b/owe/ Posit/on$

    Connt8. sonColumn,

    Columns o.Colnmn , 0.h and odis.

    TheMarginal umbersndLettersmaybeused,nsteadf theVisibleSpeechetters,o expresshe mechanismf soundsn commonype. Thefollowingxampleshowhe English, cotch,nd rishpronunciationsf thewords ' Visible Speech"4i8d2h 413i 4e ob3e3b

    gtn to m,Voeatil Con,.sonants.Labialo, edOlides,LabialidVoloels.

    Fro. 5a.

    shownn Fig. 9. Instead f usinghesliding lateshownat the bottom he usuallyplacedhis hand on or nearthe bell mouth n various ositionso form he differentsounds ut only obtainednon-vowelonesand latervowel-like tones with a characteristic "ah" sound nomatter howhishandswereplaced. he resultswerenotvery satisfactoryo for his secondmodelhe built aconsolesshownn Fig. 10 with 13 piano-like eyscon-

    trolling ir froma common ellowso separate assageseachcontaining reed, hususingndividualeedsorthe different voiced sounds.He tried rectangularboxessuch as the two at the left and then round boxessuchas the four at the right. He writeshe obtained omegood owel istinctionsor he irst ime, orming faira, o, andu. He alsoobtainedair consonantoundsorp, m, and1. With these e formed imple ordsike

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    9/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 159THE ENGLISH ALPHABET.

    [Thetalic ettersre heEnglishquivalentsf theVisible peechettersn thecorrespondingsections f the UniversalAlphabet.]I 2 ' 4 5 6 ? I 8 9 0 '.I IIIhe ro eels s w up ale[[ old.f boy ell

    tin near -tn* -ces* ill dk t p ask -al* airink hint lamp [arm ir an f

    .I I I I[zeal azure ay oldl[.] l v all] then pull-ure* kI z oro i i .i -n [ n m o -or m1 , 3 4 b 6' 7 8 9 0i

    The soundsmarked occuronly in unaccentedyllables;as inI[1) (mention); DC[lIffi. places; O[iil% (fatal;IDC.[[ (pleasure);OI;lD%Oi (history; [%ti (orator).

    The' glide' 5a heardbetween vowelandr; as n hero, iry, iery,glory,&c.The soundsn ' ale' and ' old' include he ' glides' $c.and $1. Thus :--[ (ale);}tl (old).R final or beforea consonant,s in air, arm, &c., is the ' Point.Glide' $d.Thus-- air); [ (arm).Accent s alwayson the first syllableunlessotherwise(ex'pre$sed).The mark is placed [( (be'fore)he syllable o which

    it G)[[OI (refers).FIo. 5b.

    Fro. 5. The "universal lphabet"of AlexanderMelville Bell with symbol-definingiagrams nd llustrationsin English. Seeopposite age or Fig. 5a.)"papa" and "mama"but noticed wo troubles;first, hesounds id not blend together n a natural way, andsecond,owelsn particular ameon ratherexplosively,thus adding a k-like sound.He decided that to overcome the first trouble heshould oawaywith hismultiplicity f reeds, roducingall voicedsoundsrom the same eed.Then, to preventthe explosive ncoming f vowels,he lined the reedandthe edge t beat againstwith thin soft glove eather

    around he closure n bothsides. e experimented ithan adjustable ire to changehe effectiveengthof thevibrating eedand thusalter the pitch but he found tdifficult o obtain a variablepitch so he satisfied im-selfwith a monotone,tating hat he eft this mprove-ment to later workers n the field. In this connection,it is of interest o note that J. R, Ewald developedaJ. R. Ewald,"Zur konstruktiononpolsterpfeifen,"fltigersArchiv f. die gesamte hysiologie52, 171 (1913).

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    10/16

    160 HOMER DUDLEY AND T. H. TARNOCZY '

    somespringed ushionso vibrate in a pipe in closeresemblance to the human mechanism and that theartificial arynx inventedby R. R. Riesz 6gave a fairrange of pitch variation according o the air pressureapplied o the vibrating eed, ncreasingressure ivingincreasing itch; also that the method yon Kempelentried had long been successfullypplied to pipe organreeds.Having thus failed in his first two attempts to makea speakingmachine along the lines representednFigs.9 and 10, yon Kempelen hen startedanew.Thisthird effort resulted n his final speakingmachine, heessentials f which are shown with a scale ndicatingsize n Fig. 11, a reproductionof his Plate XXV. ThebellowsX shown n part at the top is used to set upan air pressuren "wind-box"A. With by-pass eys sand sch closed, he excess ressuren box A can onlybe reduced y leakage s discussedater and by passageof air against he reededge, hussetting t in vibratorymotion.A brief descriptionwill be given of how the differentsounds re produced ccordingo yon Kempelen.Theoperatorests is ightarmon hebellows andpumpsit with an up-and-downmotion,speech eingproducedon the down motion. The fingersof the right hand areset to operate he special onsonant ontrolsmarkedr, sch,n, m, and s. The left hand s placedpalm inwardbefore he openingbc of bell C. The vowelsare pro-ducedby working he bellowswith the right elbowwhile blocking he nostril-imitating ubesm and n byfingers f the right hand,with the left handset n suchposition eforeC as listening nd practice ndicatedbest or the particularvowelbeingproduced. or sounda the hand is kept distant from the mouth opening;for e the hand s hollowed lightlywith its bottom edgeagainst he mouth and its top edgeabout one inchaway; for o the top of the hollowed and shouldbe

