US3652801A

US3652801A - Circuit arrangement for synthesis of acoustic elements

Info

Publication number: US3652801A
Application number: US813948A
Authority: US
Inventors: Walter Tscheschner; Siegfried H E Fuchs
Original assignee: Elektronische Rechenmaschinen Wissenschaftlicher Industriebetrieb VEB
Current assignee: Elektronische Rechenmaschinen Wissenschaftlicher Industriebetrieb VEB
Priority date: 1969-04-07
Filing date: 1969-04-07
Publication date: 1972-03-28
Anticipated expiration: 1989-03-28

Abstract

The artificial speech initiating electrical signal is in the form of periodic triangular vibrations or of a noise signal. The signal is selectively applied through a coordinate control matrix to input or excitation points of a conventional speech synthetizing potential field device which distributes the potentials of the signal along a number of output electrodes. Each of the electrodes is connected with a band-pass filter, respectively, and the output spectral portions are combined into the shaped spectrum of the desired articulated sound.

Description

United States Patent Tscheschner et al.

CIRCUIT ARRANGEMENT FOR 51 Mar. 28, 1972 3,127,476 3/1964 David ..179/1 SYNTHESIS OF ACOUSTIC ELEMENTS pfimary Examiner xathleen H Clam, [72] inventors: Walter Tscheschner, Dresden; Siegfried R. Assistant Examiner-Douglas W. Olms E. Fuchs, Coswig, both of Germany Attorney-Nolte and Nolte [73] Assignee: Veb Elektronische Rechenmaschinen, Karl- [57] ABSTRACT Marx-Stadt, Germany The artificial speech initiating electrical signal is in the form of [221 Med: 1969 periodic triangular vibrations or of a noise signal. The signal is [2 1 Appl 3,948 selectively applied through a coordinate control matrix to input or excitation points of a conventional speech synthetizing potential field devicewhich distributes the potentials of [52] U.S. Cl. ..l79/l SA the signal along a number of output electrodes. Each of the [51] int. Cl. ..Gl0l 1/00 electrodes is connected with a band-pass filter, respectively, [58] Field of Search 179/ 1 SA and the output spectral portions are combined into the shaped spectrum of the desired articulated sound.

[56] References Cited u 3 Claims, 4 Drawing Figures UNITED STATES PATENTS 2,817,711 12/1957 Feldrnan ..l79/l 2 7| m l O0 2 R A H 0 EL E T .A P

SHEET 1 8T 2 n T 8 m 1. R u m E P 8 .1 6 R 55 LW 7 5 mm m T G M 2 m mm C LM w E A m w G E L llll l! L. E m f m 0 L m 2 mm 1 PU C EM I N R R P ARTICULATION GENERATOR COORDINATION CE NTER TRIANGULAR WAVEFORM GENERATOR FILTER SYSTEM INVENTORS F NOISE GENERATOR gSPE/AKER WALTER TSCHESCHNER SlEGFRlED FUCHS m E z M 2 L Ow WT NS N ML 7 wm WM mm DO 5 FF 0 0m 5 E H CC R 9 UK OP T SS R .A R m R P ATTORNEYS CIRCUIT ARRANGEMENT FOR SYNTHESIS OF ACOUSTIC ELEMENTS The present invention relates to a circuit arrangement for the production of artificial speech.

The essential peculiarity of organic processing of acoustic elements is that the acoustic elements which serve 'for human communication have a relatively high degree of accommodation to a particular organic process of conversion. As a consequence, a highly reliable identification is provided having the capability of forming numerous combinations of these elements.

Due to the very low degree of accommodation to the peculiarities of human communication which is attainable in existing technical systems, it is very difficult to make the systems for identifying and indicating the acoustic elements cooperate to a .greater extent with the devicesfor data processing, communication compression systems, devices for rehabilitation and for the like applications.

Until now, there has been no satisfactory solution of this problem. Even the devices which are extremely costly have a very limited performance.

A substantial shortcoming of all of the known speech synthesizers is that the processes for shaping acoustic elements along the amplitude, frequency and time axes are not, due to the physical limitations of existing technical systems, controllable to the degree necessary for the articulation and especially, for the perception process. At the same time, it is very difficult to make a technical model having properties and control structures of organic partial systems and the complexity of the entire organic system.

FIG. 1 shows schematically in a block-circuit diagram an arrangement for the organic generation of acoustic elements originating, as rule, from discrete information. The information excites the independently operating generators and coordinates the cooperation of the generators according to the intended effect via process controlling systems.

Whereas in organic systems the individual physiological organs for controlling positions, such as, for example, positions of the jaw or tongue, or protrusion and the like, influence the peripheral physical system and the action thereof such as,'for instance, the formant frequencies of the first up to fourth formants, the band-widths of the first to fourth formants, and the amplitudes of the formants, it is necessary in technical models to individually modify each of those parameters to attain analogous effects.

