US2181265A

US2181265A - Signaling system

Info

Publication number: US2181265A
Application number: US160759A
Authority: US
Inventors: Homer W Dudley
Original assignee: Bell Telephone Laboratories Inc
Current assignee: AT&T Corp
Priority date: 1937-08-25
Filing date: 1937-08-25
Publication date: 1939-11-28
Anticipated expiration: 1956-11-28

Description

Nov. 28, 1939..

H. w. DUDLEY SIGNALNG SYSTEM Filed Aug. 25, 1937 2 Sheets-Sheet 1 mf VEA/TOR H. W DUDL EV Nov. 28, 1939. H. w. DUDLEY SIGNALING SYSTEM Filed Aug. 25, 1937 2 Sheets-Sheet 2 A 7` .TORNE Y Patented Nov. 28, 1939 UNITED sTATEs PATENT oFFicE I mames l lHorner W. Dudley.

to Bell Telephone New York, N.' Y.,

Garden City, N. Y., asslxlor Laboratories, Incorporated, a corporation of New York Application August z5, 1931, serial No. 160,759

11` (ciglia-100.1)

This invention relates to signaling systems and has for its main object to provide apparatus for making a visual record of a complex speech frequency message in a form capable of being readily interpreted by the eye.

In the two most common forms of speech communication the spoken word when Areceived byl the ear gives complete information, While the other, the written word, gives only partial information to the eye. This partial information conveyed tothe eye by the written word insures the transmission of intelligence as far as the word content goes, but it omits all "those thingsthat enter into the voice, such as stress, intonation, duration, brogues and accents, slurring and weakening of sounds and the various other characteristics which go to make -up speech. The limitations of the written word are obvious when one considers that the production of speech sounds corresponds to varying six or more elements ofthe vocal system, such as-the lung pressure, the vocal cord tension, the position of the lips, teeth, front and rear tongue and uvula, while the written Word corresponds to using a single variable to transmit the information.

In its preferred embodiment, this invention provides means for producing a visual record which gives more complete information about the of the vocal system that are involved in production); and the rate at which these vary.

Speech signals satisfy the condition for possible frequency*l range reduction iii outstanding manner for inone stage of speech production, namely, the muscular, there is a very simple set of controlled motions of the muscular parts making up the speech signal. Several muscular elements move to form a speech signal, but their rates of motion are the slow muscular or syllabic frequency rates. As an example, the lips move for ordinary-speech at a cyclic rate not ordinarily exceeding 7 cycles per second for the fundamental or basic motion. Several other parts of tbc vocal system, such as the lungs, uvula, tongue, and teeth move also, but they too move at slow rates not ordinarily exceeding even these two types have a basic rate of change nature of the message than is obtainable from the written word; more particularly `to provide for the eye in a form capable of readyinterpretation a record igiving approximately vas complete information as to the character of the message as theear receives from the spoken word. Preferably, in accordance with this invention, the complex Yspeech signal is not recorded directly,

.but is rst reduced to a low frequency range ,Qfrom 0 to 20 cycles, for example) without losing f vany of the defining characteristics of the speech With this frequencyV reduction of the message to a low frequency range the visual record may be more readily read or interpreted by the viewer than would be possible if the record 'covered the entire speech frequency range. The

J invention may, therefore, be regarded as an improved type of telegraphized speech.

The information transmitted by speech does no1; absolutely require all of the frequency'space l allotted to it by the human voice. A specific case `can be worked out as to how wide a frequency band is required as a minimum, forv example, by determining and taking account `of the number of independent variables or parameters involved in speech production, (that is, the number of the independently movable physical elements at a higher fundamental frequency is that of the `vocal cords where the fundamental frequency for men is around 125 cycles per second and for women over 200 cycles per second. This mo tion differs from the others mentioned in that it is a natural frequency of stretched cords. The tension ofthe cords is controlled voluntarily and can be changed only at slow muscular or syllabic rates. The vocal cordsnot only have a high fundamental frequency but they also give a wave shape which'is rich in harmonics up to several thousand cycles per second. The `vocal cords have asteady energy source in the lung pressure which produces air vibration at. their natural frevof not over 'I cycles per second. The first motion closing with air forced through to form a hissing sound may occur between the lower lip and upper teeth as for the f sound or between the tongue and front part of thehard palate as for the "s sound, or at other places for other unvoiced or breathed sounds; Such sounds have a continuous frequency spectrum with no definite fundamental frequency. It will bev noted that in all such unvoiced sounds the volitional control is again applied at the low frequency muscular or speech range.

syllabic rate, generally not exceeding 7 cycles per second.

