Controlled delay line signal processor for sound reproduction

'our tin

where c is said factor, f (i) is the time delay function, which represents the fluctuation of the controllable time delay.

21st The invention relates to a controlled signal processor for the processing of electrical signals according to the features of the preamble of claim 1.

Such a signal processor is from the US-PS

2 «16 71 151 known. This signal processor is used for signal compression, ie the signals are reproduced in a shorter time than they were recorded. For this purpose, the signals are supplied to an input of a delay device which consists of a circular signal path and a scanning head moved mechanically along this path. The propagation speed of the signals along the path is constant; as the scanning head moves relative to the propagating signals, a

jo bandwidth extinguisher, which serves to transmit the signal over a narrow band telephone line.

A similar signal processor is known from US-PS 34 80 737, where the signals to be transformed are sent via a delay line having along the signal propagation direction a plurality of taps, at which different delayed signals can be removed. For tapping a high-speed switch is provided, the controlled switches through each different taps.

A similar signal processor with switched abgeed delay device is also known from US-PS 25 45 371. Another mechanical signal processor for signal compression and expansion is described in US-PS 23 52 023, where a mechanically rotated head with a plurality of Einzeltonköpfen is used. This sound head is rotated so fast that the relative speed between the sound head and the tape during playback corresponds exactly to the relative speed between the tape and the unmoved head when recording. A similar signal processor is also described in the journal "Elektronik" no. 17 of April 27, 1957, page 139.

DE-AS 19 40 384 describes a vocoder synthesizer for speech synthesis. The stored words should be played faster or slower than they were recorded. In a vocoder synthesizer, however, the frequency does not change. Only the rates of change of the individual frequencies of a word are treated there. To limit the rate of change steepness) of the signal components, a frequency-dependent filter is proposed there, which controls the b5 synthesizer. A frequency transformation thus does not take place.

The US-PS 34 09 736 is concerned with the correction of phase and frequency errors, the

occur periodically. Ks is a voltage controlled, variable delay line is used, whose control voltage is related to the total phase error of a certain number of previous. Synchronization pulses.

A voltage-controlled delay device is also described in US-PS 32 02 769. This is used there to change the timing characteristics of time-varying, periodic signals and in particular of television synchronization signals. Specifically, the line changes of the horizontal sync pulses are measured with respect to an average line scan frequency. From this an error signal is derived representing the synchronization time error of the corresponding scan line. This error signal is corrected before the beginning of each image line by a time delay, so that the synchronization error is compensated. For this purpose, the corresponding line signal is sent via the controllable delay line, at whose control inputs said error signal is applied. Next this Durckschrift also deals with audio signals. Here, a sound track and a reference track are written on a magnetic tape. The reference track contains signals of constant frequency, e.g. B. a kHz. From this reference track a Fehlcrsignal is derived, which is a measure of wow and flutter of the tape drive. This error signal then controls the delay means through which the audio signal is passed. It then takes place a corresponding frequency transformation, which eliminates the frequency errors caused by the jitter. Corrections can be made there only to a limited extent, which is due to the "length" of the delay line. If, for example, the tape is running at too high a speed, more information is continuously input to the input of the delay line than is picked up in the output. It will therefore occur after a short time, an "overflow" of the delay line. Thus, there must be balanced in the mean value of the wow and flutter, ie fluctuate with positive and negative deviations to a constructively predetermined mean. A frequency correction in purely statistically occurring signals, which should take place substantially long-lasting in one direction, is thus not possible there.

The US-PS 30 93 796 deals with a signal compressor. Binary signals that are to be compressed in time are sent through an N-stage shift register. The clock inputs of the shift register are supplied with a signal which is a multiple of the frequency with which the signals to be transformed are sampled in a sampling circuit. The shift register thus serves as a delay device with a constant delay.

The invention is intended to coincidentally occurring electrical signals, for example, analog representations of audible sounds or sounds, such. Human speech or music, are, indeed, transformed to a natural frequency component distribution but at time intervals different from the original duration of the signals. In general, such signal processors are also referred to as signal compressors or expander. These⁽ Naturally, compressors and expanders work with delays of finite "length"; This means that not the entire signal sequence (eg the signals of a several-hour audio tape) can be "stored" in the delay device. Consequently, the delay devices operate periodically by processing (short) portions of the entire signal sequence to be processed in one cycle, then reverting to an output state and processing the next section accordingly. The "joining together" of the individual processed sections now causes problems. When ι a signal compression sections have to be "cut out" mentally, the then to be added to each other sections do not match exactly. In an expansion, on the other hand, there are "gaps" that have to be filled in some way. In both cases, these "connection points" are audible, which is reflected in a noise or interference signal.

The object of the invention is to provide a controlled signal processor referred to in the preamble of patent claim 1, in which using a signal-controlled delay noise and noise caused by the signal processing can be eliminated.

This object is achieved by the features specified in the characterizing part of claim 1. Advantageous embodiments and further developments of the invention can be found in the dependent claims.

Consider mar. Once again the portion of the signal sequence to be deleted, it can be stored and subsequently "destroyed" in the delay means or it can be divided so that it does not reach the input terminal of the delay means. In any case, this signal and the transitions generated by resetting the delay device must be cleared when the variable delay device is repeatedly controlled by the delay signal between the maximum and minimum delay values. Resetting of the delay and erase occur simultaneously, with two originally spaced portions of the burst being brought close together.

In one embodiment of the invention, the input and output bandwidths are controlled for those voice frequencies that must be reproduced for ease of understanding in relation to the compression ratio, thereby excluding those frequencies. which would cause distortions and intermodulation due to incomplete sampling sequences or excessive phase shifts per high frequency delay stage and unfavorable extinction lengths to obtain frequency conversion for the lowest frequencies.

Furthermore, maximum deletion intervals are provided, which are determined by the maximum "length" of the delay means, which deletes the original message with respect to the Pressverhältnisverhältnisgrenzen to a value at which the loss of noticeable transmission commands or transitions is minimized. The loss of information is thus kept minimal.

Further, measures are taken to process the signal at a junction of two reproduced signal portions to suppress deflection noise components and also to avoid the introduction of false transmission commands that transferred them to the subsequent signal portion

To modify information. Finally, if necessary, to suppress such noises, the transition between successively reproduced signal portions may be modified by transfer function selection or control, or the transition may be made simpler by introducing artificial or signal portions derived from the original signals, to approach a smooth transition within a time interval.

