Transcription of Recordings - p. 2
SSSHP 8: "VOTRAX VS4, VS6, VS6G2 SYNTHESIZERS, 7/3/88
R.T. GAGNON"
SOURCE: Received 5/88 from R. T. Gagnon, Gagnon Electronics and
Research Co., 2097 Addaleen, Milford MI 48042, (in SSSHP USA
VOTRAX file). Both sides, no dolby, cassette.
CONTENTS: Side A, narration by R.T. Gagnon
1. VOTRAX 6.2 SYNTHESIZER, Apr 1975.
(syn: "Zero, one, two, three, ... 100, January, February, ...
December, Sunday, ... Saturday, Alabama, ... Wyoming, alpha,
bravo, ... zulu. To be, or not to be, ... A bird in the hand
... Four score and seven years ago ... The only thing we have
to fear is fear itself.")
2. VOTRAX VS6G2 (GERMAN) SYNTHESIZER, Aug 25, 1975.
(7 sen, x2, words: " ... ya, nein.")
3. VOTRAX VS6 SYNTHESIZER, Apr 1974.
(syn: "Zero, one, two, three, ... 100, January, February, ...
December, Sunday, ... Saturday, Alabama, ... Wyoming, alpha,
bravo, ... zulu. To be, or not to be, ... A bird in the hand
... Four score and seven years ago ... The only thing we have
to fear is fear itself"; from movie Mary Poppins, different
speeds: "Supercali ... shus"; phonetic sounds of "hello", with
explanation; 6 var of stress: "Goodbye.")
4. VOTRAX VS4 SYNTHESIZER, Mar 7, 1972. Synthesis based on Carl
Bixby's phonemes.
(syn: "Zero, one, two, ..., ten.")
5. VOTRAX VS6 SYNTHESIZER, Apr 1, 1974. Three demonstrations of
systems using the Votrax synthesizer.
(syn: "Welcome to the Raytheon Data Systems Exhibit. I am the
RDS500 Super Mini-computer. I am speaking ... Votrax ... all
aboard.")
ARC police response system, programmed by Interface Systems, Inc.
(syn: "Officer 136 inquiry on ... Officer 136 you have a ...
hit on licence.")
Pillsbury Order Entry System.
(syn, x2: "This is the Pillsbury Data Center ... Order 57631
... May 22 ... Order 57631.")
Side B, narration by R.T. Gagnon
6. VOTRAX VS4.
(Gannon: "The voice synthesizer you are about to hear is what
we call the VS4. The VS4 has been developed by Cytronics Corp.,
which is a member of the Vocal Interface Group consisting of
Cytronics, Interface Systems, and Federal Screw Works. The VS4
is ...")
(syn: "Hello, zero, one, ..., ten. Four, six, seven ... seven,
five, one.")
(Gannon: "Hello, the following ... VS4 ... Mar 20, 1972, ...
several short familiar quotations.")
(syn: "To be or not .. Four score and seven ... the only thing
we have to fear ... Early to bed ... A bird in the hand ... ")
(Gannon: "To illustrate words used in common applications ...")
(syn: "Zero, one, two, ... 100. Credit balance ... climb to
altitude ... barometric pressure at ... in service."; words
at random: "... zero.")
END: **********
SSSHP 9: "SPEECH SYNTHESIS, W.A. AINSWORTH, UNIVERSITY OF KEELE,
17/8/88"
SOURCE: Received 8/88 from Dr. William A. Ainsworth, Dept. of
Comm. and Neuroscience, Univ. of Keele, Keele, Staffordshire,
ST5 5BG, United Kingdom (in SSSHP UK Univ. of Keele file). 7"
reel, 7 1/2 ips, synthesizer has 50 Hz hum at 0 vu level.
CONTENTS:
1. NURSERY RHYMES, ETC., about 1970. System for phonetic spelling
to sound conversion.
(syn: "Hickory, Dickory, Dock. The mouse ran up ... Dock. Jack
and Jill ran up the hill ... after. I have a deep voice."; syn,
2 voices: "Humpty Dumpty sat on a wall ... together again.";
syn: "One, two, three, ... twenty. A, B, C, ... Zed. Friends,
Romans, Countrymen, lend me your ears ... Caesar. Good morning,
Madam, ... so have I.")
2. NORTH WIND AND THE SUN, about 1972. System for text to sound
conversion.
(syn: "The North Wind and the Sun were disputing ... Sun was
the stronger of the two.")
END: **********
SSSHP 15: "TELEPHONE INTERVIEWS - H.D. MAXEY"
SOURCE: H.D. Maxey's interviews with contacts for the Smithsonian
Speech Synthesis History Project. Cassette.
CONTENTS:
DR. ANDRE MALECOT, Dept. of French, Univ. of Calif. at Santa
Barbara, Santa Barbara, CA, 11/16/88. Discussion of work of
Dr. Pierre Delattre, a mentor of Dr. Malecot. The PB2
synthesizer used by Dr. Delattre has been disassembled and lost.
END: **********
SSSHP 17: "TSI SPEECH READING SYSTEM ANNOUNCEMENT (VC 003T),
12/12/77"
SOURCE: Received 1/30/89 from Dr. Gabriel F. Groner, Speech Plus,
Inc. See SSSHP USA Telesensory Systems, Inc. file. Cassette, good
quality.
CONTENTS:
Announcement of development project.
(human, 50 sec: "Computers are getting to sound more human
every day. ... hand-held camera across page (Opticon) ...
and sounds like this.")
(syn, 40 sec: "While the Opticon Tactile Reading Aid developed
by ... at low cost.")
(human: "TSI plans to incorporate ... $10,000 .. late 1979."
(syn, 2:41 min: "Good morning. On behalf of Telesensory
Systems, Inc., ... scientists at MIT ... my human
colleagues."
END: **********
SSSHP 18: "TELESENSORY SYSTEM & SPEECH PLUS COMPILATION OF
1979-86 TAPES, made 9/11/88"
SOURCE: Received 1/30/89 from Dr. Gabriel F. Groner, Speech Plus,
Inc. See SSSHP USA Telesensory Systems, Inc. file. Use right
channel for Dr. Groner's narration; left channel for synthesis.
90-minute cassette is more fragile than 60-minute cassettes.