From the foregoing detailed discussion of they mechanics of speech sound production, it is seen that the various speech sounds are produced by voluntarily controlled variations in the muscular systems at slow syllabic frequency rates of 7 cycles perscond or less. The important muscular elements or variables used ln speech production are eight in number as follows: lung pressure; vocal cord tension and position; rear mouth resonance chamber; front mouth resonance chamber; opening from front to rear resonance chamber; mouth opening; position of uvula; position of any stricture in the sound path.

Since the important muscular variables are only eight in number, it is seen that the total frequency range required to produce sounds for the vocal systems is very limited, it being limited in Afact to the number of such'variables multiplied by the frequency range required to express the motion of each, which may be about 14 cycles per second if the reasonable assumption is made that the fundamental rate of change plus its first upper harmonic defines the-motion reasonably well.

A simple ideal system for frequency range reduction in speech vwould beto analyze the speech sound to determine what'the important motions are and then prescribe a narrow frequency band, say, of 20 cycles in width to dene each motion, but it is difficult, if not impossible, to do it in this simple way because of the analyzing diiiiculties. Thus, for example, it is very dimcult to determine from a speech sound just what position the tip of the tongue had in its production.V

However, it is reasonably simple to do this by the use ofequivalent parameters. So long as the parameters are entirely independent we know mathematically that we can use any parameters we choose. Not only can they be chosen in any fashion'provided they are independent, but if they are not entirely independent, more parameters can be chosen to make up for the lack of independence. Thus, as an example, the sounds can be analyzed into independent variables that are easily determined such as the power in each of several small: frequency subbands within the The power in each subband is not entirely independent `of that in the others, but is sufficiently so that the average power level in say eight or. ten subbands of the speech range will give us a very satisfactory set of parameters for speech definition. 'I'his analysis ofthe power level in small frequency subbands is preferably done electrically by circuit arrangements described hereinafter. '4 A In one specific aspect of this invention the speech message is analyzed electrically for its fundamental frequency andthe average power in properly chosen subbands of frequency. In analyzing for the fundamental frequency a unidirectional current is produced whose magnitude defines the frequency of the fundamental frequency and whose magnitude varies with the sy1-' labic rate of change in the fundamental frequency. The magnitude of the unidirectional current will also be indicative of the presence of any unvoiced sounds which, having no definite fundamental frequency, will result in la substantially zero value of the defining unidirectional current. An analyzer may -also be arranged to produce separate unidirectional currents each defining the average power level in a chosen subband of the frequency range of the message.

'I'hese defining unidirectional currentsin accordimpressed upon separate recording devices having needles arranged to make a written record on a moving strip of paper either in the form. of a. scratch on the surfaceor in the form of an inked or pencilled line. The plurality of separate recv ords so formed may be'viewed collectively and readily interpreted after a little training. Providing there is a defining current for the average power level in, say, eight to ten subbands of the speech frequency range the visual record produced in the above manner will give to the viewer substantially Vas complete information as if the -viewer hadheard directly the spoken word. This composite record will also be much more readily interpreted and understood than if one attempted, without frequency range reduction, to make directly a single record of the complex speech 4wave for the entire frequency rangeof speech. The present invention is therefore particularly adapted for conveying information to the deaf since the information is far more complete than could be obtained by lip reading or by the written word. Other uses, however, are contemplated such as a substitute for` the existing type of telegraphy where the telegraph record comprises merelya single variable usually a series of dots and dashes according to a particular code.

