US2928902A - Signal transmission - Google Patents

Description

March l5, 1960 F. vlLBlG SIGNAL TRANSMISSION 4 sheets-sheet 1 Filed May 14, 1957 kwmwkh March l5, 1960 F. vlLBlG SIGNAL TRANSMISSION 4 Sheets-Sheet 2 Filed May 14, 1957 Il l@ INVENTOR.

March 15, 1960 F. vlLBlG SIGNAL TRANSMISSION 4 sheets-sheet :'s

Filed May 14, 1957 ll bio Xghk ww# March 15, 1960 F. vlLBlG SIGNAL TRANSMISSION 4 Sheets-Sheet 4 Filed May 14, 1957 United States Patent a SIGNAL TRANSMISSION Friedrich Vilbig, Cambridge, Mass., assignor to the United States of America as represented by the Secretary of the Air Force Application May 14, 1957, Serial No. 659,183

8 Claims. (Cl. 179-1555) (Granted under Title 35, U.S. Code (1952), sec. 266) The invention described herein may be manufactured and used by or for the United States Government for governmental purposes without payment to me of any royalty thereon.

l This invention relates to methods and systems for communication, particularly methods and systems for analyzing speech by converting it into electrical signals and then causing the signals to function in such manner as to reproduce the speech. As used herein, the term speech embraces all vocal utterances.

Speech-band compression aims to miniaturize the frequency-band without impairing the desired informational content. Spectroscopie investigation of speech has shown a pitch-harmonic line-spectrum for vowels and a noiselike spectrum for turbulent consonants (excitation function). The envelope of both types of spectra represents the system-function.

Voice coding systems may achieve a compression factor of 10 to 20. By a spectrum analysis on the transmitter side, a pitch frequency is extracted and a system-function is derived. Both are transmitted to a receiver where speech is synthesized.

Speech contains audible frequencies up to about 16 kc. Speech transmission will be nearly perfect in respect to fidelity and intelligibility if a 16 kc. wide band is used for transmission, and if there are no distortions, either linear or non-linear. Conversation will sound as though the persons are facing each other. Due to economical or technical consideration, the transmission band is reduced to about 3 kc. by low-pass filtering. By this limitation, the intelligibility, is not inuenced very much, but it would be decreased considerably upon further limitation to lower frequencies.

To arrive at a better utilization of the transmission lines the speech-band is compressed without losing the necessary information content. Ideally performed, the process should not be detectable by the listener. Such a compression is obtainable by a scan coding process at the transmitter, and a scanV decoding at the receiver side. This process requires at the transmitter side an analyzer and on the receiver side a synthesizer for speech.

There are other advantages connected with speechband compression such as a reduction of noise and a certain security against undesired listening.

In order to achieve speech-band compression certain characteirstics of speech vibrations may be utilized. An analysis of speech vibration in the case of a vowel will show a typical repetition of certain configurations of vibration where the repetition frequency is called pitch frequency. lf unvoiced sounds are considered, a more or less irregular vibration is indicated.

In an inspection of the frequency-spectrum distribution of speech, in the case of vowels, a large number of harmonics of the pitch frequency, the fundamental frequency excited by the larynx, is indicated. This means we have a line spectrum, while non-vowel sounds cause a noise spectrum. The amplitudes are represented by difvowel case is energy concentration in certain frequency areas which are called the formant areas.

Generally, a vowel has three significant formants and the position and amplitude of the formants are characteristic for a particular vowel.

The analysis of speech vibration results, as already mentioned, in some physical characteristics; in the case of vowels, there is a harmonic line-spectrum, where the lines have the distance of pitch frequency. In the case of nonvowel sounds, there is noise spectrum. Last, there are the energy distribution and the formants indicated by the envelope. Furthermore, an analysis produces indications` about the electric response of the time functions of the sounds which can be used. For example, discrimination may be achieved between continuous consonants as s and f which can be produced over a longer time interval compared with such stop sounds as p, t, and k which have pulse-like time functions.

It is possible to describe the speech vibrations by merely physical values, such as frequency, amplitude, energy content, time functions and so on. Methods using exclusively these physically measurable values for speech compression may, therefore, be called physical methods.