CONTENTS:
1. TSI READING MACHINE PROTOTYPE, mid 1979. Announcement of the
system, using a prototype synthesizer. Demonstration narrated by
Jeff Moyers. Text was hand-tracked with the Opticon. Tape was made
from microphone at a conference and includes audience sounds and
interviews with the audience.
(syn, 1 sen: "Greetings from Telesensory Systems. As you know
... disabled people.") Use left channel.
(human: "After over three years of ... in August.")
(syn and audience, 11:34 min: ... local field representatives.")
2. PROSE 2000 V1.0, Apr 13, 1982. Conversion of ASCII text to
speech for computer voice output. First production machine.
(syn, 6:58 min: "I am the Prose 2000 Text-to-Speech Converter
... ; variation of controls, poetry, joke; ... provide the
answers.")
3. PROSE 2000 V1.2 and V2.0, May 18, 1983. Comparison between
V1.2 and V2.0, the first major improvement in speech quality.
(syn, several sen, V1.2 and V2.0, 4:14 min: "Here is a
demonstration of the differences ... pull and full.")
4. CALLTEXT 5000 AND PROSE 2000, V3.0 and V3.1, May 22, 1986.
Comparisons of software versions 3.0 and 3.1 over a speaker
(Prose 2000) and over the telephone (CallText 5000). CallText
5000 and Prose 2000 have the same level at the same time of
production. CallText 5000 has an additional hardware circuit
for use over the telephone. V3.0 was the second major
improvement in speech quality, and V3.1 was the first attempt
to adjust the synthesis for the telephone. Order of
presentation is V3.0 over a speaker, V3.0 over a telephone,
V3.1 over a telephone, V3.1 over a speaker, and repeat of V3.0
over a speaker. Good quality, -10vu, use right channel.
(syn, 5 samples, 2:46 min: "How San Jose lost $60,000,000 ...
airport.")
5. CALLTEXT/PROSE, V3.1 vs V3.2, Oct 1986. Version 3.2 had
improvements to the pronunciation of homographs and proper
names. Good quality, -10vu, use right channel.
(syn, V3.1 and V3.2: "He read the documents so the President
could read them to the customers. Please wait a minute so I
can tell you some minute details. He will project that the
last project will be completed")
(syn, 17 proper names, V3.1 and V3.2)
6. CALLTEXT/PROSE MULTILANGUAGE, May 30, 1986. Release of Spanish
and French synthesis for preliminary testing. German synthesis
not quite ready for testing. Good quality, -10vu.
(syn: "This is a demo tape prepared on ... the order is
Spanish, German, French, and English.")
(syn, 4 languages, several phrases each)
END: **********
SSSHP 21: "Xerox/Kurzweil Personal Reader Demonstration Tape"
SOURCE: Received 1/3/89 from Mr. Randall Stern, Kurzweil
Computer Products, Inc. See SSSHP USA Kurzweil Computer Products
file. Cassette, Side A duplicated on Side B. Good quality. Circa
1988.
CONTENTS:
Kurzweil Personal Reader presenting a description of itself, using
the DECtalk synthesizer (see SSSHP USA DEC file.)
(music, syn, 4:30 min: "Now hear this! Kurzweil announces the
latest breakthrough in technology for ... 617-864-4700, in
Massachusetts.")
END: **********
SSSHP 22: "KRM Series 400 Demonstration", 1984.
SOURCE: Received 1/3/89 from Mr. Randall Stern, Kurzweil
Computer Products, Inc. See SSSHP USA Kurzweil Computer Products
file. Cassette. Good quality.
CONTENTS:
Kurzweil Reading Machine, Series 400 model, using Prose 2000
synthesizer (see SSSHP USA Telesensory Systems file). Interview
with Kurzweil product representatives demonstrating the Reading
Machine. Synthesis examples alternate with human speech, with the
synthesizer seeming to respond to the interviewer. Over 400
installations at this time.
(music, human introduction, 1:08 min: "... Why don't we let the
Series 400 explain that?")
(syn, 1 min: "Thank you, David. My operation is quite simple,
... poem by Robert Frost ... miles to go before I sleep.")
(human discussion of use, 1:38 min: ... how it works.")
(syn, 1:25 min:"I read the way everyone else does, I scan
the page ... 4,827.")
(human discussion of use, 1:33 min: "... information about the
Series 400?")
(syn: "They can contact Kurzweil ... 617-864-4700.")
END: **********
SSSHP 26: "Speech Synthesis by Rule, Electrotechnical Laboratory,
1969"
SOURCE: Received 12/20/88 from Dr. R. Teranishi, Khushu
Institute of Design. Flexible 33 1/3 RPM record, 3 copies in
SSSHP JAPAN Electro Technical Laboratory file.
CONTENTS: See SSSHP 39 Cassette.
END: **********
SSSHP 29: "NTT-1, NTT Human Interface Laboratories,
Sadaoki Furui, Jan 13, 1989"
SOURCE: Received 1/13/89 from Dr. Sadaoki Furui, NTT. See SSSHP
JAPAN Nippon Telegraph and Telephone Corp. file. 7" reel.
CONTENTS:
1. PARTIAL AUTOCORRELATION (PARCOR) VOCODER, 1978. English
sentences by PARCOR vocoder. Three data rates each, in the
following order: 9.6, 4.8, and 2.4 kilobit/second. Processed human
speech. Good quality, -10vu.
(human, male and fem: "Isn't it a fine day? Yes, it's
beautiful. Where are you going? To the country. My
grandparents have a farm there. Oh, that's very nice. Where
are you going? I'm going to the country, too. My train leaves
at ten o'clock.")
2. LINE SPECTRUM PAIR (LSP) VOCODER, 1981. English sentences by
PARCOR and LSP vocoders, analog recording and three data rates
each. Order of presentation is all male sentences followed by
all female sentences. Good quality, -10vu. Within each gender,
the order is:
a. Original recording
b. PARCOR vocoder (no quantization of transmitted signals)
c. LSP vocoder (no quantization of transmitted signals)
d. PARCOR vocoder (9.6 kilobit/sec quantization)
e. LSP vocoder (9.6 kilobit/sec quantization)
f. PARCOR vocoder (4.8 kilobit/sec quantization)
g. LSP vocoder (4.8 kilobit/sec quantization)
h. PARCOR vocoder (2.4 kilobit/sec quantization)
i. LSP vocoder (2.4 kilobit/sec quantization)
(human: "The ship was torn apart on the sharp reef. Sickness
kept him home the third week. The box will hold seven gifts at
once. Jazz and swing fans like fast music.")