In the following the invention will be explained in more detail with reference to embodiments in conjunction with the drawings. Show it

Figs. 1 (a) to 1 (h) are diagrams showing the reproduction of a message recorded on a magnetic tape and also the operation of the system according to the present invention at different press ratios;

F i g. Figures 2 (a) and 2 (b) are plots of compression and expansion, respectively, showing input-output signal time relationships;

Fig. 3 (a) shows a number of curves representing the relationship of various parameters occurring in the processing of speech at different press ratios greater than one;

Figure 3 (b) shows similar relationships for ratios less than one (ie, for expansion or expansion);

F i g. 4 (a) to 4 (f) are waveforms valuable for the description of processing forms of a transition between adjacent reproduced speech samples;

Figures 5 (a) to 5 (d) show a number of curves representing the active processing of the transition between adjacent samples;

F i g. 6 (a) to 6 (e) are waveforms advantageous in describing the application of the two delay lines to effect a transition between adjacent speech samples;

Fig. 7 is a block diagram of a voice press expansion system according to the present invention;

F i g. 8 (a) to 8 (d) are waveforms used to describe the operation of the system of FIG. 7 are advantageous;

F i g. 9 is a block diagram of a dual delay line system according to the present invention;

10 (a) to 10 (d) are waveforms used to describe the operation of the system of FIG. 9 are advantageous according to a pressure;

11 (a) to 11 (d) are waveforms used to describe the operation of the system of FIG. 9 are advantageous according to expansion or expansion;

Fig. 12 is a partial block diagram of a modified embodiment;

Figs. 13 (a) to 13 (c) are waveforms used to describe the operation of the circuit of Figs. 12 are advantageous according to a pressure;

Figs. 14 (a) to 14 (c) are waveforms favorable to the operation of the modified embodiment of Fig. 12 in accordance with expansion or expansion;

Fig. 15 is a partial block diagram of a binaural dual delay line system;

Figs. 16 (a) to 16 (f) are waveforms used to describe the operation of the system of Figs. 15 are advantageous;

17 is a partial block diagram of a speech processor according to the present invention using an analog shift register as the variable delay element;

FIG. 18 is a block diagram showing gap fills in signal continuity in a system similar to that shown in FIG. 17 illustrates;

19 is a partial view of a modified embodiment of the invention, wherein the variable delay by an r-bit digital

Parallelο parallel shift register is provided;

Fig. 20 shows another embodiment of the invention wherein the variable delay is provided by a digital serial shift register; F i g. Fig. 21 shows another embodiment using an analog memory array to provide a variable delay;

F i g. Fig. 22 shows another embodiment using a direct access digital r-bit memory;

Fig. 23 is a logic circuit of a direction null signal value control unit;

F i g. 24 waveforms used to describe the operation of the circuit according to F i g. 23 are advantageous;

25 shows a graphical representation of the clock frequency and the maximum signal frequency for the system according to FIG. 17;

F i g. Figure 26 is a block diagram of a dual delay line memory system using analog shift registers using separate read and write strobe signals; and

F i g. 27 (a) and (27 (b) are waveforms of the touch control signals for the system of Fig. 3. The description of the preferred embodiments is preceded by an explanation of the parameters of the speech signal, particularly with respect to the parameters that the speech pressure for reproduction relate to a given linguistic message in a shorter period of time. Due to the fundamental and unavoidable limitations encountered in voice compression, the following discussion refers particularly to the method and apparatus used in the pressing operation or the pressing operation phase. The temporal biasing mode results in cancellation of a portion of the original information which is directly proportional to the press factor, which is also the factor by which the time is reduced to represent a given speech sequence. The

However, the method and the apparatus can also be used for expansion, and the aspects occurring in this case for the reproduction of the signals which occupy or take up a greater length of time than the original linguistic utterance, will be explained later. The system also has the capability of frequency transformation, without a corresponding time change, to achieve a desired frequency signal, as is the case when generating a speech in a medium having a propagation velocity other than air

F i g. Figure 1 shows a particular system using a delay line providing a maximum time delay of 6 milliseconds for the final portion of the sample. Assuming that the speech signal to be processed is limited to frequency components between 333 and 5000 Hertz certain parameters of the

Define playback playback system. On a magnetic tape 21, the speech signal is recorded, with the lowest frequency component at 333 Hertz is picked out by the sine wave 22, wherein the tape is guided past a scan converter 23 and is wound onto a take-up reel 24 at the speed S. The electrical signal generated by the transducer 23 passes through a pressing processor 25 and is reproduced as an audible signal to the speaker 26.

The corresponding to the line (a) of F i g. 1, system 23-26 generates a recording signal on the tape 21 without frequency or time change as the speed of the take-up reel 24 passes the tape at the transducer 23 with the recording speed S, for which condition the processor 25 performs a fixed constant time delay of any value. Line (b) of F i g. 1, where c = 1, reproduces the 333 Hz sinusoidal signal with no change except for the fixed phase delay (which was neglected).

For voice compression, the tape speed is increased by a factor of c, and the processor 25 linearly alters the delay by a minimum value to a maximum value. Like curves (c), (d) and (e) of FIG. 1 For example, at a compression ratio of c = 2 for a 6 millisecond final signal delay requiring an 8 millisecond delay line, a 12 millisecond portion of the originally recorded wave 22 is stored, now half and leaving half of the magnitude of the signal unblocked. which originally occupied 24 milliseconds of recording time. This retained portion is called "chunk" and is in line (c) of FIG. 1 illustrates in schematic form prior to the processing process and has cycles numbered 1, 2, 3 and 4. With respect to the pressing operation, and since the maximum final signal delay is kept at 6 milliseconds corresponding to the 8 milliseconds in the delay line at the end of the sample, a 6 millisecond portion of the original information representing 12 milliseconds at the recording speed is erased Section is labeled "discard" in line (c). This erase section contains cycles 5, 6, 7 and 8 of the original wave 22 and represents the gap in the information content between successive so-called "packs" that are reproduced as audible signals. This audible output is in line (d)

Table I

Typical parameters for voice presses

with the squeeze taken out as a piece of tape 31 which has been played at a speed 25, and which squeeze contains cycles 1-4, which after processing are effectively stretched into a piece of tape 32 occupying the original 12 milliseconds of recording time and contains cycles 1-4 with their original recording frequency. At line (d), it should be noted that the next reproduced .V-jus is the cycle number 9 of the original waveform. after cycles 5 to 8 have been deleted. The representation in line (d) of a smooth transition between the end of cycle 4 and the start of cycle 9 should not be regarded as a characteristic for real signal conditions, which results from a consideration of an actual signal in opposition to the one shown in FIG. 1 idealized signals would emerge.

The lines (f). (g) and (h) of F i g. Figure 1 illustrates the situation that is prevalent. when the press ratio is 5, in which case the tape speed past the transducer 23 is 5 times the recording speed 5. With a final signal delay of no more than 6 milliseconds, this press ratio will result in a 1.5 millisecond bond length, with 2 '/: cycles of the 333 Hz wave 25 of line (a) included, and again a 6 millisecond erase interval equal to the final signal delay, corresponds to a delay line length of 10 milliseconds at the end of the sample. However, the information gap has grown to the point where the last half of the cycle number 3 and the first half of the cycle number 13 and all intermediate information in the originally recorded time have been lost in the erasing section, and this gap represents 30 milliseconds of the originally recorded one linguistic statement.

The relationships among the parameters of a speech-compression system and those responsible for the information content of the coded speech are interrelated so that optimal conditions can be specified, and they set limits to the operation of the systems of the invention for a given intelligibility factor. These parameters can be checked for a given system for different press ratios, and for this purpose the system parameters for a system should be given in the following table, with a delay line having a maximum final signal delay ΔT_ma»Will be omitted from ö milliseconds.

