U.S. patent number 3,676,595 [Application Number 05/030,123] was granted by the patent office on 1972-07-11 for voiced sound display.
This patent grant is currently assigned to Research Corporation. Invention is credited to Ladislav Dolansky, Nathan D. Phillips.
United States Patent |
3,676,595 |
Dolansky , et al. |
July 11, 1972 |
VOICED SOUND DISPLAY
Abstract
Voiced sounds are displayed by means of a pitch period
indicating circuit including circuits for the suppression of those
errors which arise from a minimal number of parameters. Automatic
sequencing is provided in a compact setting to provide equipment
ideally suited for instruction of young students.
Inventors: |
Dolansky; Ladislav (Weston,
MA), Phillips; Nathan D. (Stoneham, MA) |
Assignee: |
Research Corporation (New York,
NY)
|
Family
ID: |
21852630 |
Appl.
No.: |
05/030,123 |
Filed: |
April 20, 1970 |
Current U.S.
Class: |
704/276;
704/E11.006 |
Current CPC
Class: |
G09B
19/04 (20130101); G10L 25/90 (20130101) |
Current International
Class: |
G09B
19/04 (20060101); G10L 11/04 (20060101); G10L
11/00 (20060101); G10l 001/14 () |
Field of
Search: |
;179/1VS,1SA ;324/77B
;35/35C |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
F Anderson, An Experimental Pitch Indicator for Training Deaf
Scholars, J.A.S.A. vol. 32, 8/1960 p. 1065-1074.
|
Primary Examiner: Cooper; William C.
Assistant Examiner: Leaheey; Jon Bradford
Claims
We claim:
1. Apparatus for displaying patterns derived from voiced sounds
comprising in combination means for producing a sequence of pulses
whose frequency is that of the pitch of the voiced sound to be
displayed; a cathode-ray variable-persistence display having gating
means to block or deliver the display on the cathode-ray screen, a
beam-sweep means adapted to start the sweep across the display in
response to the start of the voiced utterance, and means for
deflecting the beam in a direction transverse to the direction of
sweep at a rate slow enough to display pitch variation within the
utterances; a P-A Analog Circuit for producing an analog voltage
the magnitude whereof is proportional to the logarithm of the
frequency of said pulses; means for applying said voltage to said
beam-deflecting means of said cathode-ray display; and means for
producing a control signal to actuate said gating means to block
the display until a certain time interval after completion of a
sweep.
2. Apparatus according to claim 1, wherein said means for producing
a signal also includes means to produce a signal blocking said
pulse whenever the time interval between this pulse and the
immediately preceding pulse corresponds to a pulse repetition rate
greater than a certain fixed rate.
3. Apparatus according to claim 1, wherein said means for producing
a signal also includes means to produce a signal blocking said
display whenever the time interval between immediately successive
pulses is more than a certain time interval.
4. Apparatus according to claim 1, wherein said
variable-persistence display includes means for storing said
display, and wherein said means for producing a control signal
includes means to actuate said gating means, after said gating
means has blocked the display, to deliver the display during at
least one additional sweep of said beam and thereafter to block the
display again until a certain time interval after completion of
said additional sweep.
5. Apparatus for displaying patterns derived from voiced sounds
comprising in combination a means for producing a sequence of
pulses whose frequency is that of the pitch of the voiced sound to
be displayed; cathode-ray variable-persistence display having
gating means to block or deliver the display on the cathode-ray
screen, a beam sweep means adapted to start the sweep across the
display in response to the start of the voiced utterance, and means
for deflecting the beam in a direction transverse to the direction
of sweep at a rate slow enough to display pitch variation within
the utterances; an exponential-decay circuit for producing a
voltage output which decays exponentially from a fixed voltage to
which each pulse restores it, so that a sample of said output at a
fixed time interval prior to each succeeding pulse is a measure of
the logarithm of the pulse frequency; means for storing said sample
until production of the following sample, thereby producing an
analog voltage the magnitude whereof is proportional to the
logarithm of the frequency of said pulses; means for applying said
voltage to said beam-deflecting means of said cathode-ray display;
and means for producing a control signal to actuate said gating
means to block the display until a certain time interval after
completion of a sweep.
6. Apparatus according to claim 5, wherein said means for producing
a signal comprises means to compare each sample with the stored
preceding sample, and means to produce a signal whenever the
samples thus compared differ by more than a predetermined
amount.
7. Apparatus according to claim 5, wherein said exponential-decay
circuit comprises two RC circuits each adapted to receive pulses at
the junction of its resistance and capacitance, said resistances
being joined at the respective ends thereof remote from said
junction so as to deliver a current between pulses which is the sum
of the discharge currents from the capacitances.
8. Apparatus according to claim 7, wherein said exponential decay
circuit comprises a first RC circuit and a second RC circuit,
wherein the RC constant of said first RC circuit is 0.0113 sec.,
wherein the RC constant of said second RC circuit is 0.0019 sec.,
and wherein the magnitude of said second resistance is twice that
of said first resistance.
Description
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to cathode ray display equipment and in
particular to circuitry for automatic display of patterns dervied
from voiced utterances.
2. Prior Art
The invention relates to the visual display of voiced utterances
and related sounds.
Previous pertinent developments fall into two problem areas: (A)
pitch-signal extraction, (B) visual display.
A. PITCH EXTRACTION
Several methods of extracting a meaningful pitch signal have been
developed in the recent decades. Of these, the signal envelope
detection type appears to be most suited for the present
application: accordingly, the previous developments in
pitch-extraction are limited to the most essential developments in
this area.
In 1949, Gruenz and Schott described a method for extraction and
portrayal of pitch of speech sounds. (See Gruenz, O.O., Jr., and
Schott, L.O. "Extraction and Portrayal of Pitch of Speech Sounds,"
B.T.L. Mono B-1698. Also: J. Acous. Soc. Amer. Sept. 1949). They
claimed reliable pitch extraction for frequencies from 100 to 600
Hz. In order to reduce errors in producing the basic pitch-period
indicating pulses, an input filter was used to limit the pass-band
for the speech signals.
The need to obtain pitch-period indicating pulses with almost no
delay precluded the use of such filters and led to the development
of the "instantaneous pitch-period indicator (IPPI)." (See
Dolansky, L. "An Instantaneous Pitch-Period Indicator," J. Acous.
Soc. Amer., Vol. 27, No. 1, pp. 67-72 (1955). Anderson, who wanted
to use a pitch indicator for deaf children, used the basic circuit
of the IPPI but incorporated a voltage-holding circuit, used
between pitch-indicating pulses, and also included a logarithmic
transformation, which made a more desirable display possible. (See
Anderson, F. "An Experimental Pitch Indicator for Training Deaf
Scholars," J. Acous, Soc. Amer., Vol. 32, No. 8, pp. 1,065-1,074
Aug. 1960). In Sweden, Tjernlund also developed a similar pitch
extractor based on the envelope-detection process. (See Tjernlund,
P.A. "A Pitch Extractor with larynx Pick-up," Quarterly Progress
and Status Report, Speech Transmission Laboratory, Royal Institute
of Technology, Stockholm, Sweden, pp. 32-34 (Oct. 15, 1964). Other
developments of pitch-extraction methods and devices are not
considered to be essential for inclusion in this brief summary of
pertinent developments. However, an excellent summary of such
efforts is to be found in McKinney's report on laryngeal frequency
analysis, and shorter summaries elsewhere. (See McKinney, N.P.
Laryngeal Frequency Analysis for Linguistic Research, Report No.
14, Contract Nonr 1224(22), N R 049/122, The University of
Michigan, Ann Arbor, Mich. (September, 1965); Dolansky, L. Aids for
the Deaf and Some Related Problems, Final Report, VRS and USEC
Fellowship Grant, (PL 87/256, Fulbright-Hayes Act), Northeastern
University, Boston, Mass. (Sept. 30, 1967); Dolansky, L. and
Manley, H. Speech Analysis (A Survey Report), Sylvania Elec. Sys.,
Applied Res. Lab., Waltham, Mass. Proj. 72-401 (Oct. 1, 1960).
B. VISUAL DISPLAY
From the point of view of practical usefulness, when an intonation
display is used in any kind of teaching, it is essential to be able
to keep the display as long as desired, and to discard it instantly
when it is desired to continue with the next task. During the early
attempts only long-persistence crt displays and photographic
reproduction were available. (See Gruenz, supra) The former did not
have sufficient duration to permit, say, a comparison of the
student's and the teacher's pattern. Photography was not a
satisfactory solution for a rapid sequence of teaching tasks.
