US3573374A

US3573374A - Formant vocoder utilizing resonator damping

Info

Publication number: US3573374A
Application number: US700542*A
Authority: US
Inventors: Louis R Focht; James M Loe
Original assignee: Philco Ford Corp
Current assignee: Space Systems Loral LLC
Priority date: 1968-01-25
Filing date: 1968-01-25
Publication date: 1971-04-06
Anticipated expiration: 1988-04-06
Also published as: DE1903623B2; JPS4912009B1; DE1903623A1; DE1903623C3; GB1261552A

Abstract

A formant vocoder comprising formant resonators and means for damping the resonators just prior to each successive pitch pulse. Damping is achieved by changing the Q of the resonators.

Description

United States Patent [72] Inventors Louis R. Focht 1 [56] References Cited gm G UNITED STATES PATENTS James 2,s17,707 12/1957 Weibel 179/1(AS) 211 Appl. No. 700,542 [22] Filed Jan 25 1968 2,928,901 3/1960 Bogert 179/1(AS) 3,268,660 8/ 1966 Flanagan 179/ 1(AS) [45] Patented Apr. 6,1971 3 431 362 3/1969 [73] Assignee PhikmFord Corporation 1 er 179/1(AS) Philadelphia, Pa. Primary Examiner-William C. Cooper Assistant Examiner-Jon Bradford Leaheey Attorney-Robert D. Sanborn [54] FORMANT VOCODER UTILIZING RESONATOR DAMPING 3 Claims, 3 Drawing Figs.

[52] U.S.Cl 179/1 [5 1] Int. Cl Gl0l l/00, ABSTRACT: A formant vocoder comprising formant resona- GlOl 1/04 tors and means for damping the resonators just prior to each [50] Field of Search 179/ 1 (AS), successive pitch pulse. Damping is achieved by changing the O 15.55 of the resonators.

2 l :2 mzauzA/c'r I V d 1 r u'g' r c :53 (ml [Mme 7 4 I /4 1a 2; I mam fi/nw I v raw xiv/rm l' fidillllf mum 0:71am? I an. M00. Ilia 170A ,i

3/ l 4 22 2a 6 l' /0 I z I 1 v IW/fc'l/ zfi/xmam 7;; p Z ra/cm'a 6,4 mm M00. anon 42m 1 IETICTUR (/RC'lI/f I 1 I Z4 30 I I g k fir/7M 241 xwm'r M00. 5 Avid/V420,? I MM: 1 l 6!. I f/WTblI/Zifl I J'Pflt'l Ii/4'44 FORMANT VOCODER UTILIZING RESONATOR DAMPING In the process of human speech, excitation of vocal chords excites the various resonant cavities of the human speech mechanism producing a speech wave which is characterized by concentrations of acoustic energy in the frequency domain, called formants. Glottal damping decreases the Q of the resonant cavities from a high value to a low value just prior to each successive excitation point (pitch pulse) to prevent carryover of the ringing signals in the resonant cavities to successive pitch pulses.

In the formant vocoder, signals representative of the frequencies and amplitudes of the formants of a speech sound are transmitted and utilized to control a formant synthesizer which contains three or more resonant circuits. In order to make the synthesized signal a substantial replica of the speech wave produced by the human speech mechanism, the Q of the resonant circuits must have a high value. This, without damping, produces carry-over" of the signals in the resonant circuits between successive excitation points and hence a low quality synthesized speech wave. Attempts to improve the quality of the synthesized speech wave without damping by adjusting the resonant circuits so that they have a Q value somewhere between the damped and undamped values have not been totally successful.

It is therefore an object of the present invention to improve the quality of speech waves produced by a formant vocoder. A further object of the present invention is to provide a system in which the Q of the resonant circuits of the synthesizer of a formant vocoder is regulated so as to prevent carryover" ringing of the resonant circuits of the synthesizer between successive pitch pulses.