A consideration of instantaneous pictures of a speech spectrum displays two typical extremes; a harmonic linespectrum and a noise-spectrum. The one which is produced depends on the excitation of the speech sounds. Beyond this, the envelope of the spectrum is of importance, due to the resonant structure of the sound source system. The spectrum curve contains the characteristic of the excitation function and of the systemfunction. In sound-excitation there is also a distinction between voiced sounds or vowels and unvoiced sounds or non-vowels. If a vowel is produced, the air current coming out of the lungs will be rhythmically modulated by the vibration of the larynx. The fundamental frequency of the so excited air pulses has already been defined as pitch frequency. It varies for different persons, differing in its frequency and inflection. In the spectrum of vowels we find a number of frequency lines which are harmonics of the pitch frequency. This is the picture of a harmonic line spectrum. Also, we may say the appearance of a harmonic line spectrum is characteristic for a vowel excitation.

Uttering a non-vowel sound, e.g., a hissing sound, will not excite the larynx, as may be perceived. A hiss will be produced by the air current which is represented by a noise spectrum. In general, we may say the appearance of a noise spectrum is characteristic for a nonvowel excitation.

The transmission of a so-called pitch-hiss-signal indicates either the presence of a voiced sound or vowel with the pitch frequency as fundamental in the line spectrum or of an unvoiced sound with a noise spectrum. Beyond the pitch-hiss discrimination, the pitch-frequency signal is another important excitation characteristic. It indicates the frequency and the alteration in pitch frequency. But it is not absolutely necessary to transmit the pitch frequency too, since it is possible to add an artificial pitch Y frequency which is somewhat different from the original pitch on the receiver side. This would avoid using a frequency-band for transmission of the pitch signal, without any important loss of intelligibility. Also, it is possible to combine the pitch-frequencysignal with theA pitch-hiss-signal, for instance, by the assumption thatk sented by the spectrum envelope. The so-called formant;

Patented Mar. 15, 1960 ranges are resonant areas which result from the resonant positions of the different cavities in the vocal tract. In speaking vowels, the positionof the tongue, the lips, and so forth, will change the cavity resonances. If these cavities come to vibration by excitation of the pitch modulated air pulses, direct and coupled resonance frequencies will be excited. These build up the already mentioned formant frequencies. Each formant frequency will appear as a damped oscillation periodically excited by the pitch-period. This procedure can also be seen Linder the aspect of formant frequencies modulated by the pitch frequency and its harmonics, which result in a line spectrum.

In non-vowel sounds the excitation of the different cavities of the vocal tract is accomplished by a continuous air current modulated by noise. This noise modulation results from friction and turbulent excitation in the mouth, especially at the teeth. Very high resonances will appear in speaking s and f sounds, for instance, since .the small slot between teeth and lips is an acoustical resonator of high frequency. The resuit is that also in nonvowel sounds, the envelope is a picture of the systemfunction.

It has been seen that speech contains a pitch-harmonic line-spectrum for vowels and a noise-like spectrum for turbulent consonants (excitation function). The envei lopes of both types of spectra represents the system-function. Also it has been shown that the transmission of a so-called pitch-hiss-signal indicates either the presence of a voiced sound or vowel with the pitch frequency as fundamental in the line-spectrum or of an unvoiced sound with a noise spectrum. Beyond the pitch-hiss discrimination, the pitch-frequency signal is another important excitation characteristic. It indicates the frequency and the alteration in pitch frequency. Finally we may combine the pitch-frequency signal with the pitch-hiss-signal by assuming that there is hiss in the absence of pitch. During intermission in speech, a system-function is not found for the hiss spectrum, resulting in no output.

In accordance with the present invention a scan voice coding method and system is provided by performing a spectrum analysis of the speech spectrum, extracting a pitch-frequency signal therefrom including the alteration in its frequency, deriving a system-function, transmitting both to a receiver for speech synthesizing.

More specifically, a communication system is provided by the present invention which achieves a substantial compression in bandwidth. In this system we derive by utilizing a pitch extractor at the input of a transmitter the larynx or pitch frequency. This frequency is then transmitted to a receiver speech synthesizer and controls a pitch generator therein. This generator excites not only the pitch frequency but also a large number of'its harmonies each having the same amplitude. At the input speech analyzer of the transmitter we also provide means to distinguish between hiss and larynx signals. These two events are also transmitted to the speech synthesizer located at the receiver and these control either a hiss or pitch generator. The aforementioned pitch frequency is extracted by applying the principle that a vowel spectrum contains harmonics of the pitch frequency. By the frequency difference of these harmonics, pulses at the' pitch frequency are obtained.