END: **********
SSSHP 30: "NTT-2, NTT Human Interface Laboratories, Hirokazu Sato
& Sadaoki Furui, Dec 27, 1988"
SOURCE: Received 1/13/89 from Dr. Sadaoki Furui, NTT. See SSSHP
JAPAN Nippon Telegraph and Telephone Corp. file. 7" reel, good
quality.
CONTENTS: Synthesis by Rule from 1966
1. TERMINAL ANALOG - VCV SYNTHESIS, 1968. Simulation of terminal
analog (formant-type) synthesizer with on-line computer.
Synthesis based on vowel-consonant-vowel segments. Synthesized
Japanese words.
(syn: "sakura, kaigi, saiban, amai, issai")
(syn: "karewa sakanawo tsuriagemashita. Karewa inuni esawo
yarimashita. Akiwa kakino miga minorimasu. Umini shiroi namiga
tachimasu.")
2. PARCOR - VCV SYNTHESIS, 1978. Speech synthesis on the basis of
PARCOR VCV concatenation units. Synthesis of female voice.
(syn, fem: "Tsubamega tonda. Medakaga ogawawo oyoide iru.
Hasuno hanawa asa hirakimasu. Ojiisanwa yamae shibakarini
ikimashita. Kinou dokoe ikimashitaka? Kinouwa yamani
noborimashita.")
3. TEXT TO SPEECH CONVERSION, LSP - CVC SYNTHESIS, 1986. Speech
synthesis using consonant-vowel-consonant units and excitation
waveform elements. Synthesis of male voice.
(syn, male: "Shinkansenno tabiwa kaitekideshita. Sumireno
hanaga sakimashita. Jishinno youna toppatsugenshouwo
yochisurukotowa muzukashii. Jishinto iumonoga hontouni
toppatsutekide atte, soreno okoru maeni, zenchoutekina monoga
zenzen nanimo arawarenai toiu monode arunaraba, yochiwa
genritekini fukanoudearu.")
END: **********
SSSHP 31: "TSI Demo, Jared Bernstein, Nov. 11, 1980"
SOURCE: Received from Dr. Jared Bernstein 11/11/80, Telesensory
Systems, Inc., Palo Alto, CA. See SSSHP USA Telesensory Systems,
Inc. file. 4" reel, 7 1/2 ips. (Maxey Tape T80.7)
CONTENTS: Prose 2000 text to speech development system (TTS-X).
(syn, 2:05 min: "The purpose of this recording is to
demonstrate ... Palo Alto, California.")
END: **********
SSSHP 32: "TEXT-TO-SPEECH HISTORY, D. Klatt, ASA Demo, Copy 2,
2/87"
SOURCE: Received from Dennis H. Klatt, MIT, Feb. 1987. Identical
to source tape for 33 1/3-rpm demonstration record to accompany
"Review of Text-to-Speech Conversion for English", D.H. Klatt,
JASA 82.3, Sept. 1987, 737-793. (Online, this site.)
The tape was made directly from Klatt's computer via D/A conversion
of the speech samples that had been previously stored by A/D process
of 10,000 samples/second, 12-bit samples. Low-pass filter was TTE,
Inc. #J97E, H8917, Cutoff Freq.=4800 Hz, stop band of -60 db at 5760
Hz. (Personal communication from Keith North, RLE technician,
October 1990. See SSSHP USA MIT correspondence.)
Cassette, 19 min, dated November 2, 1986. Tape lacks sample (1)
("The leaves had been raked into piles.") of Entry A10. (The
SSSHP tape archives contain copies of many of the source analog
recordings from which these samples were taken. See the
referenced SSSHP outlines.)
CONTENTS: Demonstration speech samples to accompany above JASA
paper. Descriptions are from the paper's appendix.)
APPENDIX: Demonstration
The enclosed 33 1/3-rpm recording contains illustrations of some
of the milestones in the development of systems for text-to-speech
conversion. For convenience in locating and listening to examples
as they are described in the text, it may be desirable to transfer
the recording onto a cassette tape. The assistance of H. David
Maxey, Micheal Hecker, John Holmes, Patrick Nye, Joe Olive, and
James Flanagan in assembling these materials is gratefully
acknowledged. My thanks also go to Kenneth Stevens, who served as
narrator. The record has been inserted inside the back cover of
this issue.
Demonstration of several milestones in the development of systems
for text-to-speech conversion. Each of the following four Parts
represents an evolutionary stage in the technology.
PART A: DEVELOPMENT OF SPEECH SYNTHESIZERS
The objective of early research on speech synthesis was to test
whether the synthesizer design is capable of high-quality
imitations of human voices.
1. THE VODER OF HOMER DUDLEY, 1939. Dudley of AT&T Bell
Laboratories designed a speech synthesizer known as the Voder
(Dudley et al., 1939). It was demonstrated at the 1939 World's
Fair in New York (see SSSHP USA BTL file).
(Radio Announcer: "Will you please make the Voder say for our
Eastern listeners, 'Good evening, radio audience'?")
(syn: "Good evening, radio audience.")
(Radio Announcer: "And now for our Western listeners say,
'Good afternoon, radio audience.'")
(syn: "Good afternoon, radio audience.")
2. THE PATTERN PLAYBACK DESIGNED BY FRANKLIN COOPER, 1951. The
Haskins Laboratories Pattern Playback (Cooper et al., 1951) was
designed to permit converting back into sound the patterns
observed on broadband sound spectrograms (see SSSHP USA Haskins
Laboratories file).
(syn: "These days a chicken leg is a rare dish. It's easy to
tell the depth of a well. Four hours of steady work faced us.
A large size in stockings is hard to sell.")
3. PAT, THE PARAMETRIC ARTIFICIAL TALKER OF WALTER LAWRENCE,
1953." Lawrence (1953) of the Signals Research and Development
Establishment, Christchurch, England, designed the PAT (Parametric
Artificial Talker) parallel formant synthesizer. It was first
demonstrated at a conference in London in 1952 (see SSSHP UK SRDE
file).