Comp. relationship

Lead crowding / quenching ratio sample

length (playback)

»C« time)

dT_om DID_Max T_{at the}/ CAT_Max T

(ms) (ms) (ms)

<p><tgroup cols="2"><tbody><row><entry>Repeat</entry><entry>Cycles of sample</entry></row><row><entry> l / T MP </entry><entry>v / m " ⁼</entry></row><row><entry>33,3</entry><entry>10</entry></row><row><entry>55,6</entry><entry>6</entry></row><row><entry>83,3</entry><entry>4</entry></row><row><entry> 111 </entry><entry>3</entry></row><row><entry> 125 </entry><entry>2 2/</entry></row><row><entry>133</entry><entry>2 1/</entry></row></tbody></tgroup></p>

Hz)

2/9

2/5

2/3

6/5

4/3

6 2/3

7 1/5 8

9 3/5 10

<p><tgroup cols="3"><tbody><row><entry> 24/6 </entry><entry>30/7,5</entry><entry>30</entry></row><row><entry> 12/6 </entry><entry>18/9</entry><entry> 18 </entry></row><row><entry> 6/6 </entry><entry>12/12</entry><entry> 12 </entry></row><row><entry>3/6</entry><entry>9/18</entry><entry>9</entry></row><row><entry> 2/6 </entry><entry>8/24</entry><entry>8</entry></row><row><entry>1,5/6</entry><entry>7,5/30</entry><entry>7</entry></row></tbody></tgroup></p>

1/2

The basis for the frequency-time transformation used in the present invention may be derived as follows. Consider a sine wave V = ε sin a&gt; t recorded with a tape recorder. When the tape is replayed, c times the original recording speed, you get

V = E sin cwf. (1)

where c is referred to as pressing ratio. If c&gt; 1, then the time is pressed for any given speech passage, and if c &lt;1. The time is stretched by the factor e, where e = Mc.

Then, when the signal is applied to a delay line, the delay increases linearly with time at a rate d, so that the average delay of the signal c ', which represents the delay experienced by each point on the waveform, if any passes through the line, then the signal from (1) to

Line IV of c'T_m, the required amount. This will be

for the re-storage

dT "

+ dT_ou!

-DT_1n = ^-

10

15

d = c-1 c + 1

Using more general expressions, re-storage can be achieved by cumulatively delaying an input signal t "by an amount

20

/(O

'nut' ι

dt = C¹I_1n = (cl) r, "

(5)

V = Es \ n (c - c ') u&gt; t.

(2)

The original signal becomes decremented when the delay is equal

25 can be achieved.

For a linearly variable delay line, with a delay rate d, f (t) = d t, the following results:

c't =

so that

(3)

(4) c'7;_n = ic-1) 7 ;. =

T -Τ

'out' ir,

■ tdt

(6)

30

d =

Thus applies to

ά. H. the mean delay line speed, which is one-half the sum of the final and initial delay values

the delay line can be obtained, the _

then, when multiplied by the time t, the ^c ''" ~ ~

total delay c't yields.

F i g. Fig. 2 (a) shows a diagram for a given 40. Thus, signal sample of the signal output time t_:,_us relative to the corresponding input time r_e". The line I with a slope of 4 represents a signal at four times the original frequency or display speed and with¹A is the periodicity, while a line II with a slope of one represents the resulting stored or unmodified signal To such a signal (represented by the line I with the slope c = 4) in a signal corresponding to the line II with a slope of one to convert, with a corresponding decrease in frequency, it is necessary to increasingly delay the input signal ci "" by an amount c't (or (c-) t), as shown by line III, a signal elongation section Ti_n has an ordinate which intersects the line III at the ordinate value c% ", and when this value is at the time abscissa T", at the point cT_m is added on line I, the signal is set to T_OUt delayed on the line II The delay dt introduced by the delay line is represented by the line IV. Such a delay line has the effect of increasing the instantaneous signal V by a linearly increasing amount dt for the interval of t, Thus, considering the case at the signal end at time t = T, thus half of the sum of the initial delay dT, "and the final delay dTom results in a mean delay value on the 2T_oa-T,_{n Ta} 2 (C-1) 7 ;. from which one receives (4) d

(C + 1) 7; ..

In F i g. 2 (b), the corresponding relationships for expansion are shown. Line I_c with a slope of ¹A represents a signal from ¹A of the original frequency or display speed. To convert such a signal into a signal corresponding to the line II with a slope of one, with a corresponding increase in frequency, it is necessary, the input signal Ci_1n decelerating decreasing

and by an amount

from an initial delay of C¹T₀₁₁, out. This delay amount c't 'shifts the signal to df at every point! corresponding original value on the line II. The delay introduced by the delay line dt'xsl through the line IV_C shown. Such a delay line has the effect of delaying the instantaneous signal by a linearly decreasing amount dt 'for the interval from f1 to tau, as indicated by line IV_e is shown. Thus, considering the case of the initial signal at time r = 0, half of the sum results in the initial delay -d-T; "and the final delay - dT_uu, a mean deceleration value on the line IV_f from -c'Tin. This will be

c'T, s =

\ -e

rf = 2

\ + E

The process of linearly increasing the time delay can not be continued indefinitely, and therefore, from time to time, the delay line must be returned to its original "length." If this process is repeated at periodic intervals, provided that the interval is longer than the period of the lowest frequency component of the signal, reboots of the original signal are replayed with the angular frequency (c-c ') o) and the remainder is deleted. If (3) is satisfied, the system operates as if sections had been cut out of the original tape and then reassembled and then played back at normal speed. The sections of the signal are heard at the correct frequency, but the information is transmitted in a shorter time (if oil). The language was pressed to 1 / c of its original length.

The values ​​of Table I are in FIG. 3 (a) applied. For any given pressing ratio, the sampling time is given by the curve T, "", and the crowding length is represented by the curve T, ". The difference between these two curves represents the extinction or extinction section, which is equal to the final delay of the signal at the end of the sampling period (6 milliseconds in the example of Fig. 3 (a)). If one chooses a value on the curve at any pressing ratio, such. At c = 5 in F i g. 3 (a), we obtain the tape erase and erase times running at the recording speed, and orienting these values ​​to the time axis gives the actual original recording time for the respective erase and erase sections. For the case c = 5, the crowding section is 1.5 milliseconds long, and the erasing section is b milliseconds, making 7.5 milliseconds each of the recorded and reproduced information and 30 milliseconds of the erased information. This latter value is determined by the size cT_{mj &lt;} shown in FIG. 3 (a) is applied.

For a speech signal in which the lowest frequency of 333 Hz has a period of 3 milliseconds, a burst length of 1.5 milliseconds at C = 5, which corresponds to 7.5 milliseconds of recording time, contains 2.5 cycles of 333 -Hz signal. For any higher frequency components in the speech signal, then more cycles are included in the 1.5 millisecond crowding section. The length of the crowding section should exceed the period of the lowest frequency (ie should

to contain at least one full cycle), otherwise no satisfactory pressure can be achieved. As shown in FIG. 3 (a) is displayed at the bottom of the time axis at 3 milliseconds, the 333-Hz signal, when pressed in accordance with sampling periods approaching 3 milliseconds, with the samples appropriately rearranged, results in a compressed, low quality output, because the sampling causes an interrupting irregularity for almost every cycle of the 333-Hz processing signal. Sampling periods less than 3 milliseconds do not allow the completion of any cycle so that the resulting recreated output would not only contain the mentioned interrupts, but also begin to have a fundamental change in its frequency characteristic in the form of clipping by clipping and producing spurious frequencies werder. However, since this condition represents a real condition for a speech wave due to the complexity of the waveforms, it does exist and sampling periods less than the period of the lowest frequency wave in the speech signal do not provide the proper pressure.