There is simultaneously provided by the aforementioned communication system a speech spectrum analyzer at the transmitter end. If a vowel is present at the input of the speech analyzer, this vowel will excite corresponding voltages at lter outputs of the said analyzer. At the speech synthesizerrlocated at the receiver, there is created all the harmonics of the pitch frequency by the pitch extractor signal. Since the speech synthesizer excites all harmonics with equal amplitudes, control is exercised of the amplitude of the harmonics to realize an envelope corresponding to the speech spectrum envelope present at the input of the speech analyzer. This is achieved by scanning the harmonics of the speech spectrum present at the analyzer and controlling the harmonics of the pitch frequency at the synthesizer by the scanning voltages. If there is a consonant instead of a vowel to be transmitted the only difference will be that the pitch generator will be replaced by a hiss generator, and thereby there is provided a system for reproducing the original input speech spectrum at the synthesizer.

By the aforementioned scanning, the present invention provides means-.for deriving a system-function signal which requires only a single channel for transmission. The procedure of envelope transmission within a time interval is referred to hereafter as time multiplex.

The present invention also provides a system utilizing only four parameters, viz., a hiss signal, a larynx signal, the pitch frequency and a modulation envelope which corresponds to the instantaneous position of formants. Thus, a special formant extraction is unnecessary.

Also in accordance with the invention a speech band of 3 kc. is analyzed by 100 magnetostriction filters, each of which has a bandwidth of 30 c.p.s. The single output voltages of these filters are scanned by the rotating arm of a 10G-contact mercury-jet switch, rectified, and filtered through a low-pass filter having a cutoff frequency of approximately 200 c.p.s. The envelope curve thus obtained is transmitted to the receiver in a channel of about 200 c.p.s. The arm of SO-contact mercury-jet switch rotates synchronously with the analyzer switch and distributes the envelope curve voltage to the bias inputs of 5i) modulators. These modulators control the harmonics of the pitch-frequency signal, or of the bias spectra, in the out put of 50 pairs (100) magnetostriction filters. The rcsulting 'output of the synthesizer is a frequency-band whose spectrum is similar to that of the original speechband. Including transmission of the pitch signal, a frequency-band compression in the order of 10:1 is made possible.

A more completeunderstanding of Ithe invention may be gained from the following description of one of its preferred embodiments, when considered in conjunction with the drawings, in which:

Fig. l is a diagram of a scan voice coding system for achieving speech-band compression in signal transmission;

Fig. 2 shows a typical spectrum identifiable with signicant points in Fig. l;

Fig. 3 shows diagrammatically a novel pitch extractor utilized in the scan voice coding system of Fig. 1;

Fig. 4 shows the circuit for the modulator and demodulator utilized at the receiver-synthesizer of the scan voice coding system of Fig. 1; and

Fig. 5 is a graph illustrating the effect of shifting the phase of the output of three groups of modulators and then adding them.

Referring now to Fig. l, there is generated at the -output of microphone 1 electric voltage fluctuations of sound pressure present at the input of microphone 1. For example when speech is present, a characteristic speech spectrum of about 3 kc. is present at point I. rfhis is illustrated by Waveform I in Fi". 2 showing the speech spectrum of voiced and unvoiced sounds. The voiced sound creates a speech spectrum containing a large number of harmonics of the pitch frequency and also an envelope of varying amplitude as shown at Ia of Fib. 2. The unvoiced sounds cause a noise spectrum as illustrated at Ib of Fig. 2.

From point I of Fig. l the signal is fed by way of line 2 to pitch extractor 3. This pitch extractor is more fully illustrated in Fig. 3. In the pitch extractor we utilize the fact that a vowel spectrum contains harmonics of the pitch frequency and by the frequency difference of these harmonics, pulses at the pitch frequency can be obtained.