Referring to the drawings:

Fig. 1 is a schematic circuit embodying one form of this invention for making an accurate visual record of a speech message;

Fig. 1A is a fragmentary view of a type of recording device which may be used in the system of Fig. 1; f

Fig. 2 is lillustrative of the type of record obtained forlcertain spoken words by the use of the apparatus of Fig. 1;

' Fig. 3 is a simplified form of the invention; and

Fig. 4 is a representative record obtained by the system of Fig. 3.

Speech may be regarded as having a dual characteristic; On the one hand we have fixed parts or elements setting up oscillatory waves containing high frequency patterns. On the other hand we have varying parts or elements setting up modulatory'waves of low' syllabic frequency patl tern. The fixed features include (a) the existence of definite frequency subbands in which the power distribution is uniform at `any given instant; (b) the existence of a frequency spectrum that alternates from a continuous type of spectrum with no definite fundamental frequency to a discrete type with varying fundamental and with al1 the upper harmonics always present;

and (c) the .fact that time variations of the fundamental frequency and of the power in the frequency subbands occur only at syllabic frequency rates. The variable features include (a) the magnitude of the average power at each subband and (b) the nature of the signal spectrum band into which the speech signal is divided."

speech frequency band into about eight or ten.

subbands and provide means for indicating the syllabic rate of change of the average power in each subband. In Fig. 1 the speech signal is divided into ten subbands. The subbands may be of unequal width, since each subband may have a width depending upon its importance function in the production of speech as described, for exai'nple, in my United States Patent 2,151,091, issued March 21, 1939 on Signal transmission..

'I'he second-mentioned variable feature of speech concerning which information should be transmitted is the nature of-the speech spectrum as to whether it is a continuous spectrum without any denite fundamental frequency or whether it is a discrete spectrum and in the latter case as to what is the frequency of the fundamental frequency for each voiced sound analyzed. This information is transmitted and recorded in a simple manner by the apparatus of Fig. 1.

In the arrangement of Fig. 1 the speech or other vocal sound to be analyzed is picked up by a suitable high quality microphone 20 and amplied to a desired level by an amplifier 2| whose output is divided into a frequency pattern measuring circuit FP and an amplitude pattern measuring circuit AP. The frequency pattern measuring circuit FP takes advantage of the fact that l in vowels and in other sounds vhaving a vdecided fundamental frequency in the range from 80 to 320 cycles, there is a high-power level, while in sounds like the sibilant consonants where the power is in the continuous spectrum rather than 'a dicrete spectrum, the power level is much lower.

When a high level discrete spectrum is received from amplifier 2l the frequency pattern measuring circuit FP sends to the yrecording milliammeter 22 a current of a magnitude indicating what the fundamental frequency is, but not indicating anything about the amplitude of the current of the fundamental frequency. When a. lowv level continuous spectrum speech signal is received from amplifier 2| the output current from channel FP is substantially zero and hence no current is applied to the recording-device 22.

Referring more particularly to the details of the circuit FP a band-pass filter Fo selects the band from 50 to 400 cycles of the voiced signal so as to be surel to include the fundamental frequency, if any.; The output of this band-pass filter Fc is sent through an attenuation network E1 of a type oftenjtermed an equalizer which has a loss increasifng with frequency for the purpose of insuring that the fundamental frequency comes out at a higher level than any of its upper harmonics that may be present. For practical purposes this puriiies the fundamental tone. Next the output from this equalizer Ei is fed to a constant output amp Nr LA so that from this amplifier there is obtain ,i essentially a single frequency, the fundaxsitiental Ifrequency .of the speech signal at a conant p werlevel regardless of what thel frequency This fundamental frequency may be from about 50 cycles to 400 cycles. Next we pass this power y throughian equalizer En. which has a characteristie reverse -to that of .network E1 in that the output l.ffrom equalizer-Ez increases as the frequency/ increases. This output is sent through a copper-oxide detector Do which gives essentially a direct current bias which iluctuates as the fundamental frequency of speech fluctuates, that is, at a syllabic frequency. The detector output is then sent through a low-pass filter Fao cutting oif at 20 cycles so that the unwanted higher frequency products are eliminated. This output is now used as a biasing current applied to the winding of the recording milliammeter 22 whereby its recording needle 23 is moved from vits zero position an amount proportional to the defining current from filter Fao. 'I'hat is, the departure from its normal axis of the trace made by needle 23 on the moving paper strip 24 will be proportional to the frequency of the-fundamental frequency so that the fundamental `frequency will be accurately indicated on the record sheet for each speech component. For a voice with aohigh pitch the trace made by needle 23 will depart widely from its zero axis, while for a voice with a low pitch the trace will be closer its zero axis and the position of the trace will vary with any syllabic variations in the pitch of the voice.