Sampling periods greater than the period of the lowest frequency wave result in a squeeze, and in the zone in which the sampling period is only slightly larger than the period of the lowest frequency wave, there is an interruption interval, as on the line axis between 3 milliseconds and 6 milliseconds in Fig. 3 (a) is displayed. The result obtained during this interruption period consists of a distorted extended wave, where the separation effect between the samples becomes very severe when the single cycle point is reached, and this effect diminishes when the count increases. of cycles in the sample increases. In practice, two and a half cycles per sample are the desirable limit in FIG. 3 (a), however, it is generally believed that the greater the number of cycles in a sample, the lower the disturbance factor.

In order to avoid the extreme distortions created by wavelengths having a longer wavelength than the sampling period, these low frequencies should be filtered out before the speech signal enters the delay line, otherwise these separated and heavily distorted Weller will travel through the line and into the line Intermodulator with the wanted signal occurs and the quality of the system; can greatly reduce.

For pressing ratios lower than ci and at AT, "^ = b milliseconds, the length of the staple increases, with the result that the actual sample distance exceeds 7.5 milliseconds

t&gt; 5 grows and therefore more than the minimum number of cyclers for the lowest frequency component is present in the crowding section. If desired, the user can control the line for less than f

Milliseconds of displayed delay for AT_Max operated to reduce the size of the erase section.

If the "erase portion of the sample is considered to be constant for 6 milliseconds in the compression ratio in playback, the actual information loss is equal to the press ratio times 6 milliseconds, so that at C = 5 the actual erased information for each Sample equal to 30 milliseconds of recording time is as from the time axis of the F i g. 2 This is the interval from 7.5 milliseconds to 37.5 milliseconds, and the relationship of this loss of information with the intelligibility of the reproduced speech signal must be checked.

In general, human speech provides a very complex code from the relatively limited number of sounds called phonemes associated with various attributes of the linguistic code, such as the tuned-uncoordinated components, pitch, formant frequencies, and the continuum of the sound image represented by the sound energy (and the lack thereof) and constituted by all the essential transitions between the intermittently occurring components thereof, constitute an acoustic current of infinite variety and versatility. The ability of the human ear to receive this acoustic message and the ear-brain system to decode that message has not been fully explored as it appears that the comprehensive information sequence far exceeds the ear's pure acoustic response characteristic Receiver exceeds.

Fortunately, the ability of the ear-brain system to summarize the message transmitted by human speech signals is sufficiently good to allow large portions of the transient acoustic current to be lost or erased, without appreciable loss of perception or recognition and the summary of the message information content of the acoustic signal. Since the understanding of the message content decreases much faster than recognizing or capturing individual words when the message is offered to the listener at an increased speed, the problem associated with canceling a portion of the signal stream may be in favor of the viewpoint or message Understanding and can go to the point where the intelligibility of individual words ceases. This latter point is reached when the loss or change of transitions or other transmission commands (cues) that effectively results in the connection between a consonant sound and a self-sounding of the extinction of a large part or of the entire transmission command (s) that is, the apparent information content of adjacent and chained crowding sections is changed. Even before the point of absolute intelligibility loss is reached, the tolerance limit corresponding to the inconvenience in continuous listening already occurs due to the unnatural sounds and the fatigue that occurs in the intense concentration, the information in the presence of excessive time gaps or to detect clippings.

For speech compression purposes, the loss of intelligibility may be attributed to the erasure portions of the message which contains significant transmission voices or phonemes, which components change in length, the shortest being from about 10 milliseconds to 20 milliseconds in length. These short transient commands do not dominate the language, but they occur - with sufficient regularity, so that their systematic

Verlustο loss becomes undesirable and thus a desirable upper limit for the erase period should be 30 milliseconds and preferably closer to 15 milliseconds should bitrate. For this fixed limit to the intelligibility of the reproduced syllables and words, the speed of presenting a given message for any given listener can be increased to the term or detection limit and increased to the level of difficulty of the subject, with little attention to limitation or limitation which is due to loss or distortion of the word content or generating

false transmission instructions of the concatenated message suppression sections is set.

Fig. 3 (a) shows the recording time extinction ratio to the pressing ratio as a linear function cAT_mil&gt;, where the range of 18 milliseconds to 30 milliseconds indicates the uncertainty erasure range. Thus, at C = 5, the 6 millisecond erase section contains the actual recording time interval from time r = 7.5 to t = 37.5, which approaches the upper limit allowed for erasure without an excessively large loss of intelligibility, as required to not significantly contribute to the loss of intelligibility of the received message. Smaller values ​​of c provide an actual shorter erase time, and thus improves intelligibility, especially for those transmit commands that are at the lower end of the timescale, ie, in the neighborhood of 10 milliseconds.

While Table I and Fig. 3 (a) illustrate parameters for a typical speechpress system, with a final signal delay of 6 milliseconds, and define the operating limits within narrow limits, it should be noted that the

5 can also be adapted for use over a wide operating range. Thus, the change of the actual frequency band of the speech signal and the maximum length of the delay line are important design factors which influence the selection of the annulus to erase ratio and the sampling period for a given range of the compression ratio c. On the other hand, the actual frequency range of the signal has a significant influence on the design of the delay line, which must be adapted to the frequency spectrum present in the signal, in terms of quantitative and qualitative factors such as voice pitch, the presence of all or only some of the formant frequencies for an individual speech and the width of the signal spectrum, over which linear phase frequency characteristics must be obtained. The final system used, however, is the designer's wishes

*&gt; 5 of the mentioned factors within the wide limits defined here.

F i g. FIG. 3 (b) is a plot of corresponding signal strain relationships, with the initial

Spal, the output displacement section and the maximum fluctuation or change in the delay line length at an expansion ratio e for a given input sampling interval T_m The output gap occurs at the start of each sampling period, and thereafter, when the sampling period is adjusted, the frequency-wise, time-extended output crowding portion occurs. The maximum required delay dT_in is also shown as a function of the expansion ratio e.

A feature of the voice-press system, which in conjunction with F i g. 1, the audible output of the converter 26 has not been dealt with when the variable delay processing unit 25 is switched from the maximum to the minimum delay at the end of the sampling period. Shortly before the switching operation, the delay line is loaded with the speech signal which is to be canceled, and if the line is momentarily switched to a delay of zero, all information or information, if not eliminated or previously extinguished, will be in high-condensed or condensed form be contained in the output signal. In practice, using a conventional delay line using R and L or C components, a time interval is required to switch the line from the maximum to the minimum delay, and it has been found that even if the line contains no signal information, this switching operation of a line has a noticeable minimum time constant associated therewith, thereby producing an audible disturbance transition in the output signal, wherein the repetition frequency or repetition frequency of this transition is the reciprocal of the sampling period. Due to the limitations imposed by the parameters of the system, as described earlier, the switching frequency and the spectral components of the transition itself are always within the listening range, and therefore, it appears as a highly undesirable in-modulating component in the audio output size of the device , In the following, a number of transition suppression devices are proposed, as well as message gap bridging arrangements to minimize unwanted noise effects. In more detailed systems, substitution of the pseudo or real message components further improves the transition from one sample to the next, and can suitably fill a portion which is erased during the pressing process.

