DE2227476B1 - PROCEDURE AND ARRANGEMENT FOR GENERATING QUASI-SYNTHETIC LANGUAGE IN A VOCODER - Google Patents

Description

From K. Steinbuch in »Taschenbuch der Nachrichtenverarbeitung«, Springer Verlag Berlin, 1962, pp. 851 ff., Along with a number of others under the name »Vocoder« of well-known devices for the generation of quasi-synthetic speech also described the channel vocoder. With one of the that this flip-flop stage also still as start impulses from the basic speech frequency generator or the generated by the noise generator and selected by the voiced-unvoiced signal Pulses are supplied, whereby the time interval between the start pulses and the stop pulse series is doubled from row to row, and that from the output of the Flip-flop stage of the permanently modulated Kanalpuis can be removed. For further training this arrangement will produce the pulses with the various pulse widths required generated directly by a central generator. There is an arrangement at the in the channel synthesis units a buffer with one as a selector acting switching arrangement is connected, in which a central generator is provided which generates several series of pulses, the duration of the pulses varying from Row to row doubled and each of these pulse rows has a separate output of the generator is available in which these outputs are connected to the selector are and at the output of the selector the continuously modulated Kanalpuis directly or via an adjustment stage can be found. This arrangement results in addition to the easy Verifiability and the clarity a particularly large space saving.

With reference to the embodiments shown in the drawing the invention is explained in more detail below. In the drawing, Fig. 1 a channel synthesis unit with a preset counter and a 3-bit / 7-bit code converter according to the invention, Fig.2 shows an arrangement according to the invention, which has a preset Contains counters, a central code converter and pulse generators, F i g. 3 a Channel synthesis unit with a switching arrangement that selects stop pulses, FIG. 4 is a diagram of FIG. 4 in the arrangement according to FIG. 3 occurring pulses, F i g. 5 a channel synthesis unit according to the invention dung with a switching arrangement that selects certain impulses from centrally generated impulses of different duration, and Fig. 6 is a diagram of these pulses.

The channel synthesis unit shown in FIG. 1 consists of a buffer 1, the channel information C, B, A for at least one digitally stores a sampling cycle lasting about 20 milliseconds and when it occurs of a transfer pulse ÜP at input 11, the stored one labeled C *, B *, A * Forwards information to a code converter 2, which converts the 3-bit code into a 7-bit code converts the compression of the 7-bit code into cancel a 3-bit code; a counter 4 with a preset 3 and a NAND gate 5. A start pulse SI that occurs at 12 causes that by default 3 those counter levels are set to 0 for which is offered by the code converter a 1. At the output 14 of the NAND gate 5 appears then also a logical 1, on the one hand a pulse PDM of the basic speech frequency represents and accordingly the band filter, if necessary via an amplifier stage, is supplied, but on the other hand to the gate input 13 for the counting pulses Zl is returned. The counting process is interrupted as soon as all counter levels are at logical 1.

The start pulses SI supplied at 12 of the presetting are, according to the position of a switch controlled by a pitch signal, the pulses of either a noise generator or a fundamental speech frequency generator known per se. The channel pulse fed to the filters is a narrow square pulse, the maximum duration of which is to be determined by a number of considerations, both in the case of the pulse amplitude modulation used hitherto and in the case of the modulation of the pulse duration according to the invention. The expression for the Fourier series of a square pulse is: where Ky = TV / To = Tv f, the duty cycle and n = f / fo (fo = basic speech frequency, Ty = pulse width).

For frequencies at which this expression deviates sufficiently little from 1, is the harmonic spectrum. The pulse duration may therefore have a certain value do not exceed. This value results from the highest to be processed Frequency f / l. If one tolerates a drop in the intensity of the highest harmonic the amount at 1.75 db, the pulse width may not exceed TV = 64 ys for fh = 5 kHz be.

The frequency of the counting pulse is in the voiced speech segments and the lowest basic speech frequency 1 milz. The power to be transmitted increases is proportional to the square root of the basic speech frequency and is due to a decrease the pulse width Tv by increasing the Compensate for counting frequency fz h fo. At a This readjustment of the power is a relatively small range of the basic speech frequency without neglecting major disadvantages.