(syn: "What did you say before that? Tea or coffee? What have
you done with it?")
4. THE OVE CASCADE FORMANT SYNTHESIZER OF GUNNAR FANT, 1953. Fant
(1953) of the Royal Institute of Technology in Stockholm, Sweden,
designed a cascade formant synthesizer (OVE-I). It was
demonstrated at the same London conference in 1952 (see SSSHP
Sweden RIT file).
(syn: "How are you? I love you.")
5. COPYING A NATURAL SENTENCE USING WALTER LAWRENCE'S PAT
FORMANT SYNTHESIZER, 1962. James Anthony and Walter Lawrence
attempted to match a natural recording using an updated version of
PAT (Anthony and Lawrence, 1962). Demonstrated at the 1962
Stockholm Speech Communication Conference (see SSSHP UK Edinburgh
file). Compare with the OVE II version of the same utterance,
next.
(syn: "Welcome to the Stockholm Speech Communication Seminar.")
6. COPYING THE SAME SENTENCE USING THE SECOND GENERATION OF
GUNNAR FANT'S OVE CASCADE FORMANT SYNTHESIZER (OVE-II), 1962.
Gunnar Fant attempted to match a natural recording using OVE II
(Fant and Martony, 1962). Demonstrated at the 1962 Stockholm
Speech Communication Conference (see SSSHP Sweden RIT file).
Compare with the PAT version of the same utterance, above.
(syn: "Welcome to the Stockholm Speech Communication Seminar.")
7. COMPARISON OF SYNTHESIS AND A NATURAL SENTENCE, USING OVE II,
BY JOHN HOLMES, 1961. Holmes (1961) of the Joint Speech Unit of
the British Post Office used the OVE II synthesizer to generate a
close copy of a natural sentence. See SSSHP UK JSRU file. Klatt
obtained recording from John Holmes, Oct 7, 1986. Copyright
JSRU.
(syn, male: "I enjoy the simple life.")
(human: "I enjoy the simple life.")
(syn, fem: "He knows just what he wants.")
(human: "He knows just what he wants.")
8. COMPARISON OF SYNTHESIS AND A NATURAL SENTENCE, JOHN HOLMES
USING HIS 'PARALLEL' FORMANT SYNTHESIZER, 1973. Holmes did
essentially the same thing in 1973, using a more complex parallel
formant synthesizer of his own design (Holmes, 1973). See SSSHP
UK JSRU file. Demonstrated at the 1972 IEEE Conference on Speech
Communication and Processing, Boston. Klatt obtained recording
from John Holmes, Oct 7, 1986. Copyright by JSRU.
(syn: "I enjoy the simple life, as long as there is plenty of
comfort.")
(original human, low-pass filtered to 4 kHz: "I enjoy the
simple life, as long as there is plenty of comfort.")
9. ATTEMPT TO SCALE THE DECTALK MALE VOICE TO MAKE IT SOUND
FEMALE. The DECTalk "Perfect Paul" male voice has been modified
by scaling fo by a factor of 1.7 (ap=204, pr=170), by scaling all
formant frequencies by a factor of 0.85 (hs=85) and removing the
fifth formant (f5=2500, b5=2048), by increasing the open quotient
of the glottal waveform using the "richness" variable (ri=0), and
by decreasing the output level slightly to avoid overloads
(lo=81). These manipulations are not sufficient to turn Paul into
a convincing female speaker. DECTalk 3.0. See SSSHP USA MIT
file.
(syn: "I am the standard male voice, Perfect Paul. This is the
result of trying to imitate a female voice by increasing the
pitch, reducing the head size, and lengthening the open
quotient.")
10. COMPARISON OF SYNTHESIS AND A NATURAL SENTENCE, FEMALE VOICE,
DENNIS KLATT, 1986. A synthetic copy of a female speaker
producing (1) a sentence and (2) an utterance in which each
syllable of "Steve eats candy cane" is replaced by [?a] is
compared with the original recording (Klatt, 1986b). See SSSHP
USA MIT file.
(1)("The leaves had been raked into piles.") [Ed: missing from
tape]
(2)("?a ?a, ?a ?a ?a. ?a ?a, ?a ?a ?a")
11. THE DAVO ARTICULATORY SYNTHESIZER DEVELOPED BY GEORGE ROSEN
AT MIT, 1958. The DAVO (Dynamic Analog of the VOcal tract)
circuit designed by Rosen (1958) at M.I.T., augmented by a nasal
tract designed by Hecker (1962), was controlled by a tape
recording of control signals created by hand by Kenneth Stevens
and Arthur House. The demonstration occurred at the fall meeting
of the Acoustical Society of America in 1961. See SSSHP USA MIT
file.
(syn: "This is the voice of DAVO at MIT."; song:"A, B, C, ...")
12. SENTENCES PRODUCED BY AN ARTICULATORY MODEL, JAMES FLANAGAN
AND KENZO ISHIZAKA, 1976. Flanagan and Ishizaka (1976) of the
AT&T Bell Telephone Laboratories used an articulatory synthesizer
to generate two sentences, using control data derived from the
Coker et al. (1973) text-to-speech system. A two-mass model of
the vocal cords was employed, and turbulence noise was injected
automatically whenever the Reynolds number became large at the
larynx, or at a constricted section of the vocal tract. See
SSSHP USA BTL file. Klatt obtained recording from J. L. Flanagan,
Nov. 2, 1986.
(syn: "She saw the house. This is the test".)
13. LINEAR PREDICTION ANALYSIS AND RESYNTHESIS OF SPEECH AT A
LOW BIT RATE IN THE TEXAS INSTRUMENTS SPEAK-'N-SPELL TOY, RICHARD
WIGGINS, 1980. Wiggins (1980) designed a low-cost linear
prediction synthesis chip to take advantage of the ability of
linear prediction to represent critical spectral and temporal
aspects of speech waveforms efficiently. See SSSHP USA Texas
Instruments file.
(syn: "Now spell 'one' as in 'one word'. O-N-E. Correct.
Next spell 'earth'. E-R-T-H. That is incorrect. The
correct spelling of 'earth' is E-A-R-T-H.")