Now referring to Fig. 3 for a more detailed explanationof pitch extractor 3, at terminal 10 there is fed the speech signal. Where there is a vowel spectrum present as shown in Iaof Fig. 2, the lower part of the spectrum is suppressed Zby 1000 Vc.p.s. cutoff high-pass filter so that at the input of diode 2 there is present waveform 9. Diode rectifier 2 produces a smaller amplitude of the second harmonic compared to 4the pitch frequency amplitude. At the input of amplifier 3 we consequently have waveform 10. The output of amplifier 3 is fed into 80 c.p.s. low-pass filter 4. The effect of diode 2 is enchanced by passing the fundamental and second harmonic through 80 c.p.s. low-pass filter 4 having an attenuation slope of 12 db/octave. Thus the second harmonic will be satisfactorily suppressed. At the input of clipper 5 there is present waveform 11 containing pulses representative of the fundamental pitch. The input of integrator 6 receives normal pulses as shown in waveform 12, these pulses are of uniform amplitude because of clipper 5. The output signal of integrator 6 as shown in waveform 13 is integrated. Diode 7 produces a signal as shown in waveform 14. These pulses are fed to pulse generator 8 whose output is a pulse representative of the pitch frequency and which supplies the excitation function transmitted to the synthesizer shown in Fig. 1.

Again referring to Fig. l, the output of pitch extractor 3 as indicated at point V will be the envelope waveform at V of Fig. 2. From point I of Fig. 1, the speech spectrum of about 3 kc. is fed to push-pull modulator 5 by way of line 4. There is also fed into push-pull modulator 5 by way of line 6 a carrier frequency of 130 kc. which is generated in oscillator 7. The output at point II of Fig. 1 is fed simultaneously to filters 101-200 by way of line 8. Since push-pull modulator 5 is utilized, the audio frequency band is shifted to the higher position where the carrier (130 kc.) and the lower sideboard are suppressed. The resulting frequency range of 130 to 133 kc. is analyzed by magnetostriction filters 101 through 200. Each of the filters has a bandwidth of 30 c.p.s. They are designed so that they overlap each other at their half power points. The entire upper band Vextends from 130 to 133 kc. The output of each of these magnetostriction filters 101-200 is fed to its corresponding rectifier 201-300. The resulting D.C. output voltages from rectifiers 201-300 are fed into its corresponding storage capacitors 301-400. The output of storage capacitors 301-400 is fed by way of lines 401-500 to their corresponding connections 501-600 on 100-contact-mercury-jet switch 9. Rotary arm 10 of switch 9 scans the outputs` of storage capacitors 301-400. The output signal at point III thus obtained is shown as a waveform at III of Fig. 2. The voltage charges in storage capacitors 301-400 are scanned by switch 10 rotating 30 times a second. The fundamental frequency of the so developed scan-pulse is 3000 c.p.s. The amplitudes correspond with the envelope slope of the waveform at point III as illustrated at III of Fig. 2. By means of low-pass filter 11, the fast alterations in the curve are suppressed, only the smooth envelope remains at point IV of Fig. 1 as illustrated in waveform at IV of Fig. 2. The cuto frequency of low-pass filter 11 equals the bandwidth needed for the envelope transmission, at present this frequency is 200 c.p.s. The output signals at point IV as illustrated in waveform IV of Fig. 2 are to be transmitted to the speech synthesizer (also referred to as receiver).

It may be seen that at the analyzer (which may also be referred to as the transmitter) there has beerextracted coded signals representative of speech. These signals are illustrated at IV and V of Fig. 2. These signals are transmitted to a speech synthesizer (also referred to as a receiver). This transmission may be by any known means such as telephone line, or radio.

On 'the synthesizer side, at point A is present the incoming pitch-frequency signal as shown in waveform as at point V in Fig. 1. Multivibrator 12 or hiss-noise generator 13 is controlled by said incoming pitch-frequency signal.

When a voice or unvoiced sound is present at point A the waveform that is present is illustrated in V(a) of 6 Fig. 2. Duringintermissions in putas shownin V(b).of Fig. 2v.

The normal position of relay switch 16 is at point C.

speech there is 11o-out- When a pitch-frequency signal is present at point A,

`we have an input either of line spectrum B or a noise spectrum C. Modulator 17 is also fed a carrier frequency of 130 kc. which is generated in oscillator 1S. The upper side band (130 130 kc.) of the output spectrum at point D (as illustrated at D of Fig, 2) in which the carrier frequency is suppressed will receive an analysis rby magnetostriction filters 601-700. As shown in Fig. 4,` said magnetostriction filters are connected in pairs, for example 601-602 resulting in 50 double-filters. The pair'is comprised of one filter 601 adapted to respond to a positive input and the other 602 toa negative input. The 50 filter outputs are linked to corresponding modulators 701-750. The control inputs 851-900 of each ofv modulators701-750 receives its voltage from the points 751-800 of 50 contact rotating switch 19, scanning the envelope E as shown at IV of Fig. 2.