    about one-half nch from the mouth; for u the hand isheld flat with the openingof the mouth reduced o aminimumshortof stopping he reedvibration; but withthe openinggreater than for i; for i the flat hand isplaced ightly across he mouth openingand the indexfinger hen crooked, o that there appears t the secondknuckle a small opening,more air pressurebeing re-quired for this vowel than for the others. He says thepositionsor othervowelssuchas the umlautsare inter-mediate o the givenpositions nd can easilybe locatedwith a small amount of practice, His descriptionofpositions nfers he tried to modify a second esonantfrequency f the vowelspectrum s doeshis considera-tion of the two openings,ongue-to-palate nd mouth.An over-allpicture of the consonants ade on yonKempelen's machine is obtained by comparing theones he claimed with a list of the 24 consonants usedin Englishas given by Fletcher 9

    Class Produced Not producedSemivowels 1, m, n, r, ngStops p, b; t, d; k, g; ch, j (judge)Fricatives f, v; s, z; sh, zh; th (thin), th' (then).Transitionals h, w, y (German j) 'Yon Kempelenalso produced he soundof ch (ich)found in German but not in English for a total of 19consonantsounds.Of the six sounds isted as not pro-duced,none are mentionedby yon Kempden. Of thesesix sounds, nly ng is found n the German anguage.Of the four semivowelsisted as producedby yonKempelen ound1 was made ike the vowelsbut withthe left thumb curving inside the rubber mouth tocorrespondo the way 1 is producedn normal speechwith the tongue arched to divide the air stream inthe mouth. The soundof m wasproducedby closing hemouth with the left hand while leaving open bothnostrilsand soundn by leaving open only one nostrilof thespeaking,machine.o producehe adjusted is

    FIG. 6a. Recentvisible peech pectrogr.mPotter). Courtesy f D. Van NostrandCompany,nc. R. R. Riesz,Description.nd emonstrationfanartificialarynx,".Acous.oc. m. ,273 1930).

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    11/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 161

    Fro. 6b. Manualsymbolsor alphabet f recent isible peechPotter).Courtesy f D. Van NostrandCompany,nc.positionsor the following owel and then depressedthe key markedr which pusheda wire into contactwith the vibrating reed, thus giving a rattling effect

    which was claimed o producea trilled r that was notperfectbut better han manypeople ouldmake.Of the six stop consonants roduced, e made p byedistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    12/16

    162 HOMER DUDLEY AND T. H. TARNOCZYsssssssssssssssssss

    ssmmmmmmmmmmmmmmmmmmmm

    Fro. 7. Schematicepresentationf Kratzenstein'sive vowelsynthesizersfrom Young'sNatural Philosophy, 845).dlosingll openingsnd hen,when ressureas uiltup, releasing he hand suddenly rom in front of themouth. This soundwas rather weak at first so he pro-vided an auxiliary storagebellowsunder the left sideof the wind-box,not visible n Fig. 1 but clearly shownin the next figure. He found that building up the pres-sure in the "wind-box" and the mouth for sound presulted, upon release of pressure, n sufficient airpassing he reed o produce voicingeffectsohe placeda small by-passing rass ube as shown n Fig. 11 nearthe sch key between the wind-box A and the bellmouth C. This, he says ed to a goodp sound.Addingvoicing to p gave b. For t, d; k, g, he did not provideseparate means as would seem indicated but afterconsiderablexperimentatione decidedhathe couldmodify the p and b sounds n a way, that he doesnotdescribe n detail, to give tolerableresemblanceso thedesired sounds.