Later, attempts were made to use a mechanically rotating
long-persistance crt tube. (See Anderson, supra; Loos, R. Ein
Tonhoheschreiber fur die Sprecherziehung von Gehorlosen
Diplomarbeit, Inst. f. Elektrotchnik der Johannes Gutenberg
Universitat in Mainz (September, 1965). However, control of the
pattern retention time was still not available.
In 1965, with the storage oscilloscope then available, the
Northeastern University speech research group embarked on an
extended research effort in the area of teaching intonation to the
deaf. (See Dolansky, L., Ferullo, R. J., O'Donnell, M.C. and
Phillips, N.D. Teaching of Intonation and Inflections to the Deaf,
Final Report, Cooperative Research Project No. S-281, Northeastern
University (1965); Dolansky, L., Karis, C., Phillips, N and
Pronovost, W. Teaching Vocal Pitch Patterns Using Visual Feedback
from the Instantaneous Pitch-period Indicator for Self-monitoring,
Part I. VRA Project No. 1907-S, Northeastern University (Dec. 31,
1966); Dolansky, L. and Phillips, N.D. Teaching Vocal Pitch
Patterns Using Visual Feedback from the Instantaneous Pitch-period
Indicator for Self-monitoring, Final Report, Part II, VRA Project
No. 1907-S, Northeastern University (Oct. 31, 1966); Dolansky, L.,
Pronovost, W.L., Anderson, D.C., Bass, S.D. and Phillips, N.D.
Teaching of Intonation Patterns to the Deaf Using the Instantaneous
Pitch-period Indicator, Final Report, VRA Grant 2360-S,
Northeastern University (Feb. 28, 1969); Phillips, N.D., Remillard,
W. J., Bass, S. and Pronovost, W. "Teaching of Intonation to the
Deaf by Visual Pattern Matching," American Annals of the Deaf, Vol.
113, No. 2, pp. 239-346 (Mar., 1968). The essential instrument in
this study was the IPPI.
SUMMARY
The invention makes use of visual display techniques utilizing the
fundamental frequency of the signal under study (e.g. the
microphone signal for voiced utterances of speech). In accordance
with the invention a visual pattern which represents the
fundamental frequency of the signal under study as a function of
time is displayed on the face of a storage oscilloscope. Because
the intonation pattern is presented as a function of time, temporal
aspects of the signal can be studied. In addition, the intensity of
the signal is displayed by means of lights. The invention
comprehends a complete intensity display and intonation display
system capable of producing and storing intonation patterns for
extended periods of time and suitable for the quantitative study
and teaching of intonation patterns as well as for other frequency
versus time displays of variable frequency signals.
One aspect of the invention includes an error suppression scheme
based on (a) the total permissible frequency range and (b) the
permissible rate of change of the fundamental frequency of the
signal.
For signals that are as complex and varied as the speech signals,
it is virtually impossible to guarantee that occasional errors will
not occur. The occuring errors are usually in the form of an
extraneous or missing pitch-period indicating pulse, and result in
momentary displays of twice or one-half of the correct frequency.
Since considerable knowledge is available about the total range of
the pitch frequency and the rate of change of this frequency, it is
possible to devise schemes which would prevent such erroneous
indications from actually appearing on the screen in most
cases.
Another aspect of the invention comprehends a specific improved
method of conversion of the frequency describing voltage which
approximates the desired logarithmic conversion to a high degree of
accuracy.
Still another aspect of the invention comprehends a system of
automatic and semiautomatic modes of operation permitting a great
flexibility in the desired procedures for the study and teaching of
intonation patterns.
Since it is highly desirable to develop adequate voice volume
before it is attempted to obtain meaningful intonation patterns, an
intensity display in the form of four lights has been incorporated
into one embodiment of the invention. Thus it is possible to use
the equipment for (a) an intensity display, using the lights, (b)
an intonation contour display, using the storage-oscilloscope
screen display, (c) both displays simultaneously.
When it is used as an intonation-contour display, it may be desired
to produce patterns continuously, and have them fade at a constant
rate, or it may be desirable to produce either one or two patterns,
keep them unchanged for observation for a brief time, and then have
the patterns automatically erased by the built-in sequencing
program. Alternatively, it may be desired to perform repeated
attempts, and subsequently erase all patterns by actuating a
push-button switch. While the system of the invention can actually
produce two intonation patterns (one from either of its complete
pitch-extraction circuits) at the same time, in most cases it is
desirable to produce only one pattern at a time (e.g. the
teacher's) and when the circuits become ready to process another
signal, to so indicate and at the same time unblock the other input
channel, so that the new pattern can be properly produced and lined
up against the first pattern.
A few of the possible applications of this apparatus are given
below:
A. Auditory Feedback Replacement
Especially in the case of profoundly deaf persons, the absence of
the usual auditory feedback makes it difficult for these
individuals to acquire natural intonation in their speech. If,
however, the visual intonation pattern is available to them, they
can learn to imitate the teacher's intonation pattern in which case
their intonation sounds also quite natural to other persons. This
use was the primary reason for developing this equipment, and
encouraging results have been obtained in the experimental research
with previous, less developed models.
B. Speech Correction
In attempts of correcting speech defects in speech-correction
clinics, the system of the invention may be quite helpful if the
corrective process is concerned with pitch, intonation, intensity,
or duration aspects of the utterances.
C. Teaching of Music (Voice) (voice)
Such aspects as vibrato and other attributes of pitch can quite
clearly be seen and studied in a quantitative way, both with
respect to frequency and duration.
D. Teaching of Languages
It is often difficult to teach proper intonation in a foreign
language. The system of the invention is a welcome aid to show
differences in intonation in a most straightforward and
quantitative way. This may be of special importance in languages in
which the meaning is changed when the intonation pattern
changes.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention may best be understood from the following detailed
description thereof, having reference to the accompanying drawings,
in which:
FIG. 1 is a front view of the external physical layout of a system
embodying the invention;
FIG. 2 is a graph in which the pitch frequency is plotted against
time;
FIG. 3 is a graph showing the pulsed output of the pitch detector
and illustrating regions of error suppression;
FIG. 4 is a block diagram of a circuit embodying the invention;
FIG. 5 is a circuit diagram showing the preamplifier, mixer and
pitch detector used in the system of the invention;
FIG. 6 is a circuit diagram of the pulse rate to analogous log
pitch frequency converter used in the system of the invention;
FIG. 7 is a circuit diagram showing the logarithmic conversion
circuit and the sample and hold circuits of the system of the
invention;
FIG. 8 is a circuit diagram of the .DELTA.T comparator of the
system of the invention;
FIG. 9 is a circuit diagram of the tri-level indicator or frequency
range comparator of the system of the invention;
FIG. 10 is a multiple graph showing the wave forms of the various
portions of the P-A analog circuit;
FIG. 11 is a circuit diagram showing the horizontal sweep control
and presentation sequencer of the system of the invention;
FIG. 12 is a circuit diagram showing a trigger generator for use
with the horizontal sweep control and presentation sequencer of
FIG. 11;
FIG. 13 is a circuit diagram of a sweep generator suitable for use
with the horizontal sweep control and presentation sequencer of
FIG. 11;
FIG. 14 is a multiple graph showing the timing diagram for delay
and sequencing logic in the MFRT mode as employed with the
horizontal sweep control and presentation sequencer of FIG. 11;
and
FIG. 15 is a circuit diagram showing a clock and channel switch
suitable for use in the system of the invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring to the drawings, and first to FIG. 1 thereof, a display
screen 1 displays an intonation pattern representing the pitch
frequency of a human voice which it is desired to analyze. Input
signals from a microphone, into which the person whose voice is to
be analyzed speaks, are connected either to microphone jacks 2 or
to the auxiliary input terminals 3 via an external pre-amplifier.
The auxiliary input terminal may be used to receive signals from
the output of a tape recorder rather than from the output of a
microphone pre-amplifier. An earphone jack 4 is provided to provide
an output for an earphone. Two habitual pitch level controls 5, one
for each channel, make it possible to shift the intonation pattern
on the screen 1 up and down as needed. If it is desired to replace
the ordinary intonation pattern by a double line pattern indicating
the limits of acceptability of the response, the control for this
purpose may most conveniently be provided by an auxiliary plug-in
trace width control 6 which can be connected to the side of the
oscilloscope unit. A loudness display represented by four light
bulbs 7,8,9 and 10 is activated by depressing a loudness switch 11.