In accordance with the present invention there is provided in the synthesizer of a formant vocoder a glottal damping generator. This generator, which is connected to an output of the pitch generator of the synthesizer and an input terminal of at least one of the formant resonators of the synthesizer, reduces the Q of the resonators to which it is connected for a brief period of time before the occurrence of each successive pitch pulse. This damps the resonant circuits at the same time and to substantially the same degree that the glottis damps the resonant cavities of the human speech mechanism and thus makes the synthesized speech wave a better replica of the spoken speech wave.

For a better understanding of the present invention together with other and further objectsthereof reference should now be had to the following detailed description which is to be read in conjunction with the accompanying drawings in which:

FIG. 1 is a block diagram of a formant vocoder in accordance with the present invention;

FIG. 2 is a graph illustrating first formant speech signals produced by the human speech mechanism and formant vocoders; and

FIG. 3 is a schematic diagram of a portion of the circuit of FIG. 1.

Referring to FIG. 1, there is shown a formant vocoder in accordance with the present invention. A speech signal such as that produced by a standard high-quality microphone (not shown), is supplied to a formant frequency and amplitude detector 2, a pitch detector 4 and a voicing detector 6. Detector 2 produces signals representative of the frequencies F F and F of the first three formants of the speech wave and signals representative of the amplitudes A,, A and A of the first three formants of the speech wave. A suitable formant frequency and amplitude detector for the system of the present invention is described in U.S. Pat. No. 2,458,227, issued to R. Vermeulen et al. on Jan. 4, 1949, entitled Device for Artificially Generating Speech Sounds by Electrical Means.

Detector 4 produces a signal the amplitude of which is representative of the frequency of vocal cord excitation, i.e. the the pitch frequency. Voicing detector 6 measures the regularity of the spacing between adjacent pitch pulses and produces a signal representative of the ratio of harmonic to nonharmonically related energy present in the speech wave. This information distinguishes voiced sounds from unvoiced sounds. Suitable pitch and voicing detectors for use in the system of the present invention are described in U.S. Pat. No. 3,488,442 of Louis R. Foc'ht, issued Jan. 6, 1970, entitled Single Equivalent Formant Speech Analysis System." I

The signals generated by detectors 2, 4 and.6 are trans mitted by a conventional communication channel to a synthesizer network. For example, the detector signals can be transmitted by continuously varying the amplitude of a radio frequency carrier signal in accordance with the amplitude of the detector output signals. I

The output of detector 4 is supplied to a frequency-controllable pitch oscillator 8 which generates a signal having a frequency which is a function of the amplitude of the control signal supplied thereto from detector 4. Since, as previously stated, the amplitude of the pitch signal is a function of the pitch frequency, the signal generated by oscillator 8 has a frequency equal to the pitch frequency of the input speech signal.

The output of voicing detector 6 is supplied to a gating circuit l0. Gating circuit 10 is also coupled to pitch oscillator 8 and to a noise generator 12. Depending upon the amplitude of the voicing signal and accordingly upon whether the speech signal is voiced or unvoiced, gate 10 will pass either the signal from oscillator 8 or the signal from generator 12.

The output signal of gate 10 is supplied in parallelto modulators-I4, l6, and 18. Modulators l4, l6, and 18 modulate the amplitude of this signal in accordance with the amplitude of the signals representative of the formant amplitudes A,, A and A respectively. The amplitude modulated signals produced by

modulators

14, 16, and 18 are supplied to tuna- -ble

resonant circuits

20, 22, and 24, respectively. The tuning of

circuits

20, 22, and 24 is controlled by the formant frequenc'y signals F,, F and F respectively. The output signals of the tunable

resonant circuits

20, 22, and 24 are processed through band-pass filters, 26, 28, and 30, respectively, which are identical to those used in detector 2 and which remove undesirable harmonic components introduced during the processing of the speech wave, and then combined additively in circuit-31 to form the synthesized speech signal.

The circuit described heretofore is that of a conventional formant vocoder. In accordance with the present invention, a glottal damping generator 32 has an input terminal coupledto. pitch generator 8 and an output terminal coupled in parallel to

tunable resonators

20 and 22. As will be explained in detail presently, generator 32 generates damping signals which reduce the Q of the first and

second formant resonators

20 and 22. The timing of these signals is set to reduce the Q of

resonators

20 and 22 just prior to each successive pitch pulse. Probably because of its physical placement in the human vocal tract the third formant resonant cavity of human speech appears not to be clamped by glottal opening and therefore the third formant resonator is not damped in the synthesizer.