However, when the frequency pattern circuit FP receives a speech signal having a continuous spectrum such as whena sibilant consonant sound lis impressed on the microphone 20 the entire frequency spectrum is at a low level with no frequency emphasizedover the others. Hence, for such sound there will be substantially no' energy output from low-pass filter F3n and hence milliammeter needle 23 will remain substantially on its zero axis.

There remains to be described the apparatus for analyzing the speakers energy in the different frequency subbands of the speech frequency range in order to determine the amplitude pattern characteristic of each speech signal.- The amplitude pattern measuring circuit AP of Fig. 1 is essentially a circuit for measuring how much power there is in the speech signal in chosen small frequency bands and for transmitting this information by proportional currents to a plurality of .recording devices, one for each subband. The amplitude pattern circuit AP at the output of amplifier 2l is divided by suitable band-pass filters-F1 to F1o into ten frequency subbands. As shown on the drawings, channel AP: receives the vspeech frequencies in the band from zero to 250 cycles; channel AP-z receives the speech frequencies in the band from 250 to 550 cycles; channel AP: receives the speech frequencies in the band from 550 to 850 cycles; channel AP4 receives thespeech frequencies in the band from 850 to 1150 cycles; channel AP receives the speech frequencies in the band from-1150 to 1450 cycles; channel APs receives the speech frequencies in the band from 1450 to 1750 cycles; channel AP'z receives the speech frequencies in the band'from 1750 to 2050 cycles; channel APs receives the speech frequencies in the band-from 2050 to 2350 cycles; channel APg receives the speech frequencies in ythe band from `2350 to 2650 cycles; and channel APm receives the speech frequencies in the band from 2650 to 2950 cycles. These bands have been chosen for the purpose of illustration. Obviously, other bands may be found preferable for particular circuits and may be so used without departing from the principles involved. s

Considering the channel APi, for example, the output from the zero to 250 cycles band-passv filter F1 is fed to detector D1 which may be, for

are passed-through a 20 cycle low-pass filter Fn fest subband. zero to 250 cycles,

and used to energize a recording milliammeter 4| so that the amplitude of the vibrations of needle 5| of device 4| deiines the average amount of pwer of each speech component in the frequency band from zero to 250 cycles and the amplitude of the needle vibrations will increase or diminish at a syllable rate in response to the syllable vari- :ions of the po'wer level in the designated sub- Channels AP: to APm are similar to channel AP; just described except for the frequency range analyzed by each channel. The defining currents from channel APz are recorded by needle 52 of milliammeter 42; the defining currentsl from channel AP: are recorded by needle 53 of milliammeter 42; the deiining currents from channel AP; are recorded by needle 54 of milliammeter 44; the dening currents from channel APs are recorded by needle 55 of milliammeter 45; the defining currents from channel APs are recorded by needle 56 of milliammeter 45; the dening currents from channel AP: are recorded by needle 51 of milliammeter 41; the defining currents from channel AP. are recorded by needle 58 of milliammeter 48:-the defining currents from channel APa'are recorded by needle 6 9 of milliammeter 49; and the deilning currents from channel APm are recorded by needle 60 of milliammeter 50. Filters F1 to F1o are alike except as -to the frequency band passed, as indicated on the drawings. Detectors Du to Din may be all alike and lters Fao to F40 are all alike in that they suppress all frequencies above 'cycles.