It is, therefore, indispensible in synthetizing systems operating according to a channel-vocoder principle to apply great quantity of controls. .ln systems having glide band-pass filters and operating on the formant-vocoder principle, the reproduction of dynamic processes is complicated as a result of the employment of the plurality of controls and also due to the physically conditioned persistency of the system. Ifitis attempted in such systems to adjust the physically conditioned relationships in order to simplify the control action, it will be found that the functional relations which are required for the improvement of communication cannot be accomplished.

Considerable advantages for speech forming composition of control data can be attained by means of a potential field device coupled with a channel system. A potential field device or potential distribution device as hereinafter discussed is comprised of an electrically conductive medium having a plurality of regularly arrayed spaced apart inputelectrodes connected to the medium, and a plurality of spaced apart output electrodes also connected to the medium. In this type of device, the potential at any given output electrode is a function of potential applied to the input electrodes as well as the geometry of the medium between the output electrode and the input electrodes to which potentials are applied. The potential field device or potential field matrix operates similarly as the so-called electrolytic tank in which a plurality of input electrodes or excitation points are regularly arrayed in an electrically conducting medium. Respective excitation points within the electrolytic tank or other potential distributing means are connected to a switch means for selectively applying electric potentials thereto. The polarity of those potentials is applied via a plurality of output electrodes operatively disposed at-the device to control modulators forming the frequency envelope of individual sounds. TI-le potential field device makes it possible to provide both the composition of control data andsuch physical limitconditions in the frequency and time axes which result in an excellent reproduction of speech. However, it has the disadvantage that the amplitude and time conditions thereof are subject to unfavorable limitationsof physical nature and also that the unavoidable modulators due to their relatively very small operative range and unstability in time, may'reduce the quality of speech.

Therefore, the object of the present invention is to provide a circuit arrangement for a speech synthesis which avoids the aforementioned disadvantages of the systems operating according to a channel or formant synthesis principle.

Moreparticularly, the object of the present invention is to provide a circuitarrangement in which'the technical parameters can be improved in such a manner that the artificial speech signals can be readily and easily produced.

In accordance with the present invention the speech stimulating signal which is applied to excitation points of a conventional potential field device, has the form of periodic triangular vibrations or of a noise signal. A n outputs of the potential field, there are coupled n band-pass filters, respectively, through which separate information pertaining :to each of the sound elements are composed into acoustic spectrum of articulated andacoustically modified sounds.

By superposing a plurality of'potential field devices,'theamplitudes of source currents or voltages can be processed in uniformlydesigned channel stages and the limiting conditions can be compensated to obtain more suitable form.

The invention has the advantage that a number of very critical structural stages can be avoided, and the arrangementof the potential field can be still better adjusted to operatingconditions of organic information processing systems.

The presentinvention will nowbe explained in more-detail by way of an example with reference to accompanying drawings, in which:

FlG. 1 is a block-circuit diagram of an organic speech transmitting system;

FIG. 2 is a blockdiagram of atechnical speech synthetizer;

FIG. 3 ,is a graph showingthe envelope of the sound'a (Curve l4) and of a simulated function of a Curve I5);-and

FIG. 4 is a schematic diagram of an embodiment of .the speechsynthetizer according to the present invention.

FIG. 1 shows a block-circuit diagram of an organic speech transmitting system which usually serves as a modelfor technical embodiments. The speech forming process is initiatedby selecting a series of elementary phonological units -l-whichare mostly superimposed with super-segmental units and revoke motive programs 2 throughwhich therealization of the speech is effected in

peripheral generator systems

5,6 and 7. The control commands for lungs generator 5, vocal generator 6 and articulation generator 7 are released 'by neuro-motive coordination centers 3 whereby the execution of theientire speech realization is mainlyperformcd by neuro-motive feedback 4 originating at tactile and acoustic sensing organs.

'The actual technical synthesizers'have-a substantially'lower degree of complexity. Also, the mutual control of partial systernsis relativelydecreased. In the majority of technical synthesizers (as seen in FIG. 2) the-time'functions'of spectra of naturalspeech are simulated by means of filter system :13 controlled according to a time function for providinga spectral envelope with variable formants. Control system 10 determines the-excitation of noise signals'from noise generator 11 for .unvoiced sounds or consonants, or the excitationof triangularperiodic functions from triangular-waveform generator 12 for voiced sounds, thus determining the position and shape of formants for an articulated sound. The:input of control system 10 can be supplied with results of a continuous physical speech-analysis (vocoder), with a'seriesof vocal demonstrations (voder), with digital information (phonetic output),

with a series of sounds (from a speech machine) or with any other derived speech information 9. There must always be provided a number of various control signals to accomplish the synthesis of speech signals in the given synthesizer system, such as a signal for the actuation of triangular-waveform generators for producing the vocal articulated sounds, a signal for the actuation of noise generators for producing the unvoiced sounds, and further signals for the adjustment of frequency of the triangular-waveform generators. Besides, there must also be provided information of the position, bandwidth and the amplitude of formants.

With reference to the evaluation of discrete singular points of envelope curve 14 of the instantaneous spectrum as seen in FIG. 3, at least three formants are necessary, i.e., six formant control signals, eventually as many channel control signals as to enable a sound coordination according to the produced formant positions. As to the shape of function 15 defining the sound perception, at least ten channels must be controlled or, eventually, as the case may be in the controlling of a potential field device, two pole sites controlled along two coordinate axes.