14. COMPARISON OF SYNTHESIS AND A NATURAL RECORDING, AUTOMATIC
ANALYSIS-RESYNTHESIS USING MULTIPULSE LINEAR PREDICTION, BISHNU
ATAL, 1982. Atal of the AT&T Bell Laboratories demonstrated a new
formulation of linear prediction, known as multipulse LPC (Atal
and Remde, 1982) at the 1982 Paris ICASSP. See SSSHP USA BTL
file. Klatt obtained recording from Joe Olive on Sept 12, 1986.
(LPC syn, speaker 1: "Where is Dennis sitting?")
(original human: "Where is Dennis sitting?")
(LPC syn, speaker 2: "This field of beets is ripe and ready.")
(original human: "This field of beets is ripe and ready.")
PART B: SEGMENTAL SYNTHESIS BY RULE
The first synthesis-by-rule programs concentrated on the
development of rules for phonemic synthesis, and did not include
rules for the automatic specification of phoneme durations and
fundamental frequency. Since prosody was specified by hand to
match a natural recording, these demonstrations sound
significantly better than they would if all information had been
derived by rule.
15. CREATION OF A SENTENCE FROM RULES IN THE HEAD OF PIERRE
DELATTRE, USING THE HASKINS PATTERN PLAYBACK, 1959. A stylized
spectrogram of the desired sentence was painted on a transparent
plastic plate by Pierre Delattre, and then played by the Haskins
Pattern Playback. See SSSHP USA Haskins file.
(syn: "I painted this by rule without looking at a spectrogram.
Can you understand it?")
16. OUTPUT FROM THE FIRST COMPUTER-BASED PHONEMIC SYNTHESIS
BY RULE PROGRAM, CREATED BY JOHN KELLY AND LOUIS GERSTMAN, 1961.
Kelly and Gerstman (1961, 1962) of the AT&T Bell Laboratories
demonstrated the first phonemic synthesis-by-rule program in 1961
at a meeting of the Acoustical Society of America. See SSSHP USA
BTL file.
(syn: "To be, or not to be. That is the question. Whether it is
nobler in the mind to suffer the slings and arrows of
outrageous fortune or to take arms against a sea of troubles,
and by opposing, end them.")
17. ELEGANT RULE PROGRAM FOR BRITISH ENGLISH BY JOHN HOLMES,
IGNATIUS MATTINGLY AND JOHN SHEARME, 1964. Holmes et al. (1964)
of the Joint Speech Research Unit in England demonstrated an
impressive phonemic synthesis-by-rule program for British English
at the fall meeting of the Acoustical Society of America in Ann
Arbor, 1963. See SSSHP UK JSRU file.
(syn: "A bird in the hand is worth two in the bush. It was the
last thing I expected to find there. Did you come by motor
car? I'm going home now. Someone, somewhere wants a letter
from you.")
18. FORMANT SYNTHESIS USING DIPHONE CONCATENATION, BY REX DIXON
AND DAVID MAXEY, 1968. Dixon and Maxey (1968) of IBM at Research
Triangle Park demonstrated a diphone concatenation method for
construction of control parameter time functions for a formant
synthesizer at the 1967 M.I.T. Conference on Speech Communication
and Processing. See SSSHP USA IBM file.
(syn: "The number you dialed, ME1-5280, has been changed. The
new number is PA6-1347. This is a recording. I'm sorry,
you've reached this office by mistake. Please consult your
directory and dial again.")
19. RULES TO CONTROL A LOW-DIMENSIONALITY ARTICULATORY MODEL, BY
CECIL COKER, 1968. Coker (1968) of AT&T Bell Laboratories created
a method of generating speech from an articulatory model. The
system was demonstrated at the 1967 M.I.T. Conference on Speech
Communication and Processing. See SSSHP USA BTL file. Klatt
obtained recording from Joe Olive on Sept. 12, 1986.
(syn: "This is computer vocal tract speaking. You are
listening to the voice of a machine. The number you have
reached, 464-1078, has been changed.")
PART C: SYNTHESIS BY RULE OF SEGMENTS AND SENTENCE PROSODY
The next synthesis-by-rule programs include a complete set of
rules for going from phonemes, stress marks, and some syntactic
information to an output speech waveform.
20. FIRST PROSODIC SYNTHESIS BY RULE, BY IGNATIUS MATTINGLY,
1968. The synthesis-by-rule program of Mattingly (1966; 1968) of
the Haskins Laboratories was demonstrated to accompany his Ph.D.
thesis. See SSSHP USA Haskins file.
(syn: "You are listening to speech synthesized by rule. It was
made at Haskins Laboratories with a computer. On the computer
typewriter, the experimenter types a phonemic transcription.
The control signals are calculated for the synthesizer and the
synthetic speech can be heard immediately.")
21. SENTENCE-LEVEL PHONOLOGY INCORPORATED IN RULES BY DENNIS
KLATT, 1976. Klatt (1976b) of the M.I.T. Speech Communication
Group created a phonological component to generate segmental
durations and a fundamental frequency contour, as well as
sentence-level allophonic variation, from a phonemic input
augmented with stress and syntactic symbols. See SSSHP USA MIT
file.
(syn: "T'was the night before Christmas, and ... sugar plums
danced in their heads.")
22. CONCATENATION OF LINEAR-PREDICTION DIPHONES, BY JOE OLIVE,
1977. Olive (1977) of AT&T Bell Laboratories controlled a
linear-prediction synthesizer from stored reflection coefficients
for a set of diphones. The system was demonstrated at ICASSP-77.
The recording is from about 1980, and includes prosodic rules
provided by Liberman and Pierrehumbert. See SSSHP USA BTL file.
Klatt obtained recording from Joe Olive on Sept 12, 1986.
(syn: "This paper describes a small real-time speech
synthesizer. The synthesizer requires as its input a string of
phonemes, and the associated duration, pitch, and amplitude
parameters. The synthesis scheme uses a large table of stored
transitions, dyads, between phonemes. These transitions are
stored in terms of LPC-derived area parameters.")