    Of the five fricatives and three transitionalspro-duced, he made f from the leakage of air with allopenings losedwhile exertingstrongbellowspressure.He made h in the same way but with the mouth leftopenand with lesspressure n the bellows;Germanchwas made like h but with the bellowspressed lightlyharder but not enough o vibrate the reed. He made vlike f except hat a small escape f air waspermitted atthe mouth openingbc, sufficient o vibrate the reed.With less air but a larger percentageof vibrationalpower from the reed, the soundbecamew. The soundof s was made by depressinghe key marked s witheverything else closed; n this case the depression fkey s openeda by-pass o air at the side of the reedwith the air escaping hrough the small funnel shownunder he s key, this funnelbeingdesigned f a resonantsizeand shape o make a hiss ike the sounds. Addinga little voicing o s gavez. Similarly,sh (alwayswrittenschby yon Kempelen)wasmadeby depressinghe keymarked sch, whereupon h.e air was by-passed o theother side of the reed through the escape ube shownat the bottom which tube was designedor an sh soundresonance.Adding voicingproducedy, i.e., German j.An intermediate amount of voicing could have pro-

    duced zh, one would think, but von Kempelen doesnot mention zh at all.The by-pass ube from wind-boxA to bell C evidently

    wasted someof his air for he tells how he could onlyproduceshort combinations f sounds n connectedspeech, from a complete depressionof the bellows

    Fro. 8. Von Kempelen'sschematic howingessential eatures ormechanizing roductionof B and D sounds.

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    13/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 163although he bellowshad six times the air capacityofthe lungs.Because f this, he was imited to producingshortphrases t a time suchas "LeopoldusSecundus,"while the human vocal system can produce phrasesseveral times longer on a single expiration of thebreath.

    Summarizing,of the 19 consonant ounds on Kem-pelen ells of producing,we find six (p, b; t, d; k, g)

    FIG. 9. Von Kempelen's irst vowelsynthesizer.

    FIo. 10. Von Kempelen's onsoleor some owels nd consonants.

    FIG. 11. Von Kempelen'sinal speakingmachine.were stop consonants adewith an arrangement e-signedor p; five,consistingf the three ricatives,, v,and Germanch, and two transitionals, and w, weremadeby using scapingir having hissing oundsimilar o f; seven s, z; sh, y; m; n; r) weremadebyswitchingn fivedifferent esonant assages;ndone,1,was made, as in human speech, y splitting he air-stream in the mouth.This completeshe description f yon Kempelen'swork on his speakingmachine.While he mentions isbelief that the final machinecan easilybe fitted withkeys ike a piano the secondmodel,Fig. 10, had keysbut not a single eed)and says n conclusionhat if hefoundime o improvehemachineewould ontinuehis writing, explainingwhat he had done,we have nofurther ecord f any suchwriting rom the date of thepublication f his book n 1791 o 1804whenhe died.Onemustadmire he patien'ce f the man who worked22 yearson hisspeech achine efore ompletingt tothe pointof wanting o describet in writingand alsothe perserverancef onewho twicediscards ome woyears'workand startsovercompletely fresh.

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    14/16

    164 HOMER DUDLEY AND T. H. TARNOCZY

    He states hat in three weeks ime a personcan makeastonishing rogressn playing he machine f he limitshimself to the Latin, French, and Italian languages.German is much harder becauseof the prevalenceofconsonants in German.Von Kempelenwas not only a skilledmechanicianbut he had an acute and observantear for speechsounds as well as a lot of common sense. Some of his

    first-hand observations eveal a true understandingofthe manifold nature of speech nd its interpretation.The similarity of all vowelswhen sustainedor a con-siderable time is a characteristic the modern worker onspeech s likely to observe,perhapswith annoyance;its disappearancen dynamicspeech s cheering.Thedifficultyof startinga vowel soundwithout an accom-panyingexplosive ffect s alsowell known.He againobserveshe difficulty of getting a substitute or thevocal cords that sounds at all voice-like. He mentionsthat. selectinga high pitched voice is an advantagebecause a child's voice is not criticized. He made useof the lack of sameness from voice to voice when hemodified the p-b sounds o obtain his tolerable re-semblanceso the t-d and k-g sounds.Finally he ob-served hat peoplewould nterpret soundsmuch moreeasilywhen given somecluebeforehand.In conclusion,we shall round out the picture by

    briefly mentioningsome significant urther develop-ments f speechynthesisromhis imeon.About hissame ime, Kratzenstein nd alsoAbb Mical of Franceare said to have built speakingmachineswhich theyexhibitedn Pariswith pin-cylinder rivesas in musicboxes.These were presumably nferior to yon Kem-pelen'sn qualityof speech roduced. rofessor heat-stone, rom yon Kempelen's escription,uilt a speak-ing machine7 which he demonstratedn the Dublinmeeting f the BritishAssociationor the Advancementof Sciencesn August, 835. This s shown chematicallyin Fig. 12. In 1846 a certainProfessoroseph aberof Viennademonstrated speechmachinesadvertisedas "Euphonia" n the Egyptian Hall, Piccadilly,London.The device s pictured n Fig. 13. A ticketcostingone shillingpermitted he bearer o hear theperformanceonsistingf ordinary peech, hisperedspeech, onversation,nd the singing f airs endingwith "God Save the Queen" (Victoria's eign). Thismachine s said to have been a big improvementoveryon Kempelen's, articularly rom the standpoint fhavinga variablepitch that permittedsinging.Helmholtz with a seriesof tuning forks, Koenig 'with a shaped iren,Miller2and StumpF with setsofpipes,Preece nd Strohswith gearedwheelsmakingphonographicecord, ndmanyothers ave ynthesized