Referring to Fig. 4, a portion of the 333 Hz wave at the transition point shown in Fig. 1 (d) is reproduced showing cycle 4 and cycle 9 of the originally recorded 333 Hz wave as a smooth, unbroken sine wave are. The connection between the end of cycle 4 and the beginning of cycle 9 at point 41, although shown as a continuous or uniform section of the sine wave, is in reality, as already mentioned, in most cases never so related to non-selective periodic sampling independent complex waveforms, so that, instead of a soft transition point 41, a separation must be expected between the end of the one bundling section and the beginning of the next bundling section in successive samples. This disconnection can undoubtedly be adapted without loss of intelligibility if the transition through the switching of the line (either loaded or unloaded) need not be dealt with precisely at this time. Since this transition is responsible for a very annoying audible output of the system, it must be eliminated, and for this purpose a key signal as in FIG. 4 (b) is indicated symmetrically with respect to the transition point 41

angelegto created to produce the output signal generated in Fig. 4 (c). By making the keying sufficiently long to encompass the transition resulting from the switching of the line, the audible noise thus generated is eliminated. The improvement achieved by this aid, while significant, is not ideal since the insertion of the key signal into the audible range itself is audible as a repetitive separation gap resulting in intermodulation with the audio signal.

This effect can be reduced by using an output filter designed for the particular repetition rate and sweep width to smooth the sudden transition shown in Figure 4 (c), where the response of that output to F i g , 4 (d) is displayed.

Further improvement is possible by using a key signal in the form of a gain control signal, and also by tapering the "off" and perhaps the "on" transitions of the gate, so that a gradual transition of the audio output from "off" to "on" is reached and results in a relatively smooth or smooth transition, as indicated in Fig. 4 (f). The goal is to minimize the column effect, which in itself has an audio characteristic and how a transmission command (cue) can act. This is greatly assisted by tapering or sharpening the trailing edge of the keying, but preference is given to anticipated start (or relative delay of the speech signal) for gradual onsit for the leading edges. With these relatively simple measures, it is possible to achieve a smooth transition between adjacent urging sections, which are lined up one after another by the operation according to the press-erasing process, to the extent that the device is satisfactory for many applications.

With reference to F i g. 5 are the more outdated

so arrangements for bridging the gap between the samples will be described. Fig. 5 (a) shows a separation transition corresponding to the expected standard, and this represents a sharp irregularity in the message signal, and this transition is superimposed on the noise transition resulting from the switching of the line, as described earlier. By introducing a key signal according to FIG. 5 (b) with sufficient width, so as to cover the line switching path.

Knowing or embracing, and by arranging the key signal to coincide with a zero value and in the same direction of change for the adjacent signals being processed, a zero-valued tactile transition can be achieved.

n ") as shown in Fig. 5 (c). This transition, which is free of line switching noise and substantially continues an existing zero amplitude signal value during the interval of the sample signal, results in a

little or no interference for the average listener, as stated.

Due to the nature of the human hearing, especially because of the ear's ability to assemble the message it focuses on, even in the presence of noise, in some cases or circumstances, it may be advantageous to have a pseudo or real message component in the zero value interval! as shown in Fig. 5 (c). For this purpose, a suitably selected noise or signal components of nearly the same amplitude and frequency may be introduced or inserted where there is otherwise a gap interval in the message stream, and this arrangement according to the invention is in FIG. Figure 5 (d) illustrates Where the gap is to be filled with noise components, a suitable source and balanced switching may be provided to introduce the noise from the source into the signal channel immediately during the sample interval.

F i g. FIG. 6 illustrates a preferred form of gap fill using two signal-controlled delay lines. The speech signal is applied to both delay lines corresponding to the channel labeled A and the channel labeled B to the F i g, respectively. 6 (a) and 6 (b), and these two lines are signal controlled to achieve symmetric complementary gain characteristics and overlapping variable delay characteristics, as shown in Figs. 6 (c) and 6 (d). Here, as shown in Fig. 6 (d), the delay-controlling signals are arranged in phase so as to overlap at least an amount corresponding to the transition portion of the gain control characteristics shown in Fig. 6 (c). The outputs, both delay channels A and B, are connected to produce the combined output shown in Fig. 6 (e).

Generally, the length of the delay lines used for the channels A and β in Fig. 6 is one full delay line length and one relatively shorter delay line length to store the signal used for gap filling purposes. By this arrangement, the cost of the device is reduced, which is due to the many delay line sections required to obtain the required maximum delay length for the requirements of the system. On the other hand, for systems where the cost issue is not primary, two equal variable full length delay lines may be employed and their control signals alternately fed so that one signal channel is used first and then the other delay line, providing a full signal period is to switch the inactive delay line back to minimum delay, before using it for a new signal transmission. For such symmetrical delay lines, it may still be advantageous to provide an overlap during the transition, as shown in FIG. 6 (d) is indicated, with appropriate gain control signals applied, as in FIG. 6 (c).

It should now be on F i g. 7 and a basic voice press expansion system will be described with the controlled signal processor in question. This system includes a: variable speed playback device 51 consisting of a tape transport having a manually selectable speed control input 52. The signal derived by transporting the tape past a magnetic transducer is applied to an AVR amplifier 53 which passes the signal through a bandpass filter passes, which has an adjustable low and high cutoff frequency. The choice of cutoff frequencies for the filter may be made by the hand controller 52 in conjunction with the choice of playback speed for the playback device 51. The hand controller 52 also sends an amplitude control signal to a speech pitch fine tuning control 54, which then sends a signal to control the end amplitude of the linearly increasing waveform on the line 55 which controls the variable delay line, as will become apparent.

After passing the amplifier and filter 53, the signal enters a variable delay line 56 which can be signal controlled between the minimum and maximum delay limits. This control signal, which comes on line 57, is derived from a sawtooth value amplitude shifter 58 which receives as input either a triangular pressing waveform on line 59 or the inverse amount of expansion of waveform 59 appearing on line 61, after passing through an inverter 62 has arrived. One or the other of the lines 59 and 61 is energized with a sawtooth waveform, which depends on the setting of a switch 63 which supplies the basic sawtooth waveform from a sawtooth pulse generator 64. The repetition period of the sawtooth waveform can be selected by means of a manual control 65. A pulse coinciding with the reset of the linear portion of the sawtooth waveform appears on line 66 and is applied to a blanking pulse generator 67 to produce a blanking pulse output whose width can be controlled by handset 68 and which synchronizes with the input pulse on line 66 is.

The output of the variable delay line 56 is supplied to a blanking circuit and an amplifier 71 which transmits or blocks the signal, depending on the blanking pulse (B) supplied on line 72 from the generator 67 and if the blanking pulse is absent (B), the delay signal is supplied to a voice bandpass filter 73 whose output is supplied to an audio transmitter 74.