23. CONCATENATION OF LINEAR PREDICTION DEMISYLLABLES, BY CATHRINE
BROWMAN, 1980. A synthesis-by-rule program with prosodic rules,
called Lingua, was designed by Browman (1980) of AT&T Bell
Laboratories, using the demisyllable inventory collected by
Fujimura and Lovins (1978). Demonstrated at ICASSP-80. See SSSHP
USA BTL file.
(syn: "Hello. I am a language interpreter named Lingua. I
have been used to synthesize speech from demisyllables by rule.
I start by breaking your sentences up into the chunks called
demisyllables and then combining them. That's not all I have
to do however. I must also figure out how loud, how long, and
what pitch each part of the sentence should be.")
PART D: FULLY AUTOMATIC TEXT-TO-SPEECH CONVERSION
24. THE FIRST FULL TEXT-TO-SPEECH SYSTEM, DONE IN JAPAN BY
NORIKO UMEDA ET AL, 1968. The first demonstrated text to speech
system for English was created by Umeda et al. (1968) of the
Electrotechnical Laboratory in Japan, and was based on an
articulatory model. It included a syntactic analysis module with
sophisticated heuristics. Demonstrated at the 6th International
Congress on Acoustics, in Tokyo in 1968. See SSSHP Japan ETL
file.
(syn: "One upon a time, there lived a King and Queen who had no
children. Not a day passed but that the Queen did not say, 'If
only we had a child.' One day as the Queen was walking beside
the river, a little fish lifted its head out of the water and
said, 'Dear Queen, your wish shall be fulfilled.'")
25. THE FIRST BELL LABORATORIES TEXT-TO-SPEECH SYSTEM, BY CECIL
COKER, NORIKO UMEDA, AND CATHRINE BROWMAN, 1973. Coker et al.
(1973) of AT&T Bell Laboratories demonstrated a text-to-speech
program based on the Coker (1967) articulatory model. The system
was demonstrated at the 1972 International Conference of Speech
Communication and Processing in Boston. See SSSHP USA BTL file.
(syn: "I can read stories and speak them aloud. I do not
understand what the words mean when I read them. But I can
quess which words are important and which words are not by
rules I have been given. Some day I may be able to provide
many kinds of information by telephone.")
26. THE HASKINS LABORATORIES TEXT-TO-SPEECH SYSTEM, 1973. The
Haskins Laboratories text-to-speech system (Cooper et al., 1973)
used the Mattingly (1968) phoneme-to-speech rules coupled with a
large dictionary. See SSSHP USA Haskins file.
(syn: "Early Research in Animal Communication, by Mark Twain.
Animals talk to each other, of course. There can be no
question about that. But I suppose there are very few people
who can understand them. I never knew but one man who could.
I knew he could, however, because he told me so himself.")
27. THE KURZWEIL READING MACHINE FOR THE BLIND, RAYMOND KURZWEIL,
1976. Kurzweil (1976) began selling a reading machine with an
optical scanner in the late 1970s. The system was demonstrated on
the CBS evening news in about 1976. See SSSHP USA Kurzweil file.
Klatt borrowed videotape from Kurzweil Computer Products on Nov 8,
1986.
(syn: "Hello, I am the Kurzweil Reading Machine. Welcome to
mid-Manhatten Library. I have been placed here so that an
employee or visually-handicapped person ... Please enter a
command.")
28. THE INEXPENSIVE VOTRAX TYPE-N-TALK SYSTEM, BY RICHARD GAGNON,
1978. The Votrax low-cost Type-n-Talk text-to-speech system
combines a single-chip synthesis-by-rule program and formant
synthesizer (Gagnon, 1978) with a version of the Elovitz et al.
(1976) letter-to-sound rules. It was demonstrated at the 1978
ICASSP Conference. See SSSHP USA Votrax file.
(syn: "The juice of lemons makes fine punch. A box was thrown
beside a parked truck.")
29. THE ECHO LOW-COST DIPHONE CONCATENATION SYSTEM, ABOUT 1982.
The Echo low-cost text-to-speech system concatenates linear
prediction diphones using the Texas Instruments's TMS-5220 linear
prediction synthesizer chip. See SSSHP USA TI file.
(syn, 4 sen: "The birch canoe slid on the smooth planks. Glue
the sheet to the dark blue background. ... These days a chicken
leg is a rare dish.")
30. THE M.I.T. MITALK SYSTEM, BY JONATHAN ALLEN, SHERI HUNNICUTT
AND DENNIS KLATT, 1979. The MITalk-79 laboratory text-to-speech
system, developed at the Massachusetts Institute of Technology by
Allen et al. (1979, 1987) and many others. The system was
demonstrated in its final form at the 1979 meeting of the
Acoustical Society of Anerica in Boston. See SSSHP USA MIT file.
(syn: "Speech is so familiar a feature of daily life that we
rarely pause to define it. It seems as natural to men as
walking, and only less so than breathing. Yet it needs but a
moment's reflection to convince us that this naturalness of
speech is but an illusory feeling. The process of acquiring
speech is, in sober fact, an utterly different sort of thing
from the process of learning to walk.")
31. THE MULTI-LANGUAGE INFOVOX SYSTEM, BY ROLF CARLSON, BJORN
GRANSTROM AND SHERI HUNNICUTT, 1982. The Infovox commercial
text-to-speech system (Magnesson et al., 1984) is an
implementation of the Carlson et al. (1982a) multilanguage system
that was developed at the Royal Institute of Technology in
Stockholm by Rolf Carlson et al. Versions of the system were
demonstrated in 1976 and 1982 at ICASSP conferences. See SSSHP
Sweden RIT file.
(syn: "The SA 101 built their own 16-bit microcomputer, the MC
68000, and the signal processor, the MAC 7720. The device can
initially be connected to a normal computer terminal. You can
also connect the SA101 to a computer just like a printer. This
flexible synthesis system makes it possible to use spoken
information in many applications.")
32. THE SPEECH PLUS INC. 'PROSE-2000' COMMERCIAL SYSTEM, 1982. The
Prose-2000 commercial text-to-speech system was first developed in
conjunction with a reading machine for the blind project at
Telesensory Systems by James Bliss and his associates (Goldhor and
Lund, 1983; Groner et al., 1982). Speech synthesis technology
came from the 1976 Klatt synthesis-by-rule software and the
MITalk-77 and MITalk-79 laboratory systems at M.I.T. The
recording is of Version 3.0 of the Prose-2000. See SSSHP USA
Telesensory file.