This envelope voltage curve is representative of the modulator curve of the original speech spectrum. Arm

20v of switch 19 synchronously rotates with rotating arm` 10 of switch 9 of the analyzer. Thus the envelope curve voltage is sampled for 1,6500 second, but remains at the input of each of modulators 701-750 by means of storage capacitors 801-850 until rotating arm 20 comes again to this pointrafter lf3() second. If, between samples, the envelope changes, the modulator gets the corresponding new voltage. l

The outputs of modulators 701-750 are connected so that the output of modulators 701, 704, 707 749 are connected together, outputs 702, 705 750 are interconnected and 703, 706 '748 are also linked together. Thereby establishing a grouping of modulators 701-750 having three outputs 21-23 which are fed to a phase circuit 24 which will shift the phase 0, 120, and 240 respectively and then the resultant outputs arev added in circuit 24. By these means, the modulation effect of any modulator practically is restricted to the frequency range of the corresponding filter as shown in the curve of Fig. 5. This is very important since otherwise the envelope of the side band present at point F of Fig. 4 (illustrated at F of Fig. 2) will be malformed and high distortions appear. The complex output voltage of three groups is fed to mixer 25. Mixer 25 also receives a carrier frequency of kc. The complex voltage of the three groups is mixed with the carrier and at point G of Fig. 4 there is present an output with the waveform as illustrated at G of Fig. 2. This voltage output is fed to demodulator 26 from which we receive a demodulated output at point H which is illustrated at H of Fig. 2. The audio frequency band at point H is very similar to the one at point I of Fig. 1. We, thereby, reproduce at the output of the demodulator a waveform which is similar to the original speech spectrum present at the input of the analyzer.

In effect we have at point IV of Fig. 1 derived a modulation curve. This modulation curve has been transmitted to the synthesizer. The pitch frequency and its equal amplitude harmonics are controlled by filters which are similar to that at the analyzer. The switched frequencies at the synthesizer are injected into a number of modulators (corresponding to number of pairs of filters). The modulators are also controlled by voltages corresponding to the instantaneous position of the scanner which runs synchronized to the scanner which is located at the analyzer. When there is transmitted to the synthesizer scanner the voltages being scanned at the analyzer, there is 7 obtained from 'the modulators the original input spectrum. If there is a consonant instead of a vowel to b e transmitted the only difference will be that the pitch generator will be replaced by a hiss generator. y t

Many variations and modifications ofthe invention will occur to those who are skilled in the art to which the invention relates, and it is accordingly intended that the claims that follow shall not be limited, but only illustrated, by the details of the embodiment of the invention shown and described herein.

What is claimed is: y

l. A communication system including a speech analyzer and synthesizer, said system being comprised of transducer means located at said speech analyzer to convert input speech into electrical signalsV representative of the speech spectrum, means to extract from said electrical signals pitch frequency pulses yequal to the differencer between the harmonic and basic frequency components of said electrical signals, means to simultaneously extract the modulation envelope from said electrical signals, said envelope containing only direct current voltage components, time multiplex means to transmit said direct current voltage envelope in the form of pulses on a single frequency channel, means located at said synthesizer to generate noise, means at said synthesizer to generate pulses having an audiorate of repetition, means controlled by said pitch frequency pulses toselect either said audio pulse means or said noise generating means, modulator means adapted to receive the output signals from said noise generating means or said audio pulse generating means, said modulating means simultaneously receiving a fixed frequency carrier signal, means to divide the modu' lated signal from said modulating means, means to arnplitude modulate said divided signals, said amplitude modulating means controlled by said transmitted direct current voltage envelope, and transducer means to combine said amplitude modulated signal to reproduce speech.

2. A communication system as defined in claim l wherein said means to extract pitch frequency pulses is comprised of first filter means adapted to receive said` electrical signals, means to rectify the output signal from said first filter means, second means to filter said rectified signal, means to clip said filtered rectified signal, means to Vintegrate said clipped signal, means to rectify said inytegrated signal, and pulse generator means controlled by said rectified integrated signals.