In addition to the ampule deflection established by the linear sawtooth voltage signal from the generator 64 controlled by the hand controller 52, the absolute value of the applied voltage can be controlled by the value adjuster 60. The variable delay line 56 may be of any known type and, in particular, may consist of 360 RC filter stages, wherein the shunt resistor may be provided by a FET (Field Effect Transistor) or other semiconductor device which varies the resistance in response to a control voltage or current , Such delay lines generally give best results in terms of distortion of the signal passing therethrough when

delay per stage is kept well below the maximum possible value of 90 °. As a result, the line may be designed to operate at 45 ° to 60 ° maximum phase delay per stage, and the number of stages will then be greater than the size: W&gt; (6 or 8) c (f_{m n}) ΔΤ ,,,_{; η}, In the above inequality, numerals 6 and 8 represent the numbers of steps per electrical cycle of the highest frequency to be transmitted, corresponding to a phase delay of 60 ° or 45 ° as the maximum of the phase shift per step to be used ; the size c is the pressing ratio; the size f,_mx is the highest frequency that is to pass through the line; and AT,_m% is the desired maximum signal delay, which is specified by the maximum possible deletion interval, as explained earlier. There are many other embodiments of delay lines that can be controlled by a signal, and the present invention is not limited to any particular form of delay line.

With reference to the figure. 8 (a) and 8 (b), the operation of the system of FIG. 7 will now be described. Sampling period waveform 81 has an adjustable set of periods which may be achieved by controller 65 to produce an asymmetric sawtooth waveform 82 which results in a relatively long negative going linear voltage to which a shorter positive going linear voltage follows. This waveform on the line 59 is used directly for the voice pressure, and the inverted size thereof, after an inversion in the inverter 62 appearing on the line 61, is used for stretching. The strain waveform is indicated by dashed lines at 83 in Fig. 8 (a). For a variable delay line 56 where the delay increases as the control voltage becomes increasingly negative, the waveforms 82 and 83 are in the right direction to control the delay intervals, and the magnitude of the delay is adjusted by an amplitude controller 52 relative to the voltage value the value adjuster 60, determined. In this way, the operating point in the deflection of the waveform 82 is selected for a given compression ratio in conjunction with the sampling period, which represents a particular combination for a given compression ratio, assuming that the maximum delay T_mnx in the line 56 has a fixed value, which by selecting the line length corresponding to the value d ■ T_ouis given and given in Fig. 3 (a) and Table I for the desired pressing ratio. If the maximum delay experienced by the signal is not kept constant, the erasing period changes accordingly, as can be seen from the description of Fig , 1 and corresponding adjustments in the amplitude of the wave are required to achieve the slope d required for a compression ratio c. Similar considerations apply to the slope of the curve 83 which must be set to its corresponding value d for an expansion ratio e.

The operation of the blanking pulse generator 67 is such that a pulse 84 corresponding to F i g. 8 (b) is generated with a certain width depending on the start pulse of the sampling period signal 81.

which is received by the line 66. This pulse may be applied to the circuit 71 for gain control, with a modified trailing edge as previously described to reduce the transient signal and to provide a gradual onset of the speech sound signals to the transducer 74 be directed. The width B of the blanking pulse is selected by means of the controller 68 and is normally made sufficiently wide or of sufficient duration to allow the short linear portion of the sawtooth waveform to have the possibility of Return the delay line 56 to its zero state or minimum delay state and to destroy the signal energy contained therein (or caused by the switching of the line itself transition) before the commissioning of the signal channel, which channel energizes the converter 74 with the subsequent speech signal segments.

The blanking period B and the standby period B for the expansion operation are shown in Fig. 8 (c). The extended crowding sections with an initial exit nip are in F i g. 8 (d).

The system according to FIG. 7 can also be used to substitute noise or pseudo signal gap filling signals, according to the system of FIG. 5. For this purpose, a source 75 is provided for such signals and feeds the input signal to a filter 73 during the blanking interval. With the help of a switch 76, this gap filling process can be optimally designed during the blanking interval. The gap fill signal 75 may also be derived from the message signal output of the amplifier 53.

F i g. Fig. 9 now shows a modified embodiment of the controlled signal processor in question which is particularly suitable for implementing the various gap filling methods for speech compression systems described earlier. Sections of F i g. 9, which are substantially the same as those of F i g. 7, are provided with corresponding reference numerals, and therefore only the additions and additional changes will be described in more detail. In addition to the variable delay line 56, a second variable delay line 91 receives a signal wave from the amplifier 53. The outputs from the delay lines 56 and 91 are supplied to complementary blanking circuits 92 and 93, respectively. The signals passed through these blanking circuits 92 and 93 are amplified and filtered in the element 73 and are directed to the acoustic encoder 74 as previously described.

A pulse generator 94 generates a pulse train as shown in FIG. 10 (a) at a selectable pulse repetition frequency that can be determined by adjustment of the hand controller 65, thereby specifying the basic sampling period. The output pulse from the generator 94 is delayed in the delay unit 95 and applied to a first sawtooth generator 96 and fed in undelayed form to a second sawtooth generator 97. The sawtooth generators 96 and 97 are subjected to waveform value control by the manual adjustment element 60 and linear sawtooth amplitude control by the manual adjustment element 52 subjected. As stated earlier.

For example, the fine pitch adjustment 54 may be provided to easily change the sawtooth pitch as a voice pitch adjustment by effectively changing the frequency conversion over a small range. Additionally, the width of the blanking interval of each generator is adjustable by means of the controllers 68 and 70, respectively. The output of the ramp generators% and 97 are applied to the delay lines 56 and 91, respectively, to control the time delay of the signals passing through the respective lines in accordance with the supplied control signals. With the aid of c or e selection control devices, the direction of the slope of the sawtooth waveforms can be selected for a squeeze or an expansion.

The value and amplitude controllers for adjusting the respective sawtooth generators% and 97 are preferably relatively adjustable to enable selection of the relationship between two sawtooth waveforms. By making the delay and phase adjustment of unit 95 adjustable, any desired delay line overlap can be achieved. It is also possible to rearrange the components to provide complementary gating at the inputs of the two delay lines 56 and 91, the outputs being switched so that they are connected or combined in a common channel to the amplifier 73 , This alternately clears the portion of the speech signal that is not used by each line before it enters the line, eliminating the need to cancel those portions when switching the lines between active periods.