FIG. 4 illustrates schematically the principle of a circuit arrangement according to this invention for accomplishing a synthesis of vocal sounds at one pole site. Although the embodiment of FIG. 4 relates to the formation of a shape of a single spectral envelope, it can be employed also for the production of a number of formant positions. The trianglelike speech excitation signal from the triangle signal generator 12 is applied through resistor network 17 to coordinate control field or matrix 19 which defines the position of actuated excitation points 27 in potential field device 20. According to the position of actuated excitation points, a potential distribution of the triangular-waveform voltage at the n-number of output electrodes 25 along the X-axis of potential field device 20 is produced as a function of the position of the actuated excitation points in the potential field device 20.

The positions of the maxima of the potential distribution 0 the compound triangular signal are, therefore, determined by positions of the excitation points on the x-axis and the shape thereof is defined by the position of the excitation points of the y axis. A separate filter 22a to 22!: of filter set 22 is connected to each of the n electrodes 25. The filters of filter set 22 always select the magnitude of the corresponding spectral portions of the broad band frequency spectrum as determined by the potential field device 20. Consequently, the combined signal voltages at the output 26 of filter set 22 form the synthesized sound spectrum which may be amplified in amplifier 23 and reproduced by speaker 24.

Since respective filters 22a to 2221 in filter set 22 are arranged in consecutive order according to corresponding frequency groups or bands, the spectrum is in a form which is suitable for the process of acoustic resolution and evaluation.

Coordinate control matrix 19 can be designed in various manners, either electronically or by means of electromechanical coordinate switches. The preferred embodiment is based on the electronic solution, in which two counters or

selectors

18 and 18 are in operative relation to respective column and row conduits 2811 and 29n arrayed in a matrix-like pattern. Respective points of intersection of the conduits are coupled to the excitation points and corelated output electrodes of the potential field device through a plurality of diode switching means 16 which actuate the electrodes in response to the action of selectors 18.

Resistors 17 at inputs of coordinate control matrix 19 and resistors 21 between respective output electrodes of the potential field device and the filter set 22 modify in accordance with the distribution of their values the potentials at respective terminals along the X and Y coordinates to compensate technical limit conditions of the entire circuit arrangement. Furthermore, by means of resistors 17 and 21 it is possible to simulate in the potential field device the position function of the formants as predetermined by various configurations of the spectrum of an articulated sound. For a consecutive amphtu e-position potentlal field device (not shown), re-

sistors l7 and 21 can be combined to substitute the resistor set 17. The application of a second amplitude-position potential field device is particularly advantageous when different modes of the speech excitation require different nonlinear amplitudeposition functions. The coupling of the second potential field device to the first one is accomplished by means of coordinates control matrix similar to the matrix 19.

What is claimed is:

l. A circuit for the production of artificial speech from acoustic elements comprising signal generator means for supplying a signal having a wide spectrum of frequencies and amplitudes, coordinate control matrix means having a plurality of input conduits and a number of intersection points, signal selector means for connecting said input conduits to said generator means, a potential field device comprised of an electrically conductive medium having a plurality of regularly arrayed spaced apart input electrodes and a plurality of spaced apart output electrodes connected thereto, whereby the potential at a given output electrode is a function of potentials applied to said input electrodes and the geometry of said medium between said output electrode and input electrodes to which potentials are applied, means coupling said input electrodes to said intersection points for providing a potential distribution of said signal along the output electrodes in response to the signals applied to said input conduit by way of said selector means, a plurality of band-pass filter means coupled to said output electrodes to select the magnitude of a spectral portion of the signal as determined by the action of said potential field device, and means for coupling the outputs of respective filter means to produce a synthesized sound spectrum.

2. The circuit according to claim 1 wherein said signal generator means is a noise signal generator.

3. The circuit according to claim 1 wherein said signal generator means is triangle-waveform signal generator.

Claims

1. A circuit for the production of artificial speech from acoustic elements comprising signal generator means for supplying a signal having a wide spectrum of frequencies and amplitudes, coordinate control matrix means having a plurality of input conduits and a number of intersection points, signal selector means for connecting said input conduits to said generator means, a potential field device comprised of an electrically conductive medium having a plurality of regularly arrayed spaced apart input electrodes and a plurality of spaced apart output electrodes connected thereto, whereby the potential at a given output electrode is a function of potentials applied to said input electrodes and the geometry of said medium between said output electrode and input electrodes to which potentials are applied, means coupling said input electrodes to said intersection points for providing a potential distribution of said signal along the output electrodes in response to the signals applied to said input conduit by way of said selector means, a plurality of bandpass filter means coupled to said output electrodes to select the magnitude of a spectral portion of the signal as determined by the action of said potential field device, and means for coupling the outputs of respective filter means to produce a synthesized sound spectrum.

3. The circuit according to claim 1 wherein said signal generator means is a triangle-waveform signal generator.