The pitch display 1 may be activated by depressing a pitch control
switch 12. A mode selector switch 13 provides means for selecting
the various modes of operation. A reset button 14 is provided for
erasing the intonation pattern on the screen and where applicable
to return the circuits into the initial state preceding a new
display sequence. While this reset button 14 is a single switch it
includes two illuminated captions 15 and 16; 15 being labeled
"START" and 16 being labeled "CONT." (i.e., "continue"). A mirror
17 permits monitoring the subject's facial movements as well as
those of the teacher. The mirror 17 may be adjusted in such a way
that the face of the teacher, the face of the student, and the
display screen 1 can all be seen by the student without moving his
head. A power switch 18 is used to switch the unit on and off.
In accordance with the invention a system of automatic and
semiautomatic modes of operation are provided so as to permit a
great flexibility in the desired procedures for the study and
teaching of intonation patterns. A representative list of modes are
listed in Table I.
TABLE I
Modes of Pitch Pattern Presentation
Mode CFFT MFFT MFST MFRT MFSD
__________________________________________________________________________
Contin- Persis- uously Minimal Min. Min. Min. tence Fading Fading
Fading Fading Fading Trigger of Free Free Seq. Rptd. Simul. Display
Trig- Trig- Trg. Trg. displayed gering gering Sequence of Events
Trig- gering 1 or 2 1 or 2 2 2 1 or 2 Channel Trig- Trig- Channel
gering gering 2** 2** 1 and 2 Displayed Channel Channel Fades
Storage Contin- 4 sec. 2 sec. 2 sec. Variable Time* uously Automat-
Man., Erasure ically -- *** go to 1 If man, go to 1, otherwise
continue Trig- gering 1 or 2 1 or 2 Channel Channel Trig. Trig.
Displayed Channel Channel Storage 4 sec. 1 sec. Time* Erasure
automatic *** If man, go to 1, otherwise go to 5 Ready for new
sequence
__________________________________________________________________________
referring to FIG. 1 and to Table I, operation in any of the listed
modes presumes that the "PITCH" switch has been activated, so that
the light under the pushbutton of the switch 12 is on. When the
"START" light 15 is on, the cathode-ray beam is on the left side of
the screen 1 - the equipment is ready to receive and display a
pattern. As soon as a sound signal of sufficient intensity is
received by the equipment, the sweep is triggered, the "START"
light 15 is turned off, and the continue light "CONT." 16 is turned
on. The latter stays on for the duration of the sweep, even if the
speech signal ceases; in that case, however, the beam is
suppressed; during the moments of silence, no trace is visible on
the screen. The sequence of events for a particular mode of
operation which is set by the "MODES" switch 13 in FIG. 1, is best
understood by following the pertinent column in Table I. For
example, when operating in Mode MFST, the sweep is triggered by the
speech signal of Channel 2 (usually used by the teacher), and
subsequently the pattern derived from this speech signal is
displayed on the screen 1 (event 2). After the pattern on the
screen 1 is completed, the equipment is not ready to receive any
additional pattern for a period of 2 seconds during which neither
the "START" light 15 nor the "CONT." light 16 is activated (event
3). This time interval of 2 seconds is usually needed to study the
teacher's pattern in preparation for an imitation attempt, whenever
he is ready (event 5). The pattern derived from the student's
speech signal is then displayed on the screen 1 (event 6), and both
patterns remain displayed for a period of 4 seconds (event 7).
Subsequently, both patterns are automatically erased, and the
equipment is ready for a new sequence as soon as the "START" light
15 comes on.
The various modes used in the operation of the equipment are
intended to serve the following purposes (the meaning of the mode
designations is evident from Table I):
Cfft: useful in situations where the subject is monitoring
continuous speech.
Mfft: useful when it is desired to display and study a complete
phrase pattern or sentence.
Mfst: suitable for automated programming, or rigidly sequenced
lessons.
Mfrt: this mode can also be used for automated programming;
however, a greater flexibility in the number of responses per
stimulus is possible.
Mfsd: useful when the teacher and the subject practice
simultaneously, which is sometimes desirable for greater
motivation. This mode can also be used in place of MFFT but in this
case a variable length of time to study the patterns is available
to the subject.
Since the speech volume of many subjects is too low, and also
because it is often desirable to show to the subject the difference
between intensity and intonation, an intensity display is also
provided. This may consist of the four lights 7,8,9 and 10. With
increasing intensity of the speech signal, more lights are
progressively turned on. Thus, for example, in order to operate the
equipment the intensity should generally be at least sufficient to
turn on light number 7. Somewhat greater intensity will turn on
both lights 7 and 8; still greater intensity will turn on lights
7,8 and 9 and still greater intensity will turn on all the lights
7,8,9 and 10. This intensity display may be used either separately
or simultaneously with intonation display since it may be
separately activated independently of the activation of the
intonation display by means of the loudness switch marked LOUD and
indicated by reference numeral 11.
Although the invention not limited to any particular type of pitch
detector design, in a preferred embodiment of the invention a pitch
detector is employed of a type which disregards the future wave
form of speech and bases the process of establishing the value of
the present pitch frequency on the past values of the wave form
alone. Use of this type of pitch detector permits simple equipment
and avoids excessive requirements for storage elements.
In accordance with the invention an error suppression scheme is
provided based on (a) the total permissible frequency range and (b)
permissible rate of change of the fundamental frequency of the
signal. It is necessary to have such a system for eliminating
erroneous pitch period indications, owing to the very nature of the
manner in which pitch periods are identified. The pitch period
detector must operate from a signal which has a relatively complex
wave form, but which has certain regularly recurring
characteristics which permit the establishment of a so-called pitch
frequency. However, the sound possesses certain random
characteristics which give rise to occasional erroneous
indications. The elimination in a simple manner of erroneous pitch
period indications is therefore of great importance. It has been
established that certain pitch periods cannot occur from the sound
of the human voice and the invention makes use of these facts in
establishing the error elimination circuit.
The first parameter used by the invention is the pitch frequency
range. Many estimates have been made of the total range of pitch
frequencies. These estimates vary greatly and frequencies as low as
33 (creaking) and as high as 3.1 kilohertz have been reported.
However, in accordance with the invention the system is designed to
accommodate only the frequency range of normal speech plus any
additional range which might occur in the speech of the deaf. This
frequency range is approximately 70 to 600 hertz (i.e. cycles per
second).
The second parameter utilized by the invention to detect errors is
the maximum rate of change of pitch frequency. This phenomenon may
perhaps best be understood by reference to FIG. 2 wherein an
example of the variation of pitch frequency plotted against time is
shown. It is seen that in this case the maximum rate of change of
pitch frequency occurred at the beginning. In general, this maximum
rate of change of pitch frequency cannot exceed a certain value due
to the physical limitations of the speech producing mechanism.
Therefore, in apparatus constructed in accordance with the
invention, a rate of change of pitch frequency in excess of .+-.10
to 20 percent of the reference frequency is regarded as an error
and is eliminated.
The system embodying the invention includes a logic design
activated by any signal outside the limited range of the
abovementioned parameters. An understanding of this operation may
be had by looking at FIG. 3 wherein f(R) designates reference pitch
frequency; T(HF) represents the period of highest possible pitch
frequency; T(LF) represents the period of lowest possible pitch
frequency: T(IR) represents the period of pitch frequency obtained
with the highest possible positive rate of change, when starting
from the pitch frequency of the preceding pitch period; and T(DR)
represents the period of pitch frequency obtained with the highest
possible negative rate of change, when starting from the pitch
frequency of the preceding pitch period.
Referring now to FIG. 3, therein is shown a sequence of pulses,
identifying the beginnings of pitch periods in time. The first two
pulses define a pitch period of a certain duration T =1/ f(R). The
logical decisions to be made about the acceptability of the next
pitch-period indicating pulse are based on the duration of the
first pitch period. In developing the mechanism for these logical
decisions, certain quantities should first be defined as
follows:
T(HF) is the pitch-period duration for the highest possible pitch
frequency (e.g. 600 hertz), while T(LF) corresponds to the lowest
possible pitch frequency (e.g. 70 hertz). A pitch-period indicating
pulse which occurs before T(HF) is terminated must be in error, and
is therefore always suppressed. The next condition that a new
pitch-period indicating pulse must meet in order not to be
disqualified as an erroneous pulse is that it must occur within the
interval .DELTA.T. As can be seen from FIG. 3, .DELTA.T is the
interval between T(IR) and T(DR). T(IR) and T(DR) are related to
the maximum permissible rate of change of the pitch frequency (e.g.