With reference to F i g. 10, the operation of the voice-press system of FIG. 9 will now be described. The pulse generator 94 generates the timing waveform of Fig. 10 (a). This pulse triggers the transition of the waveform Cl in the sawtooth pulse generator 97, which generates the blanking pulse, which in F i g. 1_0 (c), the determined width of B and B being determined by the blanking pulse width control 68. After the in Fig. 10 (b), the pulse from the generator 94 triggers the sawtooth generator 96 to produce the waveform Ci shown in FIG. 10 (b). In this arrangement, the control line Cl for the delay line 56 is timely overlapped by the waveform C2, which has the same direction slope and bridges the steep return branch of the sawtooth wave Cl. At the asymmetric time intervals shown in FIG. 10 are shown, the arrangements for a gap filling according to F i g. Perform 5 and 6. By designing the waveforms Cl and C2 to have symmetrical rising and falling portions, the arrangement for alternately switching the leads 56 and 91 becomes suitable for providing alternately pressed (or extended) urging portions of the speech sample. The choice of the relative lengths of the sample through lines 56 and 91 is generally dictated by manufacturing costs for the delay line. For a main delay line 56 of sufficient length for the desired pressure ratio, a relatively shorter line 91 can only be used for gap filling purposes, which is generally more economical On the other hand, two full-length lines, which become active alternately, to pass the voice sample crowding sections while also providing adequate time for the idle line to return to its minimum delay state, to soft or smooth transitions, to any desired overlap and to a maximum time interval for the discharge of the line the minimum delay state, before processing the

ίο next voice sample. The operation of the system of FIG. 9 in the gap-filling phase of operation is indicated in FIG. 10 (d) and corresponds generally to that described with reference to FIG. 5 has been described. The operation of the system according to FIG. 9 for a speech expansion, ie Increasing the time duration for a given speech utterance and increasing the frequency components from a donor or playback device running at a slower speed than the recording speed is shown in FIG. Here, the sawtooth generators 96 and 97 have inverted outputs to produce the extended waveforms E 1 and E 2 shown in FIGS. 1 1 (a) and 11 (c), the blanking waveform has been made symmetrical such that the delay lines 56 and 91 are alternately used for nearly equal periods. Due to the nature of the speech spread, there is always a gap in the output signal because the lines are controlled so that when a sample is started, the delay is brought from a maximum to a minimum or to a zero delay at the end of the sample. Thus, when the line is switched to the maximum delay, a time gap inevitably arises before the delayed signal comes out of the output end of the line. Turning the control sequence according to F i g. 11, the speech samples processed in lines 56 and 91 overlap so that the gap is filled as indicated in FIG

continuous line and shown with dashed line signal displacement sections E₁ and E ?. The presence of a small overlap in the reproduced signal does not particularly affect the intelligibility, since it is generally imperceptible and in the worst case may cause a slight echo effect of the type which often occurs during a telephone conversation. The time-expanded speech waveform, which, when using the mode of operation, is correspondingly indicated in FIG. 11

is advantageous for the recognition and understanding of difficult passages and for analysis and study of foreign languages ​​and the like.

The system shown in FIG. 12 represents a simplification of the system according to FIG. 9, where a fixed

Delay line 101 is used instead of the second variable delay line 91 of FIG. The control of the blanking circuits 92 ', 93' is simplified in that the variable blanking width B, derived from the pulse generator 94, in the

Output signals corresponding to column delayed by passing through the variable delay line 56 were generated. The fixed delay of the line 101 is selected so that a portion of the signal leaving the delay line 56 is further delayed by an amount sufficient to fill in the gap caused by the blanking pulse B, thereby substantially reducing a portion of each message.

is repeated while the variable delay line 56 is switched back to its minimum delay state. This sequence is not detrimental, as already mentioned, and merely results in a slight echo effect which is much less offensive than the presence of the gap in the message signal. This sequence of operations is shown in FIG. 13 2, wherein the variable urging portion C, alternates with the fixed urging portion Ci in providing the output.

The expansion mode of operation of the circuit according to FIG. 12 is in FIG. 14, in which the sawtooth signals for the extension waveform are controlled, which controls the delay line 56 to change from a maximum delay to a minimum delay via the linear sawtooth section E shown in Fig. H (a). The blanking waveform β is selected so that a portion of the signal pinning portion undergoes an appropriate delay to fill the gap between the perturbing portions in the output, as shown in Fig. 14 (c). Thus, the output is composed of urging sections Er and £ i which alternate to provide a continuous uniform signal.

The system of Fig. 12 can be further simplified by removing the delay line 101 and placing the gate 93 'in such a state as to introduce into the gap interval any pseudo or noise signal from a suitable source which matches the frequency content of the actual one Voice signal simulated. Although this version is less desirable than using the actual speech signal to fill in the gap, it is still better than reproducing the speech signal with existing message columns, since the clipping effect of the column becomes uncomfortable at detecting the message content, especially at high press ratios. This embodiment leads to an operation similar to that with optimum noise fill, as in connection with FIG. 7 has been described.

Fig. 15 shows a modified embodiment of the subject signal processor for binaural processing. The voice signal from the bandpass filter 53 is supplied to balanced variable delay lines VDL 1 and VDL 2, which are controlled by the signal generator 102. The output of VDL 1 passes as an input to gates 103 and 1OS. The output of VDL 2 passes as an input to the gates 104 and 106. The delay line VDL 1 is controlled in accordance with a linear variation of the delay in accordance with the waveform of FIG. 16 (c). The delay line VDL 2 is controlled according to a linear variation of the delay in accordance with the waveform of Fig. 16 (d). Each of these waveforms has its fast return transition at the center of the linear delay section of the other waveform.

The gates 103 and 106 are represented by the key shapes B_x The gate 103 is signal-transparent during Si and is blocked during B \. Gate 106 is blocked during B \ and is signal-transparent during B i. The amplifier 107 connects the outputs of the gates 103 and 106 and sends the combined signal to an audio transmitter 108.

The gates 104 and 105 are characterized by the Tastwellenformen B₂ and Bi, as shown in Fig. 16 (f). Gate 104 is during B₂ permeable to the signal and is during B₂ blocked. Gate 105 is during B₂ blocked and is for the signal during B₂ permeable. The amplifier 109 sets the outputs of the gates 104 and 105 and sends that

ι »combined signal to an audio encoder 110.

The system of Fig. 15 operates to reproduce the entire original signal (for a press ratio equal to two) since each delay line processes the section that is for the other line

16 (a) and 16 (b), for press ratios greater than two, message clearing occurs, and for press ratios smaller than two, overlap or message doubling occurs in the memory

- &lt;&gt; output size too. By binaural hearing the intelligibility is increased, since the total erasure is eliminated w^;rd (or is generally reduced for higher press ratios), and the overlap or repetition of message sections is not detrimental to word acquisition by the listener.

A binaural system without auxiliary gap filling (as just described) can be realized by removing the gates 105 and 106 in FIG. The lines VDLi and VDL: then send the processed signal alternately to the respective output transducers 108 and 110 to provide a binaural output.

Fig. 17 shows the controlled signal processor in question using a form of delay line capable of processing speech signals in a manner greatly reducing the problems of erase memory information in the line. The system illustrated in Fig. 17 is an analog shift register having a plurality of stages ASR \. ASR₂, ASR * having a voice signal input, line 111, and a compressed voice signal output on line 112. Alternate stages of the delay line are clocked by two-phase clock signals and these signals are applied to lines 113 and 114 derived from a shift frequency generator 115. The frequency fluctuation or variation of the generator 115 is chosen such that the inverse of the clock frequency, namely the pulse duration period, varies as a linear function of time, the frequency changing from a high frequency to a low frequency for a pressure and changes from a low frequency to a high frequency for stretching or stretching operation.