The operation of the system of FIG. 1 will now be explained by reference to the waveforms of FIG. 2 of the accompanying drawings. Curve A of FIG. 2 illustrates the amplitude spec trum of the first formant of a sound uttered by the human vocal tract. Due to glottal damping, the amplitude of this signal is greatly reduced just prior to each successive pitch pulse, P. Curve B of FIG. 2 illustrates the amplitude spectrum of the first formant signal as synthesized by a conventional formant vocoder. Unlike the speech wave of curve A, the amplitude of this signal is still large just prior to each successive pitch pulse. Thus each ringing signal of the first formant resonator is superimposed on the next ringing signal of that resonator. This superimposition distorts the synthesized speech wave.

Curve C of FIG. 2 illustrates the amplitude spectrum of the output signal of the first formant resonator of the present invention. Due to the action of the glottal damping generator 32, the amplitude of the signal from resonator 20 is highly damped before each successive pitch pulse as is the amplitude of sounds uttered by the human vocal tract, curve A. From the foregoing it is apparent that use of the glottal damping generator 32 improves the quality of a formant vocoder because a better electrical analogue of the human vocal tract is obtained.

Referring now to FIG. 3 there is shown a schematic diagram of circuits suitable for use as the pitch generator 8, the resonator 20, and the glottal damping generator 32 of FIG. 1. The signal representative of the pitch rate is supplied to the base of a transistor 33 which forms part of a charging circuit which also includes resistor 34 and capacitor 35. Connected in shunt with the charging circuit is a voltage reference network 36. The voltage appearing across capacitor 35 is supplied as one input signal to a high-gain differential amplifier 38 and the voltage at the junction of the two resistors of voltage divider network 36 is supplied as the second input signal to differential amplifier 38. Amplifier 38 saturates in the negative direction when the reference voltage is greater than the voltage across capacitor 35 and saturates in the positive direction when the reference voltage is less than the voltage across capacitor 35. v

The voltage across capacitor 35 is also coupled to th emitter of unijunction transistor 37. Unijunction transistor 37 is biased so that it breaks down, due to the voltage across capacitor 35, at the pitch pulse rate thereby producing pitch pulses across capacitor 39. These pulses are standardized in the single shot multivibrator 41.

The output of amplifier 38 is supplied to'the gate of a field effect transistor 40 and to a voltage inverter circuit 42. The source electrode of field effect transistor 40 is coupled to an input of a controllable loss compensating operational amplifier 44. The resistance of transistor 40 and a resistor 43 form the feedback loop for operational amplifier 44 and thus regulate the gain of amplifier 44.

The-first formant frequency signal F is supplied to a field effect transistor 45 of a voltage-variable high pass filter circuit 46 and to a field efiect transistor 47 of a voltage variable low pass filter circuit 48by means of potentiometer 50. The filter circuits are isolated by amplifier network 52. The source electrode of field effect transistor 45 is coupled to a second input of operational amplifier 44. The output of amplifier'44 is coupled to switch modulator 14 of FIG. 1 and through an isolating emitter-followernetwork 54 to the source electrode of field effect transistor 47. The damped first formant signal appears at the output of amplifier 44.

The operation of the circuit of FIG. 3 is as follows: The amplitude of the pitch signal determines the collector current of transistor 33 and hence the charging rate of capacitor 35.

When the voltage across capacitor- 35 approaches the- The high amplitude output signal (positive saturation) of amplifier 38 increases the resistance of transistor 40. Since the gain of the resonant circuit, which receives ringing energy from modulator 14, is inversely proportional to the resistance of field effect transistor 40, the gain of the resonant circuit is reduced just prior to each pitch pulse. Circuit 42 neutralizes transients resulting from the source-gate capacitance of field effect transistor 40. The formant output signal is accordingly damped just prior to each successive pitch pulse and thus the output signal is substantially a replica of that produced by the human vocal tract.