'I'he apparatus of Fig. .1, therefore, utilizes one milliammeter 22 to record the frequency variations of the fundamental frequency of the speech message and utilizes ten milliammeters 4| to 54 to record the average power level in the ten subbands of frequency into which the speech frequency band is divided by the band-pass lters F1 to F1o. The eleven milliammeters are prefer- Y ably so constructed and mounted that their'traces on the moving strip 24 are arranged compactly but without any overlapping. The paper strip 24 may be moved at a constant speed past the line of recording needles by any suitable mechanism such as an electricmotor, 55 geared to a l cylinder I4 around which the paper is rolled after the record is made.

'I'he character of the deilning patterns set up by the eleven milliammeters is shown in Fig. 2 for two typical words "sharp" and "void. The lowest-trace 41 is that of the pitch or fimdamental frequency of the tone as recorded by milliammeter 22. The next trace 44 gives a measureof the average energylevel in the lowas recorded by milliammeter 4|. upwardly onA the paper strip 24 the succeeding traces are for progressively higher subbands as indicated at the left of the iigure, trace 0I, for example, giving a m e of the average energy level in the highest subband, l2650 to 2950 cycles. Since the pitch record 41 for the distance from zero to .2 second measured along an arbitrary time scale lies substanuauy along the zero axis il for the milliammeter 22 it will be obvious that the recorded sound during this time interval is an unvoieed sound without any deilnite fundamen-l tal. For the same time interval the-ten traces above the pitch trace show substantially xero power level vfor the lower bands but withV deilnite Y power levels in their upper bands particularly in the 2350 to 2650 cycle band and the 2650 to 2950 cycle band indicating deilnitely an unvoiced aisance sound. The collective picture made by the traces for the ten subbands is that given by only one sound, namely, the sound sh. As the a sound is recorded beginning about the vertical line 1|, 'thepitch record 11 indicates a voiced sound with a definite fundamental of a frequency proportional to the distance between the trace 61 and its zero axis 10 while the energy level of .the sound is prominent in the lower subbands as well as in .the upper subbands. The composite traces just to the right oi' line 1| give an accurate picture of the power level distribution for the sound "a as in sharp and will be readily interpreted as representing that sound by one trained in observing such records. The record of the remainder of the spoken word sharp" is shown still farther to the right in Fig. 2.

Another portion of the strip 24 in Fig. 2 snows the deilning traces for the spoken word void as recorded by the eleven milliammeters of the apparatus of Fig. 1, Distinctive differences will be noted between the tracesfor the word sharp and the traces for the word "void.

The milliammeters employed as Velements 22 and 4| to 50 of Fig. 1 may be, for example, of the type disclosed in Fig. 1A which is of the moving coil type. 'I'he magnet structure 13 has opposed pole-pieces substantially surrounding a smalla coil 14 mounted for rotation in any suitable manner, not shown, in response to currents applied to` the coil from any one of the eleven channels of Fig. 1. 'I'he frame which supports coil 14 has a lightweight extension or wire 15 which, at its outer end, supports a holder for a recording needle 16. As previously stated, needle 16 may be arranged either to make a written record in the form of a scratch on the surface or in the form of an inked or penclled line.

It will be apparent from the above that the .apparatus of Fig. 1 produces a type of visual symbolic speech involving the principle of anaof speechV sounds. It is to be emphasized 'that the system of telegraphized speech proposed by this invention is inherently phonetic in nature. In recording the speech sounds from a talker with the system ot Fig. 1 itis only the phonetic sounds which h e produces that are recorded and not the artiilcial printed symbols that-'are ordinarily employed. 'Ihe printed word and the ordinary telegraph code use artincial speech symbols and the mind is ordinarily trained to learn a diiferent speech language for the eye from that for the ear. When one listens to speech it isl of course phonetic speech that he hears. Also the eye in lip reading is using a crude form of phonetic .terms of speech symbols rather than in terms.'

speech. -In that respect, the telegraphized speech of this invention corresponds to lip reading rather than the earlier forms of `visual symbolic speech. It should be noted that symbolic phonetic speech as given by this invention is an advantage and not a handicap because it tells the story of the speech as spoken, including the emotional content as well as the intelligence content. a

Fig. 3 shows a speech recording circuit somewhat like that of Fig. 1 but greatly simnliiied.