+10 to 20 percent an -10 to 20 percent respectively) and to the
pitch frequency corresponding to the just preceding (acceptable)
pitch period as defined in FIG. 3. This test of acceptability is
applied only if an acceptable pitch period preceded immediately.
The limits of acceptability for the rate of change were obtained
from a previous study in which an effort was made to measure the
maximum rate of change of pitch frequency that the subjects were
capable of performing. If a pitch-period indicating pulse does not
occur during the entire duration of T(LF), it is concluded that
phonation has ceased and the next pitch-period indicating pulse
will be considered to be the beginning of a new voiced
utterance.
If an expected pitch-period indicating pulse does not occur within
the interval .DELTA.T, the visual display is suppressed until a
sequence of three acceptable pitch periods develops. If, however,
an expected indicating pulse does not occur within the interval
.DELTA.T or between T(DR) and T(LF), the voiced utterance is
considered to be terminated, and the circuit settings return to
their initial state, suitable for the beginning of a new voiced
utterance.
Since various randomly occurring noises are usually present before
the beginning of a voiced utterance, and since such noises give
rise to erroneous pitch-period indicating pulses, the invention
also comprehends means for eliminating a very short part of the
initial portion of the intonation display in order to avoid
confusion. Occasionally, the interfering noise pulses cause the
pitch indicator to indicate a frequency which is one-half of the
correct frequency (i.e. skipping every other pitch-indicating
pulse). In order to prevent the perpetuation of this incorrect
indication throughout the utterance, a special circuit searches for
the correct pitch-indicating frequency, starting from the highest
permissible frequency.
The operation of the pitch/intensity display system is best
explained with the help of the block diagram shown in FIG. 4. The
microphone signals are amplified by means of either of the two
microphone pre-amplifiers 19,20 and fed to the corresponding mixers
21,22 which can also accept an auxiliary input signal, for example
from a tape recorder. Each of these two mixers 21,22 feeds a
separate pitch detection and pulse-to-analog (P-A) circuit 23,24.
The outputs of these two P-A circuits 23,24 are fed to a vertical
amplifier 25 via a common switch 26. The output of the vertical
amplifier 25 is used as the vertical deflection voltage for a
cathode-ray tube. The horizontal deflection voltage for the
cathode-ray tube is generated in the sweep generator 27, and
amplified in the horizontal amplifier 28. The mode switch 29,
indicated by dashed lines in the diagram of FIG. 4, is used to
obtain the various sequencing and display operations of the system
by adjusting the parameters in the blocks indicated by arrows on
the dashed lines. The outputs of the mixers 21,22 of channel 1 and
channel 2 are used to activate a trigger 30 which generates trigger
signals which activate a delay and sequencer circuit 31 and a
switch control 32 for the channel-selecting switch 26. In the MFSD
mode the switch 26 is under the control of a clock 33.
Referring now to FIG. 5, therein is shown a preferred circuit for
use as the pre-amplifier 19 (or 20), mixer 21 (or 22), and pitch
detector (see 23 and 24) shown in FIG. 4. This
preamplifier-pitch-detector circuit includes a microphone
preamplifier 34. In addition, another input, e.g. from a tape
recorder, can be fed into the terminal "AUX. INPUT", identified by
the reference numeral 35. Any of these two input signals pass
through an amplifier-mixer stage 36 and a 4 kHz low-pass filter 37.
This filter 37 is included to reduce the occurrence of erroneous
indications for the speech sounds [z], [ ]; these speech sounds
have a considerable high-frequency content. The filter output is
fed into the cascaded pitch-period incidating stages 38,39, which
operate in a manner similar to the stages of the Instantaneous
Pitch-period Indicator of Dolansky described in the publication "An
Instantaneous Pitch-period Indicator," Journal of the Acoustical
Society of America, Volume 27, No. 1, pp. 67-72 (1955) by L.O.
Dolansky. The output pulse at 40 serves as the input signal to the
pulse-analog circuit of FIG. 6.
Referring now to FIG. 6, therein is shown a
pulse-rate-to-analogous-log-pitch-frequency converter, suitable for
use as the P-A circuit shown at 23 and 24 in FIG. 4. The pitch
detector circuit previously described in connection with FIG. 5
generates a pulse output wherein pulses occur at a frequency
corresponding to the pitch frequency of the voice being analyzed.
This sequence of pulses then forms an input at 41 to the converter
shown in FIG. 6. During those periods wherein the logic circuits of
the converter of FIG. 6 permit complete measurement and display of
the pulse frequency the converter of FIG. 6 operates to deliver an
output at a PA analog output terminal 42 which output is a voltage
signal the magnitude whereof is proportional to the logarithm of
the input pulse frequency. In accordance with the invention the
system includes specific improved method of conversion of the
frequency-describing voltage which approximates the desired
logarithmic conversion to a high degree of accuracy. This
particular aspect of the invention is shown in FIG. 7 in detail and
is shown in FIG. 6 as the current switch 43, RC circuits 44 and
summation amplifier 45. These three items convert the incoming
pulses to a wave form consisting of a sequence of recurring peaks
corresponding to the pulses but followed by an exponential decay.
The magnitude of this voltage signal from the summation amplifier
45 just before each pulse is proportional to a high degree of
accuracy to the logarithm of the pulse frequency. It can be seen
readily that for a relatively long pitch period the exponential
decay curve will reach a relatively low value corresponding to a
relatively low frequency, whereas at higher frequencies the pitch
period will be shorter and the amount of decay will be less,
thereby generating a relatively higher voltage magnitude. A bridge
switch 46 is provided as a sample circuit to select only that
portion of the output of the summation amplifier 46 which exists
just prior to the succeeding pulse. The bridge switch 46 delivers
the sampled output to a follower 47 having a capacitor 48 (FIG. 7)
which holds the signal sampled by the bridge switch 46 until the
next signal issues therefrom. Thus, the PA analog output 42
steadily delivers a voltage corresponding to the pitch frequency
represented by the period between the two latest pulses received by
the apparatus, and this voltage remains the same until a new pulse
comes in thereby establishing a new inter-pulse period.
A gate 49 is provided between the pulse input 41 and the
logarithmic conversion circuits 43,44,45,46 so that erroneous
pulses may be suppressed. In addition, multivibrators 50,51 are
provided for converting the actual pulse input into cleaner pulses
which are easier to handle in the logic circuits. The pulses
produced by the multivibrators 50,51 are consistently of the same
duration and amplitude. The pulse in the first multivibrator 50 is
initiated at the time of the initiation of the incoming pulse
whereas the pulse of the second multivibrator 51 is produced
simultaneously with the conclusion of the pulse issued by the first
multivibrator 50. In the discussion which follows, these pulses may
be designated as an early pulse and a late pulse. The duration of
these pulses is approximately 30 microseconds while the time
interval between pulses is of the order of 10 milliseconds. The
logic circuit of FIG. 6 uses DC logic, so that events occur
depending upon whether the various logic elements are in one state
or another.
All gates are of the NAND type. The not excited condition is
represented by the numeral 0 and corresponds to a voltage of minus
12.6 volts; the excited state is represented by the numeral 1 and
represents a voltage of zero volts. The gate 49 encountered by the
incoming pulse is designed to prevent passage of the pulse if a
signal for this purpose is delivered either by the frequency range
comparator 52 or the rate of change comparator 53. The output from
this gate 49 will be 0 if all its inputs are 1 and this output of 0
is changed by an inverter 54 to a signal of 1 which serves to
trigger the multivibrators 50,51. The states of the multivibrators
50,51 when the output of the gate 49 is 1 are such that the bridge
switch 46 does not sample the output of the summing amplifier 45
and no input is delivered to the current switches 43; and this is
the case if any of the inputs to the gate 49 are 0. Consequently,
if no pulse comes in, or if the frequency range comparator 52
delivers a 0 to the gate 49, or if the rate of change comparator 53
delivers a logarithmic conversion to the gate 49, there is no input
to the converter of circuits 43, 44, 45, 46. In this way, erroneous
signals lying outside the frequency range specified or representing
a greater rate of change of pitch frequency than that specified
will temporarily cut off further incoming signals, and other
portions of the circuit of FIG. 6, to be described hereinafter,
will suppress the display or the screen 1.