In contrast, in most cases it is necessary for the synthesis to adapt the power of the noise impulses of the voiceless speech segments.

It is If the clock frequency of the noise pulse is about twice as high as the highest transmission-type channel vocoder, the speech signal on the transmission side, i.e. in the analysis part, is broken down into a number of spectral channels by a series of bandpass filters with adjoining pass bands and the envelope curve of these spectral channels is determined. Furthermore, a frequency meter determines the level of the basic speech frequency, also known under the term “pitch frequency”. The data determined are fed to the transmission channel or a storage device with the aid of a multiplex device known in transmission technology. On the receiving or output side, i.e. in the synthesis part of the vocoder, there is a separation of the individual parameter signals with regard to the level of the fundamental frequency and the decision regarding the voiced or unvoiced character of the sound as well as with regard to the amplitudes of the individual channels, whereby the pitch signal is occasionally used in the literature is also referred to as a voice signal. In a switching and control stage, depending on whether it is an unvoiced or a voiced sound, either a noise generator or a pulse generator serving as a fundamental speech frequency generator is switched on. The emitted noise pulses or the basic speech frequency impulses are then distributed to the individual spectral channels, passed through bandpasses tuned to a certain frequency range, amplitude-modulated with the information of the respective envelope curve and, after addition, form the synthetic speech signal. It is also common practice to modify this method so that the amplitude modulation occurs before the signals pass the bandpasses.

By dividing the voice band into a number of around 10 to 16 spectral channels result in a number of functional groups for each channel the synthesis side of the vocoder are available separately and as "channel synthesis units" are designated.

A channel synthesis unit has a number of individual functions. First the digital information of the channel is stored for one sampling cycle.

The digital signal is then converted into an analog one Signal, for example a stream.

This analog current then modulates that from the noise generator noise impulses originating or the basic speech frequency impulses originating from the basic speech frequency generator.

The arrangements according to the state of the art are in all cases to arrangements in which the fundamental frequency pulses or noise pulses experience amplitude modulation.

This results in a number of disadvantages, since amplitude-modulated when used Signals have special requirements for the accuracy and the constancy of the amplification the corresponding devices are to be provided, the further requirements with regard to Temperature stability, precise adjustment and testing in the form of an exact measurement entail.

The invention is based on the object of a method and implementation of the procedure to develop suitable arrangements for which no special requirements the accuracy and constancy of the amplification of the devices must be ensured, which are also easy to check and which take up as little space as possible.

The object is thereby achieved according to the invention resolved that duration the pulses are modulated.

The invention is based on the knowledge that the conversion the application of digital technology, in particular binary technology, to pulse duration modulation and thus the use of integrated digital circuits is possible. Through this result from the implementation of the method with regard to the insensitivity to interference and the stability have particular advantages. Is particularly advantageous in arrangements to carry out the method according to the invention, the small space requirement and the significantly improved testability. In addition, there is an integration of the arrangements to so-called MSI or LSI modules, which take up even less space, easily possible.

An expedient arrangement for carrying out the method according to the Invention includes a basic speech frequency generator, a noise generator, a Clock generator, bandpass filters, an adder circuit and channel synthesis units with one buffer memory and one code converter each, with each channel synthesis unit the code converter is followed by a counter with a preset and the individual Counter stages are connected to the inputs of a NAND gate, from its output a return to the counter is provided and also the duration-modulated channel pulse is removable.

A significant simplification and reduction of the effort results by combining parts of channel synthesis units into central units. The use of a central code converter has a particular effect on vocoders high number of channels and results in an arrangement for carrying out the method, which have a basic speech frequency generator, a noise generator, a clock generator, Bandpasses, an adder circuit and several channel synthesis units, each with one Contains intermediate memory in the case of the channel synthesis units on the intermediate memory A simplified counter follows with a pre-setting in which the individual Counter stages are connected to the inputs of a NAND gate, of which Output a return to the counter is provided and from which the duration modulated Channel pulse can be taken, in which depending on a voiced Stimilos signal a central code converter on the one hand either with a central pulse generator via a frequency divider or directly and, on the other hand, either with a basic speech frequency generator or a noise generator is connected and the output of the central Code converter is connected to the counting pulse input of the counter via a gate.