The simplification results from the use of only three of the eleven channels of Fig. 1 and as a result the speech denning signals of Fig. 3 may be readily transmitted over a telegraph line by amplified by amplifier 84 the output of which leads to a frequency pattern measuring circuit FP and an amplitude pattern measuring circuit AP. The frequency pattern channel FP of Fig. 3 is employed to differentiate between a voice and an unvoiced sound, while the two channels of the

amplitude pattern circuit

81 and 88 are employed to define the average power level in two selected subbands of the speech frequency range. 'Ihe frequency pattern circuit Fl? is identical with the circuit FP of Fig. 1 and comprises in tandem a band-passlter 15 passing the band between 50 and 400 cycles; an equalizer 11, a constant ontput amplifier 18, an equalizer 19, a detector and a low-pass filter 8| passing the band between zero and 20 cycles. As explained in connection with the channel FP of Fig. 1 there will be a substantial output current from lter 0| for a voiced sound but a substantially zero output for an unvoiced sound.

In the amplitude pattern circuit AP of Fig. 3 only two channels are employed, one channel 81 containing a band-pass' filter 86 passing a band between 400 and 800 cycles of the speech fre# quency message from amplifier 04 `while the other channel 88 comprises a band-pass filter 89 passing the band between 2300 and 2900 cycles.' These two channels have detectors 90 and 9| similar to detector 80 and also have lowfpass

Afilters

92, 93 similar to low-pass lter 8| in that they pass only the band between zero and 20 cycles.

The unidirectional current constituting the output of filter 92 which defines the average i power level lin the speech frequency subband from 400 to 800 cycles for each component of the speech message is impressed directly upon a line 94, such as a telegraph line, and therefore constitutes the main telegraph signal sent over the line. The higher band channel output from filter 93 is used to control the amount of a carrier frequency such as 69 cycles which is transmittedover the line. Oscillator 95 may be a source of 60 cycle current which is connected to a suitable modulator 96 through a filter. 91 which passes only 60 cycle current. The zero-to 20 cycle current from filter 93 which deiines the speech characteristics in the 2 300 to 2900 cycle band is also impressed upon modulator 96 whereby the 60v cycle carrier is modulated by the zero to 20v cycle band from channel 88. The output from modulator 96 passes through a 40 to 80 cycle band-pass filter 98 so that both sidebands as well as the carrier are impressed on line 949.10m; with the low-frequency 4signal from channel-81.

-At the remote station a recording milliammeter 99 is connected directly to line 94 and milliammeter 99 is preferably of a type that responds efllciently to a' frequency of 80 cycles ciu-higher.

'It follows that indicating needle |00 will vibrate at th'e 60 cycle rate with the height of its vibrations varying with the control current from channel 88 while at each instant the middle point of each swing will vary from the normal axis for the needle an amount controlled bythe low-frequency current from channel 81. As in Fig. 1 the indicating needle |00 is adapted to form a trace on a moving strip of paper |05, the character of the trace being illustrated more clearlyin Fig. 4. Line 94 also has a biasing battery |0| which is impressed on the recording milliammeter 99 with a polarity determined by the presence or 5 absence of the fundamental frequency measuring current from the'pitch channel FP and therefore with one polarity for a voiced sound and with the opposite polarity for an unvoiced sound. This diiferentiation is accomplished by having the out- 10 put from filter 8| connected to a relay |02 whose contacts in an obvious manner constitute a reversing switch fo'r battery |0|. With anunvoiced sound there is substantially zero output from filter 0| and hence relay |02 will remain un- 15 operated and positive battery is connected to the upper wire of line 94. vThis will produce a direct current which biases milliammeter 99, for example, so that for an unvoiced sound the needle |00 in the absence of any control current from 2

channels

81 and 68 wouid trace a straight line |06 of Fig. 4 which line will be termed the unvoiced axis. However, if the speech sound tobe recorded is a voiced sound tnere will be a substantial output current from filter 8| to operate relay 25 |02 and change the battery conncctionsso that now positive battery is connected to lower wire of line 94. This will cause needle |00 in the absence of any control current from

channels

81 and 88 to trace a straight line |01 which will be termed 3 the voiced axis.