    SpeechSoundscome outhere

    "Sh" Whistle

    \.rilsResonatoro,eatherAuxJlJaryJBellows

    Lever/Reed/ cut offLever Bellows

    Whistle

    Leather Nostril\.................................ed....................................... r -"-'-F- Compressed/2'/////'//,,, A r C ha m be rSection hroughResonatornd Reed

    Fro. 12. Wheatstone's peakingmachine.London& Westminster eview 8 (1837); TheScientific apers f Sir CharlesWheatstone1879),pp.348-367.Proc.BritishAss.Adv. Sci. Notices (1835), p. 14.zsC. M. Gabriel,"Machine parlant de M. Faber," J. de Physique , 274-5 (1879). F. Techmer,Phonetik 1870). F.Techmer,Naturwiss.nalysendsyntheseerh6rbarenprache,"nt. Zeits. . allg.Sprachwiss., 69-170 1884).Helmholtz,Sensationsf Tone 1875), translated y Ellis.R. Koenig,Qudques xperiences'Acoustique1882).D.C. Miller, Science f MusicalSounds 1916).C. Stumpf,Die Sprachlaute1926).W. H. ?reece ndA. Stroh, On hesyntheticxaminationf vowel ounds,"roc.Roy.Soc. ondon 8, 358 1879).

    edistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 200.17.141.3 On: Mon, 20 Jan 2014

    20:57:06

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    15/16

    SPEAKING MACHINE OF WOLFGANG VON KEMPELEN 165

    Fro. 13. Faber'sspeech organ.

    ?,:..:::.::-.-::-:..::::...-:........;.'::: .....:..:::.:.;.::5:-:.-'-........???)?:;;.??.;.:,m.. ...........m..M..... .'-::...";J' '->:;?:;;::3;.::..?..:.-".: .'"'7'

    . . ..--.... . .::

  • 8/13/2019 The Speaking Machine of Wolfgang Yon Kempelen

    16/16

    166 HOMER DUDLEY AND T. H. TARNOCZY

    UTH -- -RADIATOR

    ,,I

    ,,

    RANDOM I SOURCENOISE I

    --'"'""'"'---ESONANCEONTROL[ AMPLIFIErlOSt ..... ORIvo,cEDI... j II II II

    I vooEnI NERGY WITCH

    Essential parle of the Voder.

    Fro. 15. Schematic f the yoder, electricalcircuit for producingspeechwith manual controls.Courtesyof Journal of the FranklinInstitute.verted to electricalwaveswhich energize sort of boneconductioneceiver o transmit the sound hroughcarti-lageof the larynx into the throat. Silent speakingmodu-lates suchpower into soundpatternsgiving he effectofsoundproduced rom other than vocal cord tones.Themodulatedoutput waves from the mouth at low levelare then picked up and amplified to produceunusualvoice effects.Similar effectsare producibleelectrically

    ANALYZERII ORIGINALSPEECHDEFINING$1GNAI , . ,I $YNTHE$1ZEFISOUNDTREAMS,iz i ....i I

    Fro. 16. Schematic f the vocoder, utomaticelectrical peechsynthesizer.ourtesy fJournalof heA cousticalociety fAmerica.with the vocoder mentioned later. Professor Firestone 28built and on November 3, 1940 demonstrated beforethe Acoustical ocietyapparatus or projecting oundsfrom an electricalorgan or other electricalsource ntothe mouth whereuponmouthinggave a modulation oproducemulti-voicesingingand similar effects.Stewart irst set up an all-electricalnetwork as inFig. 14 for makingsomeof the speech ounds.n 1939an all-electrical peechmechanism nownas he Voder (from the key lettersof VOice DEmonstratoR)shownin principle n Fig. 15 was demonstrated y skilledtrained operatorsat the New York and San FranciscoWorld's Fairs. Figure 16 shows unctionallya corre-spondingdevice known as the vocoder t (from VOiceCODER) havingan electrical peech ynthesizerimilarto that of the yoderbut making useof controlcurrentsfrom electricallyanalyzed speech or automaticallyoperating the synthesizer nstead of using manualcontrols.

    .8 . A. Firestone,"Artificial larynx for speakingand choralsinging y oneperson," . Acous.Soc.Am. 11, 357, 376 (1940).'J. Q. Stewart, "An electricalanalogueof the vocal cords,"Nature 110, 311 (1922).a0Dudley, Riesz, and Watkins, "A synthetic speaker," J.Franklin Inst. 227, 739 (1939).alH. Dudley, "RemakingSpeech," . Acous.Soc.Am. 11, 169(1939).