A further simplification results in an arrangement according to the invention, which contains a basic speech frequency generator, a noise generator, a clock generator, bandpass filters, an adding circuit and several channel synthesis units, each with a buffer, in that the channel Synthesis units the buffer is connected to a switching arrangement acting as a selector, to which the pulse series generated by a central stop pulse generator are also fed via separate lines, so that one of the lines of the stop pulse generator is connected to a flip-flop stage through the selector, frequency must be. With a basic speech frequency f0 = 100 Hz and a highest transmission frequency f. = 10 kHz this results In this case it is also advantageous to use a counting pulse generator with a pulse frequency of about 5 MHz for the unvoiced sounds, which is connected for the voiced sounds to the input 13 of the counter for the counting pulses via a frequency divider dividing in a ratio of 5: 1 is.

In FIG. 2, in addition to the actual channel unit, there are further ones Function groups shown.

A buffer 1 also controls a binary counter 41 directly a presetting 31. The counter has only three levels, as it is a central code converter 22 is provided, from which the counting pulses Zl reach the counter 41 via the gate 13. The code converter generates from the periodic, from a 5 MHz generator 20 over a 5: 1 divider 21 and the switch S1 reached the input of the code converter 1 MHz pulses or a series of e.g. B. seven pulses, whose distance to the start impulse doubles in each case. Act as start impulses Sl while the voice pulse generator 24 or the noise generator 25 via the Switch S2 to the central code converter 22 received pulses. The switches S 1 and S2 are controlled by the voiced / unvoiced signal TS via line 23. The continuously modulated channel pulse PDM becomes the relevant one via an amplifier stage 91 Bandpass passed.

In FIG. 3 is a channel synthesis unit according to the invention shown, which from the buffer 1 for the information C, B, A, one as Selector operated switching arrangement 6 and a flip-flop stage 7 consists. at the switching arrangement 6 is a commercially available data selector switch module, which is available in integrated form under the designations SN 74151 N and FLY 121 is. The central Generator generated stop pulse series P1 to P 8 applied, the shown in FIG are. The stop pulses P1 correspond to those in the circuit according to FIG. 2 from the central Code converter emitted [pulses; they are in time according to the selected compander characteristic staggered and are started by a basic speech frequency or a noise pulse.

The switching arrangement 6 selects according to that of the buffer retrieved information one of the lines 61 to 68 and thus one of the stop pulse series off and switches the stop pulse via the output 69 to the flip-flop stage 7, the The start pulses S1 are also supplied via the line 71. The channel synthesis unit according to FIG. 3 works in start-stop mode. The modulated in its duration Channel pulse PDM is switched from the flip-flop stage to a filter matching stage or directly routed to a bandpass filter.

The in the F i g. 5 shown circuit of a channel synthesis unit consists of the buffer 1, one controlled by this, also as Selector operated switching arrangement 6 and an amplifier stage 92. In contrast for the circuit according to FIG. 3 are used in the circuit according to FIG. 5 not the stop pulses for a flip-flop stage are generated centrally, but the one in FIG. 6 pulses shown T 1 ... T 7 with different pulse widths, so that central all required pulse widths are available. From the switching arrangement 6, for which is the same component as that in the circuit according to FIG. 3 can be used can, according to the information C *, B *. A * the channel pulse PDM, modulated in its duration, directly via output 69 connected to the amplifier stage 92. which causes the adaptation to the bandpass filter.

The method according to the invention is also useful as a decoding method in PCM systems, especially those that work with very low bit rates, for example in pulse delta modulation systems. In such a case the pitch pulse is replaced by the sampling pulse of the PCM or pulse delta modulation system.

Furthermore, the inventive method is advantageous in one Can be used in combination with pulse amplitude modulation. It is then e.g.

the supply voltage of the amplifier stage 92 in FIG. 5 in their Amplitude changed.