(syn: "Four hours of steady work faced us. A large size in
stockings is hard to sell. The boy was there when the sun
rose. A rod is used to catch pink salmon.")
33. THE KLATTALK SYSTEM, BY DENNIS KLATT OF M.I.T. WHICH FORMED
THE BASIS FOR DIGITAL EQUIPMENT CORPORATION'S DECTALK COMMERCIAL
SYSTEM, 1983. The Klattalk (1982a) laboratory text-to-speech
system software was licensed to Digital Equipment Corporation as a
basis for the commercial DECtalk text-to-speech system announced
in 1983. The recording is of Version 3.0 of the DECtalk software.
See SSSHP USA MIT and SSSHP USA DEC files. Klatt produced a
waveform file directly from a DECTalk 3.0 machine.
(syn:"Text-to-speech systems are beginning to be applied in
many ways, including aids for the handicapped, medical aids,
and teaching devices. The first kind of aid to be considered
is a talking aid for the vocally handicapped. According to the
American Speech and Hearing Association, there are over one
million people in the United States who are unable to speak for
one reason or another. Any person in this group who can use a
typewriter keyboard, or point at some kind of communication
board, is a potential user of a text-to-speech system.")
34. THE AT&T BELL LABORATORIES TEXT-TO-SPEECH SYSTEM, 1985. A
new AT&T Bell Laboratories laboratory text-to-speech system (Olive
and Liberman, 1985) uses the Olive (1977) diphone synthesis
strategy in combination with a large morpheme dictionary (Coker,
1985) and letter-to-sound rules (Church, 1985). The laboratory
system was demonstrated at a 1985 meeting of the Acoustical
Society of America. See SSSHP USA BTL file. Klatt obtained
recording from Joe Olive on Sept 12, 1986.
(syn: "This paper will give a brief overview of recent
text-to-speech work at Bell Laboratories. Starting about a
year ago, we have completed a new set of computer programs that
translate English text into sound. This system constructs
speech sounds by concatenating elements from an inventory of
about 900 units, stored in terms of multipulse LPC coding.")
35. SEVERAL OF THE DECTALK VOICES. Examples of some of the
voices provided by the DECtalk 3.0 text-to-speech system. See
SSSHP USA DEC file.
(syn: "I am Beautiful Betty, the standard female voice. Some
people think I sound a bit like a man.")
(syn: "I am Huge Harry, a very large person with a deep voice.
I can serve as an authority figure.")
(syn: "My name is Kit the Kid, and I am about ten years old,
and I sound like a boy or a girl.")
(syn: "I am Whispering Wendy and I have a very breathy voice
quality. Can you understand me even though I am whispering?)
36. DECTALK SPEAKING AT ABOUT 300 WORDS/MINUTE. Example of using
the DECtalk 3.0 speaking rate command to skim material at a rapid
rate. The nominal speaking rate has been set to 350 words/min,
[:ra 350], although this 51-word passage took 11 sec to speak,
indicating an effective rate slightly under 300 words/min. See
SSSHP USA DEC file.
(syn: "The following is a list of topics in today's news. In
the sports world the Red Socks lost to Detroit. First round
matches were played in Wimbledon tennis tournament. Arnold
Palmer won the Senior Golf Tourney in La Grove, Penn. In local
news, there was a five alarm fire in Cambridge.")
(human: "End of the demonstration. These recordings were made by
Dennis Klatt on November 22, 1986.")
END: **********
SSSHP 33: "MITALK, TSI COMPARISONS (PISONI TESTS) 1980,
D.B. PISONI, 9/26/80"
SOURCE: Copy made 12/23/82 of Dr. N. R. Dixon's copy, at IBM.
Comparisons and tests of the Telesensory Systems, Inc. TTS-X
development system for the Prose 2000, and the M.I.T. MITalk79
system on which the TSI system was based. Cassette, fair quality,
some print-through echos in some sections, on both left and right
tracks. See SSSHP USA Telesensory Systems, Inc. file.
(MITalk, Northwind Passage, 35 sec: "The North Wind and the
Sun. The North Wind and the Sun were arguing one day ...
stronger than he was.")
(MITalk, Estes Chapter, 59 sec: "It should be pointed out here
that despite the fact that the speech signal may be a poor
quality ... before it is ever produced.")
(MITalk, Modified Rhyme Test, 10 words: "Fat, bean, fun ...
tick")
(MITalk, Harvard Sentences, 10 sen: "The birch canoe slid on
the smooth planks. Glue the ... hard to sell.")
(MITalk, Haskins Sentences, 10 sen: "The roan ... dog caught
the shoe.")
(MITalk and TSI, Lens Buyer Passage, 1:21 min each: "The lens
buyer must approach the problem of purchasing a lens of large
aperture ... is really needed.")
(MITalk and TSI, Noise Pollution Passage, 1:14 min each: "In
ancient Rome, Julius Caesar banned chariot driving ... could
become lethal.")
END: **********
SSSHP 38: "TAPE 4, SLEEPING BEAUTY, SYNTHESIZED BY RULE, ELECTRO-
TECHNICAL LABORATORY, TOKYO, AUGUST, 1968"
SOURCE: Received 3/2/89 from R. Teranishi, Kyushu Institute of
Design. Synthesis of fable, Sleeping Beauty, in English male and
child voices. Good quality, 3" reel, copy of copy of master. See
SSSHP JAPAN Electro Technical Laboratory file.
CONTENTS:
(syn, male, 12 sen: Once upon a time ... King and Queen ...
and fall down dead."; child, 4 sen: "Once upon ... a
daughter.")
END: **********
SSSHP 39: "SPEECH SYNTHESIS BY RULE: ETL (1969)"
SOURCE: Received 3/2/89 from R. Teranishi, Kyushu Institute of
Design. Cassette recording of the outside channel only of the
SSSHP 26 plastic diskette of the Electro Technical Laboratory,
Japan. Better quality than the diskette. See SSSHP JAPAN Electro
Technical Laboratory file. Narrated by Japanese female in
English.
CONTENTS:
1. SYNTHESIS OF STATIC VOWELS, 1966. Mechanical synthesis of
Japanese vowels with a plastic model of the vocal tract. The male
vocal tract was 7" long, the child's was 4" long. Some
printthrough.