The lower portion of FIG. 6 represents the logic circuit which
suppresses the display, in the event of an erroneous input signal,
in response to a control signal generated by the frequency range
comparator 52 or the rate of change comparator 53, and also at the
start of the utterance. However, at the start of the utterance, the
display is additionally suppressed by a delay circuit forming part
of the follower 47. The function of the frequency range comparator
52 and rate of change comparator 53 is to detect an input signal
outside the permissible range, which input signal is presumptively
erroneous, and in response to that input signal to prevent further
excitation of the display for the next three input pulses or so.
The frequency range comparator 52 delivers its signal to the
suppression circuit via a multivibrator 55 which produces a clean
signal pulse from the relatively unclean pulse out of the frequency
range comparator 52. In a similar fashion the relatively unclean
output from the rate of change comparator 53 is delivered to the
suppression circuit via multivibrator 56. Whether or not these
signal pulses will have an effect in the suppression circuit is
determined by whether or not they occur simultaneously with the
input pulses. For this purpose the early pulse is delivered to
certain portions of the suppression circuit and the late pulse is
delivered to other portions of the suppression circuit.
The operation of the P-A circuit of FIG. 6 will now be explained
with reference to FIGS. 6,7,8,9 and 10. The pitch period indicating
pulse from the pitch detector enters gate input terminal 57 via
diode 58. Diode 58 is only required in the event certain
conventional gating circuits are used. Initially gate input
terminal 59 is 1, it being assumed that the output terminal 60 of
control-gate 61 is 1, which assumption is true if either input 62
and 63 of control-gate 61 is 0. Control-gate input 63 is the same
as inverter-output 63, and this is shown as item g in FIG. 10, from
which it appears that inverter-output 63 (and hence control-gate
input 63) is 0 when the incoming pulse starts. The incoming pulse
gets through the gate 49 is inverted once in the gate 49 and again
in the inverter 54, so that it enters the first multivibrator 50 as
a state 1 trigger pulse. When the input to the first multivibrator
50 thus goes to state 1, one output 64 of the first multivibrator
50 delivers a pulse of one polarity to the bridge switch 46 and to
an input 65 of a gate 66, while the corresponding pulse of opposite
polarity is delivered from another output 67 of the first
multivibrator 50 to the bridge switch 46 and also to the second
multivibrator 51. Thus the pulse of the second multivibrator 51
starts when the pulse of the first multivibrator 50 stops. The
pulse of one polarity of the second multivibrator 51 is delivered
to the current switches 43 while the pulse of opposite polarity is
delivered to the suppression circuit so as to determine whether or
not certain activity therein will take place.
So far, we have traced the incoming pulse to the production of a
pair of early pulses and a pair of late pulses all of which are
well shaped. One of the late pulses goes into the current switches
43 which then activates the logarithmic signal portion of the
invention. This part of the invention is shown in more detail in
FIG. 7. The late pulse arrived at input 68 and turns on two current
switches 69,70 each of which charges an RC circuit 71,72 with a
different time constant. The exponential voltages are added in
summing amplifier 45, the output of which is fed to a diode bridge
switch 73,74,75,76. The output of the summing amplifier 45 is shown
in FIG. 10d from which it is seen that the late pulse charges the
RC circuits 71,72 to maximum potential and then the potential of
these circuits declines exponentially giving the output signal of
the summing amplifier 45. As each subsequent pulse from the second
multivibrator 51 arrives the voltage output of the summing
amplifier 45 returns to its maximum value and each time it decays
over the same exponential path, so that the lowest value achieved
by the exponential pattern is a measure of the distance between
pulses. Now the pulse which charges the RC circuits 71,72 is the
late pulse, while the pulse which causes the diode bridge switch
73,74,75,76 to take a sample is the early pulse. The diode bridge
switch 73,74,75,76 thus acts as a sampler of that portion of the
exponential curve which occurs just prior to the next pulse. The
sampled output of the switch, 73,74,75,76 which is a logarithmic
representation of the pitch frequency, is fed to a holding circuit
77. The holding circuit 77 is of conventional design and includes a
capacitor 48 which receives the sampled output and a follower 47
which has a very high input impedance; the sampled output signal
causes this capacitor 48 to be charged to a voltage virtually
identical to the voltage of the output of the summing amplifier 46
at the moment of sampling. The follower 47 delivers a corresponding
voltage signal of the same magnitude and this is the PA analog
output at 42.
Associated with the follower 47 is a circuit which is used, during
the initial portion of a voiced utterance, to bypass the holding
capacitor 48 and thereby cause the apparatus to deliver the same
type of signal as would be delivered from an excessive rate of
change signal. Such a signal has the effect of preventing display
and thus display is prevented during the first several pulses of
any voiced utterance. At the start of any voiced utterance, the
frequency range comparator 52 will be delivering the negative
signal characteristic of an excessively long pitch period since the
time since the last pulse will have been essentially infinite. See
FIG. 10e. This means that under starting conditions the transistor
78 is on so as to connect the holding capacitor 48 to a
(-12.6)-volt voltage source 79 via a 5.1 kilo-ohm resistor 80. This
prevents the holding capacitor 48 from acting as such and the
charge between samples will leak off through this resistor 80
thereby insuring an excessive rate of change signal from the rate
of change comparator 53. After the voiced utterance commences,
however, the frequency range comparator 52 no longer delivers a
negative signal, so that gradually the charge on the capacitor 81
of an RC load on the collector of a transistor 82 leaks off until
it attains a value sufficient to cut off the transistor 78. The
result is shown in FIG. 10f wherein it is seen that the first pulse
delivered to the holding capacitor 48 leaks off and the same is
true of the next two pulses. Thereafter, as the transistor 78
reaches cutoff the holding capacitor 48 becomes able to serve its
function of preserving the sampled voltage until the next
pulse.
RC circuit 71 comprises a capacitor C.sub.2 which is charged by the
late pulse and which discharges between pulses through a resistor
R.sub.2. RC circuit 72 comprises a capacitor C.sub.1 which is also
charged by the late pulse and which discharges between pulses
through a resistor R.sub.1. The resistors R.sub.1, R.sub.2 are
connected at the ends thereof remote from the respective capacitors
C.sub.1, C.sub.2 so that the current discharges from the capacitors
through the resistors are added and their sum delivered to the
summing amplifier 45. In accordance with the invention these RC
circuits 71,72 must be so designed that R.sub.1 C.sub.1 is
approximately 0.0113 sec., R.sub.2 C.sub.2 is approximately 0.0019
sec., and R.sub.2 is approximately 2R.sub.1 for circuits which are
to display voiced utterances. In a representative circuit R.sub.1
might be 50 kilo-ohms, R.sub.2 might be 100 kilo-ohms, C.sub.1
might be 0.226 microfarads, and C.sub.2 might be 0.019
microfarads.
The .DELTA.T comparator 53 used to compare the durations of
consecutive pitch periods is shown in detail in FIG. 8. The output
from the summing amplifier 45 and the signal at the P-A analog
output 42 are compared in a difference amplifier 83. When the
absolute value of the voltage difference between these two outputs
exceeds 0.3 volts the output of the difference amplifier 53 is -12
volts. This result comes about through the use of the diode bridge
84,85,86,87. At the peak of the summing amplifier pulse the voltage
output of the summing amplifier 45 will be about -12 volts. Let it
be assumed that the voltage at the P-A Analog Output 42 (which
represents the sample at the end of the previous summing amplifier
pulse) is at -3 volts. This being the state of affairs the left
hand terminal 88 of the bridge 84,85,86,87 is at a lower voltage
than that of the right hand terminal 89 of the bridge 84,85,86,87
and therefore current flows through diode 86 from the right hand
terminal 89 through a resistance 90 to one input 91 of the
difference amplifier 83, thus creating a voltage at this point of
about -3 volts. The potential of the other input 92 to the
amplifier 83 will correspond to the -12 volts at the left hand
terminal 88, which is transmitted through the diode 85 to that
input 92. Thus the absolute value of the voltage difference exceeds
0.3 volts and the output of amplifier 83 is -12 volts. Used as the
input to the interface circuit 93,94 it causes the output of the
latter to go to -12.6 volts (i.e. state 0), as follows: The
-12-volt output of amplifier 83 causes transistor 93 to be
conductive and hence the collector of transistor 94 is grounded
which causes transistor 94 to become conductive and puts a pulse of
-12.6 volts at the output 95 of the interface circuit, which is the
input to inverter 96. This pulse corresponds to state 0.