3. A communication system as defined in claim l wherein said means to extract the modulation envelope from said electrical signals is comprised of means to push-puli modulate said electrical signals with a carrier signal thus obtaining lower and upper sideband signals from said modulation operation, means to analyze said modulated signal utilizing a set of magnetostriction filters, said filters so arranged so that each of said filters overlaps the next succeeding one at the half power point, means to rectify the output signal of each of said magnetostriction filters to obtain a direct current voltage, means to store direct current voltage, means to scan each of said stored voltage, and means to filter said scanned voltage.

4. A communication system including a speech analyzer and synthesizer, said system being comprised of means located at said speech analyzer to convert inputV speech containing formants into electrical signals representative of the speech spectrums, means to extract from said electrical signals pitch frequency pulses equal to the difference between the harmonic and basic frequency components of said electrical signals, means to simultaneously extract from said electrical signals a modulation envelope which corresponds to the instantaneous position of said formants in said speech input, said envelope con taining only direct current voltage components, time multiplex means to transmit said direct current voltage envelope on a single frequency channel, means located at said synthesizer to generate noise, means at said synthesizerto generate pulses havingan audio rate ofrepdesses tiltion, said rate of repetition controlled by said pitch fref quency pulses, means controlled by said pitch frequencyY receive the output signals from said noise generating means or said audio pulse generating means, said modulating means simultaneously receiving a carrier signal having a fixed frequency higher than any audio frequency, means to divide the modulated signal from said modulating means, means to amplitude modulate said modulated divided signals, said amplitude modulated means controlled by said transmitted direct current voltage envelope, and transducer means to combine said amplitude modulated signals to reproduce speech.

`5. A communication system comprising a speech analyzer and synthesizer, said analyzer adapted to receive at its input speech, means to convert said speech into electrical signals representative of the speech spectrum,

means to extract from said electrical signal pitch-pulse signalsrepresentative of the fundamental pitch frequency offsaid speech spectrum, means to simultaneously modulate said electrical signals with a carrier signal of higher frequency than audio in such a manner as to produce an upper Vand lower sideband signal, means to divide said upper sideband signal by a set of narrow bandpass filtersy to obtain at the output of each of said filters a voltage proportional to its corresponding instantaneous speech spectrum, means to rectify each of said filter output voltages to obtain a direct current voltage, means to store each of Vsaid direct current voltages, means to scan said stored direct current voltages, means to filter said scanned voltages to obtain only the smooth envelope of said speech spectrum, means to transmit said pitch pulse signals and said scanned direct current voltages, pulse generating means located at said speech synthesizer controlled by said pitch pulse signals in such manner as to generate pulses having an audio rate, noise generating means normally connected to push-pull modulator means in the absence of said transmitted pitch pulse signals, relay means to connect said audio pulse generating means to said modulator means in place of said noise generating means, said relay means operative upon receipt of said transmitted pitch pulse signals, means to generate' a corner signal for said pushpul1 modulator means, said carrier means generating a signal with a frequency equal to Vthat of said carrier in said speech analyzer, means to divide the output signal from said push-pull modulator means by a set of narrow band filters, means to modulate each of said divided signals by said transmitted scanned direct current voltages, and means to combine said modulated signals to reproduce said speech present at the input of said speech analyzer.

6. A communication system as dened in claim 5' wherein the means to extract said pitch pulse signals is comprised of filter means adapted to receive said electrical signals representative of speech, means to rectify the signal, resulting from said filtering operation. means to filter said rectied signal, means to clip said filtered rectified signal, means to integrate said clipped signal, means to rectify said integrated signal, and pulseV gen-k erator means controlled by said rectified integrated signal.

7. A communication system as defined in claim 5 wherein the means to divide said upper sideband signal4 is comprised of a set of magnetostriction filters, each of said filters being arranged so that the succeeding filter overlaps at the half power point.

8. A communication system as defined in claim 5 wherein said means to combine said modulated signals to reproduce speech is comprised of a phasing circuit adapted to receive three input signals from said divided signal modulators, said phasing circuit operating to provide 0 phase shift for the first of said input signals, a D phase shift for the second of said input signals, and a 240 phaseY s'hiftfor thewthird ofpsaid input signals, mixer ,means receiving the signal output from said phasing means and simultaneously said carrier signal gene1-ated in said speech synthesizer, demodulator means References Cited in the le of this patent UNITED STATES PATENTS 2,098,956 Dudley Nov. 16, 1937 i0 Dudley Mar. 21, 1939 Dudley May 27, 1941 Craib f. June 11, 1946 Steinberg Apr. 14, 1953 Miller Apr. 5, 1955

Images