While the invention has been described with reference to a particular preferred embodiment thereof, it will be apparent that various modifications and other embodiments thereof will occur to those skilled in the art within the scope of the invention. Accordingly we desire the scope of our invention to be limited only by the appended claims.

We claim:

1. In a system for reproducing a signal representative of a speech wave, a speech synthesizer'comprising:

a plurality of resonator circuits, at least one of said resonator circuits comprising amplifier means, the bandwidth of said one of said resonator circuits being dependent on the gain of said amplifier means;

first means coupled to each of said plurality of resonator circuits for regulating the resonantfrequency of said resonator circuits; 1 second means coupled to each of said plurality of resonator circuits for supplying ringing energy to said resonator circuits upon the occurrence of each of a succession of pulses respectively representative of successive pitch pulses of said speech wave; and

damping means coupled to said one of said resonator circuits,'for controlling the bandwidth of said one resonator circuit, the improvement wherein said damping means comprises:

third means for controlling the gain of said amplifier means, said third means being responsive to a control signal to exhibit between a pair of terminals of said third means a first resistance when said control signal has a first value anda second resistance when said control signal has a second has a second value,

means coupling said pair of terminals of said third means to said amplifier means, said amplifier having a first gain when said third means exhibits said first resistance between said pair of terminals and having a second gain lower than said first gain when third means exhibits said second resistance between said pair of terminals,

a differential amplifier for supplying to said third means, in response to an input signal of given magnitude, a control signal having said first value, and for supplying to said third means, in response to an input signal of another magnitude different from said given magnitude, a control signal having said second value, and

means for supplying to said differential amplifier an input signal having said other magnitude at a first time following the occurrence of the pulse representative of one of said pitch pulses and changing to said given magnitude at a second time before the occurrence of the pulse representative of the succeeding one of said pitch pulses, whereby said bandwidth of said one resonator.

circuit is widened between said wave: time and said second time. v 2. in a system for reproducing a signal representative of a speech wave: I

a speech analyzer comprising first means for producing in response to said speech wave signals representative of the frequencies and amplitudes of a plurality of formants of said speech wave, and second means for producing in response to said speech wave signals representative of the pitch rate of said speech wave; a speech synthesizer comprising oscillator means responsive to said signals produced by said second means to generate a succession of voltage pulses at said pitch rate, a plurality of modulators, third means for coupling said modulators in parallel to said oscillators means, fourth means for supplying each of said signals respectively representative of the amplitudes of said plurality of formants of said speech wave to a difierent one of said modulators, a plurality of resonator circuits each having first and second input terminals, at least one of said resonator circuits having amplifier means, the bandwidth of said one of said resonator circuits being dependent on the gain of said amplifier means, each of said plurality of resonator circuits having said first input terminal thereof coupled to a different one of said modulators, fifth 'means for supplying each of said plurality of signals respectively representative of the sixth means for controlling the gain ofsaid amplifier means, said sixth means being responsive to a control signal to exhibit between a pair of terminals of said sixth means a first resistance when said control signal has a first value and a second resistance when said control signal has a second 'value,

means coupling said pair of terminals of said sixth means to said amplifier means, said amplifier means having a first gain when said sixth means exhibits said first resistance between said pair of terminals and having a second gain lower than said first gain when said sixth means exhibits said second resistance between said pair to said sixth means, in response to an input signal of another magnitude different from said given mag- 3. The system of claim 2 wherein said analyzer includes means for producing a voicing signal, said synthesizer includes a noise generator, and said third means includes a gating circuit controlled by said voicing signal and coupled to said oscillator means and said noise generator so that, depending upon the amplitude of the voicing signal, either the output signal of said noise generator or the output signal of said oscillator means is coupled to said modulators.

CERTIFICATE OF CORRECTION Patent No. 3,5733? Dated May 97 humor) Louis R. Focht and James M. Lee

It is certified thlt error eppure 1n the above-identified patent and (hit and Letter. Plum: en hereby corrected as shown below:

Claim 1, line &5, "between said wave: time" shouldbe between said first time Signed and sealed this 13th day of July 1971.