As the output of the summing amplifier 45 approaches the voltage at
the P-A Analog Output 42 it eventually comes sufficiently close
thereto that the diodes 84,85,86,87 in the diode bridge do not
conduct; thus the voltage difference between the input 92 and 91
becomes negligible. This causes the output of difference amplifier
83 to rise from its value of -12 volts to approximately zero volts.
When it reaches -6 volts this causes the base of transistor 93 to
have a potential equal to that of its emitter, which is maintained
at -6 volts by a Zener diode 97. At this point transistor 93
becomes non-conducting; the base of transistor 94 goes to a
potential of -12.6 volts and transistor 94 becomes non-conducting
so that state 1 results at the output 95.
In order to reduce extraneous pulses which might arise due to minor
noise present in the signal and the large amplification present in
the circuit, transistor 93 is cut off as the output voltage of
amplifier 83 rises through -6 volts, and this cutoff of transistor
93 is aided by a positive feedback via inverter output 64 which
feeds a signal through an RC circuit 98 and a diode 99. This
positive feedback causes the emitter of transistor 93 to go
negative, thereby supplementing the action of its base in becoming
more positive.
Thus the .DELTA.T comparator 53 causes state 1 to appear at
.DELTA.T output 95 during that period of time when the summing
amplifier 45 has a voltage output in the vicinity of the sampled
output of the preceding pulse. The circuit is so designed that if
the next late pulse does not exist at the time when the state of
.DELTA.T output 95 reverts to 0, a reset pulse or error signal will
be delivered. It can be seen that the time interval during which
.DELTA.T output 95 is in state 1 is longer when a longer period
between input pulses is involved, so that the size of the time
interval within which the next input pulse can exist while .DELTA.T
output 95 is in state 1 (and hence exist without causing a reset
pulse), is roughly a percentage of the length of the preceding
period. Previous studies have shown that this percentage may be in
the vicinity of .+-. 10 to 20 percent, and for the circuits shown
this result is achieved by setting the difference amplifier 83 so
as to react to input voltage differences greater than 0.3
volts.
A tri-level indicator is used to exclude pulses which would
correspond to pitch frequencies outside of the permissible range.
This tri-level indicator of T(LF)/T(HF) comparator is the frequency
range comparator 52 of FIG. 6 and is shown in detail in FIG. 9.
Referring thereto, the limits of this range are determined by the
settings of two 1-kilo-ohm adjustable resistors 100,101. The
voltage at the anode of a diode 102 connected to one adjustable
resistor 100 determines the magnitude of T(LF), while the voltage
at the cathode of another diode 103 connected to the other
adjustable resistor 101 determines the magnitude of T(HF). (See
FIG. 3). The cathode of diode 102 is connected to the anode of
diode 103 by a pair of equal resistors 104,105 the junction 106 of
which is connected to one input 107 of a difference amplifier 108.
As long as the input signal is between the two permissible limits
the voltage at imput 107 is equal to the mean of the two voltages
determined by the setting of the 1-kilo-ohm resistors. The voltage
of the other input 109 of the difference amplifier 108 is also
adjusted to the same value by the potentiometer 110. Thus, under
these conditions, the output 111 of the difference amplifier 108 is
equal to zero volts. In other words, there is a current path from a
-12.6 volt source 112 through the two 1-kilo-ohm resistors 100,101
to ground. Diodes 102 and 103 provide an additional path in
parallel with the one just mentioned through the two series
connected resistors 104,105. As indicated, this provides zero volts
at the output 111 of the difference amplifier 108. As long as the
output of the summing amplifier 45 is between the two voltages
mentioned, diodes 113 and 114 block the output from the summing
amplifier 45 from having any effect. As soon as the output from the
summing amplifier 45 exceeds either of these voltages, a signal is
delivered from the difference amplifier 108. If the output voltage
from the summing amplifier 45 is below the cathode potential of
diode 103, diode 114 conducts and a positive output is observed at
amplifier output 111.
Thus, as the magnitude of the output of the summing amplifier 45
falls after each pulse, any subsequent pulse will be rejected if it
occurs before the magnitude of the summing amplifier output has
fallen below the high voltage end of this permissible range which
might be, for example, a (negative) voltage having a magnitude of 8
volts. Again, if the period is so long that the output of the
summing amplifier 45 falls below the low voltage end of this
permissible range before the subsequent pulse is delivered, an
error signal will also result. This might occur, for example, if
the output of the summing amplifier were a (negative) voltage
having a magnitude below 1 volt. This of course always occurs after
the end of the voiced utterance.
Saturation of the differential amplifier 108 causes sluggish
subsequent response. In order to permit a rapid response,
saturation of amplifier 108 is avoided by the use of a diode bridge
limiting circuit 115. This insures that amplifier 108 stays within
the region of its linear operation.
In order to provide the appropriate voltage for the input 116 of
the logic gate 49, an interface circuit 117,118 is used. The
collector voltage of 118 is at ground potential (stage 1) as long
as the voltage at the amplifier output 111 does not exceed +3
volts. For larger voltages, transistor 118 saturates and the input
116 to gate 49 goes to state 0. As appears from the diagram of FIG.
6, this signal only occurs when the low frequency end of the range
is exceeded such as at the end of the voiced utterance.
The operation of the remaining parts of the pulse-to-analog
converter will now be explained with the help of FIGS. 6 and 10,
and Table II.
TABLE II
---------------------------------------------------------------------------
Truth Table for Gate Control Storage
FF1 FF2 FF3 Clock Com- Pulse ments * 121 211 212 213 214 215 216
217 218 219 220 In- itial 0 1 0 0 0 0 1 0 1 0 0 state b 1 0 1 0 0 0
0 1 0 0 1 0 c 1 1 1 0 0 0 0 1 0 1 0 b 2 1 1 1 1 0 0 0 0 0 1 0 ** c
2 1 1 1 1 0 1 0 0 0 0 b 3 1 1 1 1 0 0 0 0 ** 1 0 c 3 1 1 1 1 0 1 0
0 1 0 0 b 4 1 1 1 1 0 0 0 0 0 1 0 ** c 4 1 1 1 1 0 1 0 0 1 0 0 b 5
1 1 1 1 0 0 0 0 0 1 0 ** c 5 1 1 1 1 0 1 0 0 1 0 0 b 6 1 1 1 1 0 0
0 0 0 1 0 ** c 6 1 1 1 1 0 1 0 0 1 0 0 b 7 1 1 0 0 0 1 `1 0 1 0 0 c
7 1 0 0 0 0 1 1 0 1 0 1 Dis- play b 8 1 0 1 1 0 1 0 1 1 0 1 on c 8
1 0 1 1 0 1 0 1 1 0 1 No change b(N-1) 1 0 1 1 0 1 0 1 1 0 1 c(N-1)
1 0 1 1 0 1 0 1 1 0 1 Dis- play ** 1 0 0 0 0 0 1 0 0 1 1 off b N 1
0 0 0 0 0 1 1 0 1 1 c N 1 1 0 0 0 0 1 0 0 1 0 b(N+1) 1 1 1 1 0 0 0
0 0 1 0 c(N+1) 1 1 1 1 0 1 0 0 1 0 0 b(N+2) 1 1 0 0 0 1 1 0 1 0 0
c(N+2) 1 0 0 0 0 1 1 0 1 0 1 Dis play b(N+3) 1 0 1 1 0 1 1 1 1 0 1
on c(N+3) 1 0 1 1 0 1 1 1 1 0 1 No change b M 1 0 1 1 0 1 1 1 1 0 1
c M 1 0 1 1 0 1 1 1 1 0 1 Dis- play ** 1 0 0 0 0 0 1 0 0 1 1 off
*** 0 1 0 0 0 0 `1 0 0 1 0
__________________________________________________________________________
these circuits are used to suppress the visual display for a brief
period at the beginning of an utterance, or when permissible values
of certain parameters are exceeded, in which case a brief
interruption in the visual display occurs until conditions are
normal again. In addition, the effect of the limitation in terms of
the rate of change of the pitch frequency (as manifested by the
output of the .DELTA.T comparator 53) is delayed. A specific
example of operation will now be explained with the help of the
waveforms of FIG. 10.