Claims (5)

Claims: 1. Method for generating quasi-synthetic speech in a vocoder with several parallel channels of parameter signals Information about the envelope signals of the individual channels, the basic speech frequency and their amplitude as well as the voiced or unvoiced sounds for which at voice-like sounds are the impulses generated in a fundamental speech frequency generator and in the case of unvoiced sounds, the pulses generated in a noise generator, channel synthesis units are supplied, in which these pulses with the envelope signal of the respective Channel modulated and finally added, and either the modulated or the unmodulated pulses pass through bandpass filters, characterized in that the duration of the impulses is modulated. 2. Arrangement for performing the method according to claim 1, containing a basic speech frequency generator, a noise generator, a clock generator, Bandpasses, an adder circuit and several channel synthesis units, each with one Intermediate memory and one code converter each, characterized in that in the channel synthesis unit the code converter (2) is followed by a counter (4) with a presetting (3) and that the individual counter stages are connected to the inputs of a NAND gate (5), from the output of which a return to the counter (4) is provided and from which the duration-modulated channel pulse (PDM) can be removed (Fig. 1). 3. Arrangement for carrying out the process according to claim 1, containing a basic speech frequency generator, a noise generator, a clock generator, Bandpasses, an adder circuit and several channel synthesis units, each with one Intermediate memory, characterized in that in the channel synthesis unit on the Buffer (1) a simplified counter (41) with a preset (31) it follows that the individual counter stages with the inputs of a NAND gate (51) are connected, from the output of which a return to the counter (41) is provided and from which the duration-modulated channel pulse (PDM) can be taken that as a function a central code converter (22) on the one hand from a voiced / unvoiced signal (TS) with a central pulse generator (20) either via a frequency divider (21) or else directly and on the other hand either with a basic speech frequency generator (24) or a noise generator (25) is connected and that the output of the central code converter (22) via a gate (13) to the counting pulse input of the counter (41) is connected (Fig. 2). 4. Arrangement for carrying out the process according to claim 1, containing a basic speech frequency generator, a noise generator, a clock generator, Bandpasses, an adder circuit with several channel synthesis units and a buffer, characterized in that the buffer memory in the channel synthesis units (1) is connected to a switching arrangement (6) acting as a selector, the aside from that via separate lines (61 ... 68) from a central stop pulse generator generated pulse series (P1 ... P8) are supplied that through the selector a of the lines of the stop pulse generator connected to a flip-flop stage (7) is that this flip-flop stage also still as start pulses (SI) from the Basic speech frequency generator or those generated by the noise generator and by the voiced-unvoiced signal are supplied to selected pulses, whereby the time interval between the noise or noise the one generated by the basic speech frequency generator acting as a start impulse Pulses and the stop pulse series doubled from row to row, and that at the output the permanently modulated channel pulse (PDM) can be taken from the flip-flop stage (FIG. 3 and 4). 5. Arrangement according to claim 4, characterized in that in the Channel synthesis units have a buffer (1) with one acting as a selector Switching arrangement (6) is connected that a central generator is provided, the several series of pulses ...... T 7) generated, the duration of these pulses doubled from row to row, that one pulse row each via a separate output of the generator is available that these outputs are connected to the selector and the continuously modulated channel pulse (PDM) directly at the output of the selector or can be removed via an adapter stage (92) (FIGS. 5 and 6). The invention relates to a method for producing quasi-synthetic Language and arrangements for carrying out the method in a vocoder with several parallel channels of parameter signals, containing information about the envelope signals of the individual channels, the basic speech frequency and its amplitude, as well as via the voiced or unvoiced sounds, in the case of voiced sounds in a fundamental speech frequency generator generated impulses and, in the case of unvoiced sounds, those generated in a noise generator Impulse channel synthesis units are supplied in which these impulses with the The envelope signal of the respective channel is modulated and finally added, and in which either the modulated or the unmodulated pulses pass bandpasses. The electronic synthesis of human language wins for voice output from data processing systems and also for the purposes of transmission technology one always greater importance. A number of procedures and corresponding ones are already in place Arrangements for the generation of quasi-synthetic speech become known. Under the In this context, the generation of quasi-synthetic speech is a structure of the Understand speech from the basic speech frequency.