(SEAL) Attest:

EDJARD M.FLETCHER, JR. WILLIAM E. SCHUYLER, JR. Attesting Officer Oomissioner of Patents

Claims

1. In a system for reproducing a signal representative of a speech wave, a speech synthesizer comprising: a plurality of resonator circuits, at least one of said resonator circuits comprising amplifier means, the bandwidth of said one of said resonator circuits being dependent on the gain of said amplifier means; first means coupled to each of said plurality of resonator circuits for regulating the resonant frequency of said resonator circuits; second means coupled to each of said plurality of resonator circuits for supplying ringing energy to said resonator circuits upon the occurrence of each of a succession of pulses respectively representative of successive pitch pulses of said speech wave; and damping means coupled to said one of said resonator circuits, for controlling the bandwidth of said one resonator circuit, the improvement wherein said damping means comprises: third means for controlling the gain of said amplifier means, said third means being responsive to a control signal to exhibit between a pair of terminals of said third means a first resistance when said control signal has a first value and a second resistance when said control signal has a second has a second value, means coupling said pair of terminals of said third means to said amplifier means, said amplifier having a first gain when said third means exhibits said first resistance between said pair of terminals and having a second gain lower than said first gain when third means exhibits said second resistance between said pair of terminals, a differential amplifier for supplying to said third means, in response to an input signal of given magnitude, a control signal having said first value, and for supplying to said third means, in response to an input signal of another magnitude different from said given magnitude, a control signal having said second value, and means for supplying to said differential amplifier an input signal having said other magnitude at a first time following the occurrence of the pulse representative of one of said pitch pulses and changing to said given magnitude at a second time before the occurrence of the pulse representative of the succeeding one of said pitch pulses, whereby said bandwidth of said one resonator circuit is widened between said wave: time and said second time.

2. In a system for reproducing a signAl representative of a speech wave: a speech analyzer comprising first means for producing in response to said speech wave signals representative of the frequencies and amplitudes of a plurality of formants of said speech wave, and second means for producing in response to said speech wave signals representative of the pitch rate of said speech wave; a speech synthesizer comprising oscillator means responsive to said signals produced by said second means to generate a succession of voltage pulses at said pitch rate, a plurality of modulators, third means for coupling said modulators in parallel to said oscillators means, fourth means for supplying each of said signals respectively representative of the amplitudes of said plurality of formants of said speech wave to a different one of said modulators, a plurality of resonator circuits each having first and second input terminals, at least one of said resonator circuits having amplifier means, the bandwidth of said one of said resonator circuits being dependent on the gain of said amplifier means, each of said plurality of resonator circuits having said first input terminal thereof coupled to a different one of said modulators, fifth means for supplying each of said plurality of signals respectively representative of the frequencies of said formants to said second input terminal of a different one of said resonator circuits, and damping means coupled to said one of said resonator circuits, for controlling the bandwidth of said one resonator circuit, the improvement wherein said damping means comprises: sixth means for controlling the gain of said amplifier means, said sixth means being responsive to a control signal to exhibit between a pair of terminals of said sixth means a first resistance when said control signal has a first value and a second resistance when said control signal has a second value, means coupling said pair of terminals of said sixth means to said amplifier means, said amplifier means having a first gain when said sixth means exhibits said first resistance between said pair of terminals and having a second gain lower than said first gain when said sixth means exhibits said second resistance between said pair of terminals, a differential amplifier for supplying to said sixth means, in response to an input signal of given magnitude, a control signal having said first value, and for supplying to said sixth means, in response to an input signal of another magnitude different from said given magnitude, a control signal having said second value, and means for supplying to said differential amplifier an input signal having said other magnitude at a first time following the occurrence of one of said voltage pulses and changing to said given magnitude at a second time before the occurrence of a succeeding one of said voltage pulses, whereby said bandwidth of said one resonator circuit is widened between said first time and said second time.

3. The system of claim 2 wherein said analyzer includes means for producing a voicing signal, said synthesizer includes a noise generator, and said third means includes a gating circuit controlled by said voicing signal and coupled to said oscillator means and said noise generator so that, depending upon the amplitude of the voicing signal, either the output signal of said noise generator or the output signal of said oscillator means is coupled to said modulators.