(syn, various voices: "a, i, u, e, o")
2. SYNTHESIS OF CONTINUOUS JAPANESE SPEECH BY VOCAL TRACT
SIMULATOR, 1967. Analog computer simulation of a 17-section vocal
tract and fixed nasal tract. Off-line computer control via a
digital tape.
(syn, Japanese story, The Peach Boy (Momotaro no ohanashi), 52
sec)
3. SYNTHESIS OF JAPANESE TONGUE TWISTERS BY VOCAL TRACT
SIMULATOR, 1967.
(syn, four tongue twisters, 3 speaking rates each)
4. TEXT-TO-SPEECH IN ENGLISH USING VOCAL TRACT SIMULATOR MODEL 2,
1968. World's first demonstrated text-to-speech system for
English. Grimm's fairy tale, Sleeping Beauty.
(syn, English male, 4 sen: "Once upon a time ... King and
Queen...have a child, a daughter.")
END: **********
SSSHP 42: "A SPEECH SYNTHESIS SYSTEM BY RULE IN JAPANESE, SAMPLE
OF THE SYNTHESIZED SPEECH, BY R. TERANISHI"
SOURCE: Received 3/2/89 from R. Teranishi, Kyushu Institute of
Design. See SSSHP JAPAN Kyushu Institute of Design file. Cassette,
side A, good quality, but faint echo's from printthough.
CONTENTS: Example output of the Kyushu Institute of Design text to
speech system in 1988. Synthesis of Japanese male voice.
1. READING A STORY, GRAND TOUR OF A RAIN DROP. Whole story is read
in the standard speed, which is 5.6 morae (syllables)/second.
(syn, 3:24 min, Grand Tour of a Rain Drop)
2. READING A STORY IN VARIOUS SPEEDS. Initial five sentences from
Part 1 are read at the following speeds.
(syn, 5.6 morae/second standard speed, 5 sen)
(syn, 6.7 morae/second, 5 sen)
(syn, 8.3 morae/second, 5 sen)
(syn, 11.1 morae/second, 5 sen)
3. SMOOTHED EFFECTS OF CONTROL SIGNALS FOR THE LSP SYNTHESIZER.
The five sentences from Part 1 are repeated at a fast speed, with
the LSP synthesizer control signals being subjected to low-pass
filtering at the following frequencies.
LSP Sequence LSP Sound Source
Channels Channel
Filtering Filtering
(syn, none none )
(syn, none 15 Hz )
(syn, 15 Hz 15 Hz )
(syn, 15 Hz 8 Hz )
(syn, 15 Hz 6 Hz )
(syn, none none )
4. RECITING A POEM, KURODA-BUSHI, IN THREE CLASSES OF SPEED AND
STYLE.
(syn, three speeds and styles, 53 sec)
END: **********
SSSHP 44: "NTT-3, SADAOKI FURUI (NTT), Feb. 15, 1989"
SOURCE: Received 2/16/89 from Dr. Sadaoki Furui, NTT. See SSSHP
JAPAN Nippon Telegraph and Telephone Corp. file. 7" reel, good
quality.
CONTENTS: Itakura and Saito's demonstration tape for 6th ICA in
Tokyo, maximum likelihood method vocoder, 1968. A ten-sentence
dialog between an American male and a Japanese female, both
speaking English. The tape has an abrupt end on the first two
dialogs.
(syn, original human, 10 sen: "Hello, how are you? ...
accomplishment.")
(syn, vocoder output, 10 sen: "Hello, how are you? ...
accomplishment.")
(syn, vocoder output, expanded time, 10 sen: "Hello, how are
you? ... accomplishment.")
END: **********
SSSHP 45 "TEXT-TO-SPEECH IN JAPANESE (STANDARD READING),
R. TERANISHI, 10/17/85"
SOURCE: Received April, 1986, from R. Teranishi, Kyushu
Institute of Design. See SSSHP JAPAN Kyushu Institute of Design
file.
CONTENTS: Japanese text-to-speech system at the Kyushu Institute
of Design. Cassette, both sides, some printthrough echos, playback
amplitude changes during recording.
Side A:
(syn, example at standard reading rate, 3:15 min)
Side B:
(syn, example at quick speaking rate, 1:45 min)
END: **********
SSSHP 47: "GENESIS OF SPEECH SYNTHESIS IN JAPAN, RADIO RESEARCH
LABORATORIES, MPT (COMMUNICATIONS RES. LAB.), 1959-1960,
Edited by Jouji Suzuki, CRL, 2-Mar-1989".
SOURCE: Received 3/20/89 from Dr. Jouji Suzuki, Communications
Research Laboratory, Japan. See SSSHP JAPAN Radio Research
Laboratory file. 5" reel, good quality, -10vu narration, +5vu
synthesis.
CONTENTS: Terminal analog speech synthesizer controlled by a
multichannel function generator. Function generator created nine
trapezoidal control voltages, each with eight sequential
breakpoints adjustable in amplitude and time. Synthesizer was used
to study the acoustic cues of Japanese phonemes and syllables.
1. STIMULI TO MAKE F1-F2 CONTOUR OF JAPANESE VOWELS
(syn, 14 vowels)
2. SAMPLES OF FRICATIVE FOLLOWED BY VOWEL
(syn, 7 CVs)
3. SAMPLES OF SEMIVOWEL FOLLOWED BY VOWEL
(syn, 12 CVs)
4. CHANGE IN PITCH. Japanese for "Yes".
(syn, 5 pitch levels: "Hai hai.")
5. CHANGE IN INTONATION. Japanese for "Good morning" and "Yes, it
is."
(syn, 6 intonation patterns: "Ohajou.")
(syn, 8 intonation patterns: "Eh soodesu.")
6. CHANGE IN PITCH AND SPEAKING RATE. Japanese for "space
communication".
(syn, 6 patterns: "Uchuu-tsuusin.")
7. SEVERAL WORDS AND A PHRASE. Japanese for "Ministry of Posts and
Telecommunications", "radio wave", "research", "speech", and "I
don't."
(syn, x3: "Juusei-shou, dempa, kenkjuu, onsei. I jadesu.")
END: **********
|