The clock pulses for the logic are the early and late pulses
generated respectively by means of the first and second
multivibrators 50,51 and are presented in FIG. 10b,c. First the
mechanism of the initial delay in the display (at the beginning of
an utterance) will be described.
The first late pulse (FIG. 10c-1) generates a decreasing
exponential at the output of the summing amplifier (FIG. 10d).
Assuming that the following pulses occur at a frequency f.sub.min
or greater, the exponentials which restart at each pulse will never
reach the lower dashed line in FIG. 10d; thus the T(LF)/T(HF)
output will be 0v or higher (FIG. 10e). Consequently the transistor
78 in FIG. 7 exponentially approaches pinch off. As a result, in
about 40 msec, a staircase waveform results at the P-A analog
output (FIG. 10f). During this 40 msec interval, reset pulses (FIG.
10h) are generated at inverter output 119.
As described hereinabove, .DELTA.T comparator 53 causes the
interface output 95 to be in state 0 when an improper amount of
time has elapsed since the last pulse. This corresponds to state 1
at inverter output 64, and it is inverter output 64 which is
designated ".DELTA.T output" at FIG. 10g. However, .DELTA.T
comparator 53 does this by causing .DELTA.T output to go to state 0
whenever the summing amplifier output is near the P-A analog
output. Initially, the summing amplifier output is near the P-A
analog output only during delivery of each pulse to the holding
capacitor: the P-A analog output then rises and passes through the
voltage being put out by the summing amplifier. However, during
this initial period the .DELTA.T output reverts to state 1 after
the next late pulse has finished. Hence the output of the second
multivibrator 51 which is delivered to the .DELTA.T error gate 120
is in state 1 when the output of the .DELTA.T error multivibrator
56 which is delivered to the .DELTA.T error gate 120 reverts to
state 1. This results in a pulse to state 0 at the output of
.DELTA.T error gate 120 for the duration of the output pulse of
multivibrator 56. This causes a similar momentary transition of
inverter output 119 to state 0, which constitutes the reset
pulse.
These reset pulses delay the progression of the input 121 to the
first flip-flop 122, as it progresses through the first, second,
and third flip-flops 122,123, 124. For example (see Table II),
state 1 does not remain permanently at the output of the second
flip-flop 123 until pulse 6. Similarly the output of the third
flip-flop 124 does not remain in state 1 until pulse 7. Therefore,
the pulses occuring after pulse 6 are affected by the .DELTA.T
limitation. Similarly, the display begins to be presented after
pulse 7 (see Table II, column 220 ). An example of a suppression of
an erroneous indication in the middle of an utterance is presented
in the center part of FIG. 10. The circumstances shown are typical
of the situation when the magnitude of the permissible rate of
change of pitch frequency is temporarily exceeded. As a result of
this excessive rate, a reset pulse (FIG. 10h) occurs at inverter
output 119 and the .DELTA.T comparator 53 and the visual display 1
are disabled. If the following pulses are within the permissible
range, the .DELTA.T comparator 53 returns to its normal operating
state at pulse (N+1) (FIG. 10i), and the visual display reappears
on the screen at the time of pulse (N+3) (FIG. 10j). The net result
is that the visual display is interrupted for a time no longer than
four pitch periods but no erroneous indication appears on the
screen.
At the end of a voiced utterance .DELTA.T may be exceeded. In that
case, the first and second flip-flops 122,123 (FIG. 6) are reset
(see also FIG. 10i). However, whether .DELTA.T is exceeded or not,
eventually T(LF) is exceeded. At this time, the entire logic is
reset and thus becomes ready for a new utterance.
The horizontal sweep control 125 and presentation sequencer 31 are
shown in FIG. 11. One function of this circuit is to start the
horizontal sweep across the display upon receipt of a trigger
initiated by the start of the voiced utterance. The gates
126,127,128 determine whether or not trigger signals initiated by
the voiced utterance will be operative to start the sweep. It will
be recalled that one of the principal differences between the
various modes is whether signals from channel 1 or channel 2 will
be operative and when. Consequently, the mode switch 29 is also
shown in FIG. 11. Depending upon the position of the mode switch 29
the ability of a trigger from channel 1 or channel 2 to initiate a
sweep is controlled.
A first flip-flop 129 will always respond to an incoming trigger so
as to initiate a horizontal sweep. The signal to the first
flip-flop 129 which initiates the sweep has the appearance of state
0 at the output of the final gate 128. This can occur only if both
inputs to the final gate 128 are in state 1. Thus, in order for
state 0 to appear at the output of the final gate 128, the input at
130 must be 1 for all signals. The intermediate gate 127 acts
simply as an inverter, since it will have its output in state 1 if
either input to the intermediate gate 127 is in state 0, and this
occurs for any trigger coming from channel 2 and will also occur
for a trigger coming from channel 1 provided the input 131 to the
channel-1 gate 126 is in state 1. Thus, when it is desired to block
channel 1, input 131 to channel-1 gate 126 is in state 0. This
never happens in the CFFT or MFFT modes, as is apparent from an
inspection of the pertinent position of switch SAA, but it is the
case at the beginning of the cycle in the MFST, and MFRT modes.
The function of the second flip-flop 132 is to prevent initiation
of a second trace until it permits this to happen. It accomplishes
this by delivering a pulse at its output 133 which is transmitted
through switch SBB to the input 130 of the final gate 128. The
incoming signal which causes the second flip-flop 132 to again
permit a trigger to initiate a sweep arrives at the second
flip-flop 132 from switch SCB. The time of this signal is caused to
be a certain number of seconds after substantial completion of the
sweep by means of the clocking mechanism. When the sweep is
substantially completed an interface 134 delivers a signal to a
first multivibrator 135 which then produces a shaped pulse that is
delivered to a clock 136 via the switch SDA. The clock 136 then
initiates operation of various flip-flops 137,138,139 whose outputs
occur at varying times after receipt of the input signal. These
timed outputs of flip-flops 137,138,139 are also delivered to the
elements at the left-hand side of FIG. 11 some of which serve to
open channel 1 and some of which serve to cause automatic erasing
of the display.
The details of the circuit for producing the trigger inputs to the
sequencer of FIG. 11 are shown in FIG. 12. The circuit for the
sweep generator of FIG. 11 is shown in FIG. 13.
The operation of the horizontal trigger and sweep circuits is as
follows. Referring now to FIG. 12, with the input terminal 140 at
ground potential, transistor 141 is just at the cutoff point; i.e.
the base of this transistor is at approximately -6v. This
adjustment is made by means of the potentiometer 142 in the input
connection. In the presence of an audio signal, its positive values
make this transistor 141 conduct with the result that transistor
143 is turned on. The envelope-detection circuit 144 derives a
voltage from the output of transistor 143. The resulting voltage
turns transistor 145 on. Thus in the presence of an audio signal,
inverter output 146 is in state 1. Similarly, if an audio signal is
present at the "Channel 2 Audio" terminal, inverter output 147 is
in state 1. When an audio signal is present in either or both of
the inputs, gate output 148 is in state 1 and transistor 149 is
turned on. When an intensity display is not desired, transistor 149
is disabled by opening SW3.
The purpose of the sweep generator (FIG. 13) is to produce the
required sweep voltage for the visual display. The sweep-control
signal (in the form of the complement of waveform "CONT.", shown in
FIG. 14) is obtained from the first flip-flop 129 (FIG. 11). When
output 150 of the flip-flop 129 is in state 1, transistor 151 (FIG.
13) is on, and the voltage across the 25 - .mu.f capacitor 152 is
kept at 0v. With the input voltage at 150 at -12v, transistor 151
is cut off and the capacitor 152 charges with a constant charging
current. The magnitude of this charging is determined by the
voltage across transistor 153 and the setting of adjustable
50-kilo-ohm resistor 154. Stages 155,156 and 157 are used to
isolate the load from the sweep-voltage-producing capacitor 152, to
improve the linearity of the sweep, and to provide a low-impedance
sweep-voltage source. A voltage derived from the sweep output by
means of a Zener diode-resistance divider 134 (interface output) is
used to reset a number of voltages in the logic circuits so that
conditions required for the next sweep are obtained. This occurs
when the "interface output"158 reaches about -6v.