The traced record of Fig. 4 is for the words sharp and void as in Fig. 2. The i'lrst portion of the traces for the sound sh is along axis |06 showing that it is an unvoiced sound and the fact that the trace is on both sides of axis |06 shows very little energy content in the 400 to 800 cycle band as compared with the energy content in the 2300 to 2900 cycle band. The voiced sound ar causes relay |02 to reverse the battery connec. 40

tions and cause the trace to lie adjacent the voiced axis |01 and it will be noted on the arbitrary time scale shown on Fig. 4 from about .23 second to .5 second the median line for the 60 cycle trace lies entirely on one side of the axis 45 |01 at a distance controlled by the energy defin ing currents from channel 81. The character of the trace for the spoken word void will be obvious from the above description, for example, for the sound v the record for a short interval 50 is along the unvoiced axis and then jumps to the voiced axis but in each instance the high frequency vibrations of the trace are of very low amplitude indicating small energy output from both

channels

81 and 88.

Ihe composite trace produced'by the indicating needle |00 isoi' a suiliciently defining nature to enable one, after a little training. to interpret accurately the speech message recorded thereby.

The system of Fig. 3 may be termed a form of 60 direct speech telegraphy since. at the sending end, the sounds of a talker are directly converted into vdefining currents which pass over line 94 where while still obtaining a satisfactory and accurate record ofthe speech. It will be apparent that if desired the three channels of the analyzing system of Fig. 3 may have their defining currents I5 impressed separately on different milliammeters in the same manner as in the system of Fig. 1.

It is also to be understood that the defining currents from the eleven channels of theI system of Fig. 1 may, if desired, 4be transmitted over a line or other communication channel to a distant receiving station where the recording milliammeters 22 and 4| to 50 are located, although more complicated apparatus will obviously be required to enable the eleven control currents to be transmitted simultaneously.

Since the frequency pattern circuit FP of Fig. 1 and Fig. 3 tends to have more inherent delay than the associated amplitude pattern branches it will generally be advisable to introduce a certain amount of delay in common to all the amplitude pattern circuits to compensate forv this difference. Thusin Fig. l an electric delay equalizer DE is connected in channel AP so as to be common to all channels AP; to APm, while in Fig. 3 a delay equalizer |09 is common to

channels

81 and 88.

The above circuits for forming telegraphized speech symbols which can readily be read by the eye to give an automatic speech telegraph circuit are merely representative of the'invention since it will be evident to those skilled in the art that many modifications may be made without departing from the spirit of the invention as defined in the appended claims. For example, if desired, the recording milliammeters employed may be of the logarithmic type rather than the linear type.

What is claimed is:

1. A recording system for a message 'represented by a complex `electrical wave of a wide band of audible frequencies. said system comprising means for selecting a number of substantially independent characteristics'of said wave,rmeans for producing a plurality of low frequency currents each having a fundamental frequency of less than ten cycles per second and each defining one of the characteristics of said wave, and means' controlled by said low frequency currents for producing a visual record from which the word content of said message may be interpreted by the eye. q

2. A recording system for a message represented by a wide band of audible frequencies, said system comprising means'for dividing said message into a plurality of'frequency subbands each of a limited frequency range, separate means for deriving from each subband a low frequency wave of a fundamental frequency of less than ten cycles per second, which wave denes the average energy level in the subband, and means for producing a separate visual record of each of said defining waves for enabling the eye to interpret the word content of said message by the inspection of said records. v

3. A recording system ,for a speech message represented by a complex electrical wave of a wide band of. audible frequencies, said system comprising means for dividing said wave into a plurality of frequency subbands, separate means for demodulating each subband, and means controlled by the demodulated currents for producing a visual record from which the word content of said message may be interpreted by the eye.

4. A recording system for a message represented by a complex electrical wave of a wide band of audible frequencies, said system comprising means for selecting a number of substantially independent 'characteristics of said wave, means for producing a plurality of low frequency currents each having a fundamental frequency of less than ten cycles per second, and each defining one of the characteristics of said wave. analyzing means for producing from said waves. low frequency wave of a fundamental frequency of less than ten cycles per second whose amplitude defines variations in the frequency of the fundamental frequency of said complex wave, and

' means for producing a visual record of said low frequency currents and said low frequency wave for enabling the eye to interpret the word content of said message bythe inspection of said record.