The operation of the horizontal sweep control and presentation
sequencer (FIG. 11) will be explained for the MFRT mode, used as an
example. In the beginning of the sequence, the input 131 to the
channel-1 gate 126 is in state 0 and the input 130 to the final
gate 128 is in state 1. Thus channel 1 input 159 is blocked from
having any effect and the "START" light 15 is on. With a trigger in
channel 2 the input to inverter 160 is in state 1 and this causes
the output of the final gate 128 to be in state 0. This 0 signal
sets the first flip-flop 129 and resets the second flip-flop 132.
As a result, one output 161 of the first flip-flop 129 becomes
state 1, thereby turning on the "CONT." light 16 and delivering an
unblank output. In addition, the other output 150 of the first
flip-flop 129 goes to state 0, thereby initiating the sweep and
delivering a pulse to the clock 136 at the input 162 to gate 163.
In addition, the output 133 of the second flip-flop 132 causes the
input 130 to the final gate 128 to become state 0, thus turning off
the "START" lamp 15 and preventing further inputs in either channel
from triggering the sweep: in other words, the final gate 128 is
closed. In the CFFT mode this action of the output 133 of the
second flip-flop 132 is delivered directly to the final gate 128.
In all other modes, this output 133 of the second flip-flop 132 is
delivered to the final gate 128 via gate 164 and inverter 165, Gate
164 also inverts the output of multivibrator 189.
When the interface output 158 to the first multivibrator 135 (see
also FIG. 13) increases in magnitude above -6 volts, the resulting
output at 166 from the first multivibrator 135 resets the first
flip-flop 129. The resulting change in state at output 161 of the
first flip-flop 129 turns the "CONT." light 16 off, and the
resulting change in state at output 150 of the first flip-flop 129
causes the beam to return to its initial position. In addition, the
other output 167 of the first multivibrator 135 via switch SDA
triggers the clock 136 which generates a delay sequence. In the
MFRT mode the first delay equals 2 seconds and is caused by the
output of the fifth flip-flop 138 while subsequent delays are 1
second and are produced by the output of the sixth flip-flop 139.
The original rate of the clock pulses are shown in FIG. 14 and are
represented by the state of the clock output at 168. This output
changes state every half second so that the full cycle is one
second. This rate is reduced by a factor of 2 in each of the three
cascaded flip-flop stages numbered 6,5 and 4 (139,138 and 137). The
outputs of these flip-flops or binary counters are used to set the
second flip-flop 132 via the switch SCB.
In the first part of the MFRT mode input 169 to gate 170 is in
state 0 and input 171 to gate 172 is in state 1. Under these
conditions when output 173 of the fifth flip-flop 138 goes to state
0 (which occurs 2 seconds after the termination of the sweep), the
output of gate 172 goes over to state 1 and the output of gate 174
goes over to state 0. This change of state of the output of gate
174 from state 1 to state 0 is delivered through switch SCB where
said change of state is differentiated in the set differentiator
175 to produce the setting trigger for the second flip-flop 132 at
176. The effect of setting the second flip-flop 132 is to cause the
output of the second flip-flop at 133 to revert to state 1 which,
via SBB and SW2, causes the "START" light 15 to go on through the
mechanism explained previously and opens the final gate 128. In
addition, the change of state of the output 133 of the second
flip-flop 132 from 0 to 1, now opens the channel-1 gate 126 as
follows: output 177 of the third flip-flop 178, which output has
the same state as input 131 to channel-1 gate 126, has been in
state 0 so that input 179 to gate 180 has been in state 1.
Consequently, when output 133 of the second flip-flop 132 goes to
state 1, input 181 to gate 180 goes to state 1 and the output of
gate 180 goes to state 0. This delivers an input to the third
flip-flop 178 causing it to change state and its output 177 goes to
state 1. Switch SAA transmits this state 1 to input 131 to
channel-1 gate 126, thus opening the circuits to input signals from
either channel for the rest of the MFRT sequence. The other output
182 of the third flip-flop 178 now goes to state 0. This signal is
transmitted to input 171 of gate 172, while the change of the other
output 177 of the third flip-flop 178 from state 0 to state 1 is
transmitted to input 169 of gate 170. The effect of this is to
cause the delay between sweeps to be controlled now by the sixth
flip-flop circuit 139 whose output 183 for this purpose is
delivered via switch SCA to input 184 of gate 170. Note that when
switch SCA is in the MFST mode a similar route is followed from
output 185 of the fourth flip-flop 137 which provides a 4 second
delay.
In order to obtain erasure of all patterns the "START-CONT." switch
SW1 is depressed manually so that input 186 to gate 187 goes to
state 0 and the output of gate 180 goes to state 0, and this state
0 is transmitted to the toggle 188 of the third flip-flop 178.
Consequently, if the third flip-flop 178 changes state, its output
182 will go to state 1. This signal reaches the second
multivibrator 189 via SBA and the output 190 of the second
multivibrator goes to state 0. This state of output 190 lasts for
1.6 sec., during which interval the "START" lamp 15 is off and
erasure is effected via the erase-control connection.
The circuits for the clock 33 and channel switch 32,26 are shown in
FIG. 15. The clock includes two multivibrators 191,192. The output
of each multivibrator 191,192 is connected to the input of the
other one so that both are continuously triggered. In order to
start the operation of the multivibrators 191,192 initially, a
single pulse is generated by the combination of diode 193 and
transistor 194. The pulse is slightly delayed so that when it
reaches the multivibrator circuits the power on those circuits is
already at its full value. The pulse repetition frequency of the
pulses leaving the clock is 200 kilo-Hertz. This relatively high
frequency is used to cause the display to alternate between two
different values very rapidly so that to the eye it appears as a
simultaneous display of two values. This is used for two purposes.
One purpose is the simultaneous display of outputs from both
channel 1 and channel 2. Another purpose is the simultaneous
display of a double line representing a range of frequencies. The
latter purpose is employed only in channel 2 and this circuit will
now be described.
In order to produce the desired signal for channel 2 in the MFSD
mode a summing amplifier 195 is used. The summing amplifier 195
adds four different signals: (a) the voltage representing the
logarithm of the fundamental frequency of the voice, coming from
the PA circuit output 42 (FIG. 6), (b) an offset voltage at 196
used to represent the habitual pitch level, (c) a blanking signal
obtained from the PA circuit, unblanking output 197, and (d) a
fraction, at 198, of the clock output voltage. The blanking signal
(c) deflects the beam outside of the screen area when a voiced
sound is not present. The composite output signal of summing
amplifier 195 is connected to the switch output 199 by means of
field-effect transistor 200. When the gate voltage is high
transistor 200 disconnects the output 201 of summing amplifier 195
from the switch output 199. Otherwise, said output 201 is connected
to the switch output 199. The process of developing a suitable gate
voltage will be explained below. The fraction of the clock output
voltage is alternately added and subtracted to the PA output
voltage so that the signal delivered to the switch output
alternates rapidly between the value slightly above the pitch
frequency and a value slightly below the pitch frequency thereby
giving rise to a double line. The offset voltage representing the
habitual pitch level permits presentation of the display at any
vertical level on the screen, so that for example, a child's voice
may be displayed at the same level as that of a man.
The operation of channel 1 is the same as the operation just
described for channel 2 except that the output of the clock is not
added to the other three voltages mentioned above; the net result
is that while the limits of acceptability can be displayed in
channel 2 only a single line display results from channel 1. It is
seen that the voltage resulting from channel 1 is connected to the
same switch output 199 through another field effect transistor
stage 202. The necessary gate voltages for stages 200 and 202 are
developed as follows: each of the two states of bistable
multivibrator flip-flop 203 corresponds to one of the two channels:
i.e. for one state the output 201 of amplifier 195 is connected to
the switch output 199 while for the other state, the output 204 of
amplifier 205 is connected to the switch output 199. The state of
flip-flop 203 is determined by the inputs 147 and 146 (see also
FIG. 12) but a change of state occurs only when a clocking pulse
also occurs, which clocking pulse is delivered from output 206 of
the gate 128 (see FIG. 1d) or in the MFSD mode from 211 of 192 to
the clocking input 207. Output 208 of flip-flop 203 controls the
two inverter-saturation amplifiers 209,210. The output of amplifier
209 controls field-effect transistor stage 202; the output of
amplifier 210 controls field-effect transistor stage 200.
Having thus described the principles of the invention, together
with an illustrative embodiment thereof, it is to be understood
that although specific terms are employed, they are used in a
generic and descriptive sense, and not for purposes of limitation,
the scope of the invention being set forth in the following
claims.
* * * * *