5. A recording system for a message represented by a complex electrical wave of a wide band of audible frequencies, said system comprising means for dividing said wave into a plurality of frequency subbands; separate means for demodulating each subband, analyzing means for producing from said wave a low frequency wave of a fundamental frequency of less than ten cycles per second whose amplitude defines variations in the frequency of the fundamental frequency of said complex wave, and means for producing a visual record of said demodulated currents and said low frequency wave for enabling the eye to interpret the word content of said message by the inspection of said record.

6. A speech recording system comprising means for converting speech into a complex electrical wave containing a wide band of frequencies of importance in speech, means for dividing said wave into a plurality of frequency subbands, separate means for transforming each of said subbands into a narrow band of frequencies whose upper frequency limit'lies well below the average fundamental frequency of speech with the amplitude of each narrow frequency band defining the average energy level in the subbandfrom which it is derived, and means controlled by the amplitude variations of said narrow frequency bands for producing a visual record from which the word content of the recorded speech may be interpreted by the eye. 1

7. A speech recording system comprising means for converting speech into a complex electrical wave containing a wide band of frequencies of importance in speech, means for dividing said wave into a plurality of frequency subbands. separate means for transforming each of said subbands into a narrow band of frequencieswhose upper frequency limit lies well below the average fundamental frequency of speech with the amplitude of each narrow frequency band denning the average energy level in the subband from which it is derived, means for producing a Vvisual record of the amplitude variations of said narrow frequency bands, and means for producing as a part of said visual recor'd an indication distinguishing voiced portions of said wave from unvoiced portions of said wave, said indication for voiced portions of said complex wave being wholly independent of the amplitude of the fundamental frequency.. y

8. A recording system for a message represented by a complex electrical wave of a wide band of audible frequencies, said system comprising means for deriving said wave into a plurality of frequency subbands, a separate rectifier for each of said subbands, an output circuit for each rectifier, a separate filter in each output circuit for suppressing frequencies in excess f about twenty cycles per second, and means conl trolled by the filtered currents from each rectifier for producing a visual record from which the mismos word content of said message may be interpreted by the eye.

9. A transmission system comprising means for converting speech into a complex electrical wave containing the frequencies of importance in speech, filter means for dividing 'said wave into a plurality of frequency subbands each of a limited frequency range a rectifier for each of said subbands, a recording instrument simultaneously responsive to the output of said rectifiers A and additional means for controlling said instrument to distinguish between voiced portions and unvoiced portions of said wave.

10. In a transmission system, means for converting speech into a complex electrical wave containing frequencies of importance in speech, filter means for dividing said wave into a plurality of frequency subbands each of a limited frequency range, a rectier for each of said subbands, a limited frequency line connected to the output of vone ofsaid rectiers, a source of carrier frequency, means for impressing on said line said carrier frequency modulated by the low frequency output of another of said rectiers, a source of direct currentfmeans for vconnecting said direct current source'to said line with one polarity for the voiced portions of said wave and the opposite polarity for the unvoiced portions of said wave, and a receiving station connected to said line comprising a recording device for producing a visual record of the combined currents received from said line.

11. In a transmission system, a telephone transmitter for converting speech into a complex electrical wave, a line connected to said transmitter, a plurality of channels connected to said line, each channel comprising a lter for passing to its respective channel only a limited band of speech frequencies, each lter passing a different frequency band, a rectier in each channel, a

movable message tape, a plurality of recording devices for simultaneously making separate visual .records on said tape on adjacent portions of said tape, each of a plurality of said devices being responsive to the output of one of said rectiers, means for producing a low frequency current the amplitude of which defines thevfrequency of the fundamental frequency of the voiced portions of the converted speech signal while lproducing a substantially zero current for the unvoiced portions of the speech signal, a separate one of said receiving devices being responsive to the dening current of said last-mentioned means.

HOMER. W. DUDLEY.