The analog shift register shown in Figure 17 is of the general type and is described, for example, in the article by FLJ Sangster published in the 1970 IEEE International Solid States Circuits Conference Proceedings, pp. 74-75 and 185. Such shift registers sample and send the sample from an analog signal at the clock rate through a load or load store operation, whereby the signal sample at the output of the delay line is delayed after a time delay proportional to the clock frequency.

can be recovered. Now, if the clock frequency for the purpose according to the invention modified so that the inverse of the same is a linear function of time, the delay line operates so that it expands or compresses the speech signal and by tailoring the length of the line and the repetition of the Linear control function in accordance with the principles described above, a uniform processing of random speech signals can be achieved. During the pressing operation phase at the end of each linear segment of the control function generated by the generator 115, all stages of the delay line may be reset when a reset input is available, or the line may simply be depleted during the blanking period when the line is connected to the Beginning of the next segment of the speech signal is loaded at high clock rate. This sequence can be made fast enough to make the required blanking so short that it can not be heard. In this Ausfühnme- form according to the invention, therefore, the problems are minimized, which occur in gap filling or smoothing or blanking.

The analog shift register (ASR) design parameters can thereby be obtained by applying the criteria set forth earlier, as will now be described.

The instantaneous delay r (t) of the analog shift register at time I is v (t) = dt + zo, where d is the order of change of the delay and r₀ the initial delay is. The delay dt is added in the ASR for A / stages

pl

where /, the frequency of the clock shift signal at time / is. Defining N- as W¹, so the

zoeerung

in which /₀ is the initial clock shift frequency.

The slopes of the time delay function for re-storing the original speech frequencies are the same quantities as before.

For a pressing results:

c'i = AT₁ = Total delay of the signal entering at time t = (c-1) f for a press ratio i; and

Thus, the inverse of the shift frequency as a linear function of time multiplied by / V "gives the delay required to compress or expand the speech wave, with the original speech frequencies being stored again.

Referring to Fig. 18, for the reset time i / v, to empty / V stages results

f \ =

which represents the time required for the first / V pulses to refill the line. By suppressing the sample switching transitions by filtering or blanking, or by any of the other methods described herein, and by keeping f / v less than 0.2 milliseconds, the gap-gap modulation effect discussed earlier becomes imperceptible.

It can be a limit for the sampling clock frequency determine to match the highest signal frequency f_Maei to be achieved through the line. Referring to Fig. 25, the signal frequency for the pressing decreases linearly as it is sent through the delay line indicated by the line 201. The clock frequency changes as a hyperbolic function, shown at 202, and must be equal to or greater than the values ​​on line 201 for which the relationship is given throughout the sampling period

^ zI = Zl (J - YES

c + l '\ f, fj

For an expansion results: le

c't = AT,

for the expansion ratio e, and

1-

l + e

e_ ₌ jr_ / J J_ \

et / f, fj '

2p pl

fm

(which for a two-phase ASR 4 f_Max is at least two samples per cycle of f_ma * in total.

Fig. 18 shows a modified embodiment of the analog shift register of Fig. 17, however, measures are taken for an exact substitution of processed signals at the beginning of the delay for the delay line. The analog shift register delay line 121 processes input signals from the line 122 in accordance with the variable pulse shift frequency derived from the square wave generator 123, as described earlier in connection with FIG. 17 has been described. The pulse frequency is such that the pulse spacing changes linearly, as shown by the pulse signal source 124, where the reciprocal of the frequency is linear with time. At one point 25 on the analog shift register, the line splits and merges into two shift register paths 126 and 127. The number of stages required in boxes 126 and 127 is sufficient to continue signal processing while the line 121 is being reset. The branch outputs from stages 126 and 127 are subjected to the complementary sense control units 128 and 129 and then pass as inputs to an amplifier 130 connecting these signals.

In addition to controlling the main line 121-126 from the generators 123 and 124, the branch line 127 is controlled by the gate 5131 from the pulse generator 124 which triggers a second square wave generator 132 and when this line is strobed by the button unit 133 during the β interval the trigger sequence for the generator 132 of pinpm

fixed pulse generator 134 derived. The pulse generator 134 may also be at a pulse repetition rate

- work.

The operation of the system of FIG. 18 may be described with reference to the waveform associated with the output line 135. For a given sampling period, the change in the frequency of the generator 124 begins and controls the ASR line 121, as described earlier. Under this condition, gates B pass signals and the output of stage 126 is transferred to the input of amplifier 130, thereby producing the frequency converted output signal which is displayed during the sampling period of the waveform. At the same time, the S control for the gate 131 allows the same control pulse signal from the generator 124 to trigger the generator 132, thereby maintaining the branch ASR stages 127 in synchronization with the corresponding stages 1 [26]. However, the blanking control B in the gate 129 prevents the output from the stage 127 from reaching the input of the amplifier 130. During the blanking or reset period for the main ASR line_121 and the generator 124, the B and β gates switch, interrupting the signal current from the stage 126 to the amplifier 130, and the signal current from the stage 127 to the input of the amplifier 130 can arrive. Since stages 126 and 127 are synchronous, this switching operation results in identical signals and is therefore imperceptible on output line 135 of amplifier 130. At the same time, the current of the trigger pulses from the generator 124 is interrupted by the switching corresponding to B and B in the gates 133 and 131, and trigger pulses from the generator 134 can be applied to the square wave generator 132. This switching of the control action ensures that the generator 132 continues to process the signal in the stages 127 while the generator 124 can be reset for the beginning of the next sample. At the end of the blanking period, some irregularity arises when gates B and B are returned to their original state, thereby restoring control to pulse generator 124 in that the start or commencement of the next sampling period does not result in the generation of signals. which exactly coincide with the signals that end at the end of the blanking pulse and were under the control of the pulse generator 134.

Fig. 19 shows a modified embodiment of the controlled signal processor in question, in which the variable delay line is under the control of the variable frequency generator

136 operates, in a - repetition control function exactly analogous to that as described with reference to FIG. 17. According to FIG. 19 For example, if the analog signal is not passed through successive stages of the shift register, the input signal on line 137 is first converted to a digital message in an A / D converter 138, and the parallel output thereof transmits a concurrent decision information or word the input stage of the first stage 139, this digital because then passed through the in-line stages, until it reaches an output D / A converter 140, where this value is converted to an analog signal on the output line 141. This process is completely analogous to that of the system described with reference to Figure 17, except for the coding of the information as it passes through the in-line stages driven by the variable Tat frequency, by the required frequency conversion

vorzusehenο provide. An advantage of the system of Figure 19 is the provision of a reset signal from generator 136 on line 142, which can be simultaneously applied to all registers of all stages, thereby effecting instantaneous depletion and restoration of the line at the end of the sampling period.

FIG. 20 shows a modified embodiment of the controlled signal processor of the present invention analogous to that of FIG. 19, except that the digital signal is successively processed by a serial shift register 150 after the digital output from A / D 138 in FIG Seri numsetzer 151 was serialized (serialized). The shift register 150 is under the control of the shift frequency generator 136 having the reset input line 142. The output of the serial digital shift register 150 is supplied to a parallelizer 152 (parallelizer) which simultaneously outputs the series bit string or sequence

jo converts digital decisions (parallel digital word) for conversion by D / A 140 to the required analog output signal on line 141.

Fig. 21 shows an embodiment of the controlled signal processor in question in which an analog memory array is provided with measures for addressable input and output signal storage. A load memory array 161 is shown and comprises a plurality of X write lines 162 and a second plurality of V write lines 163 whose intersections define the matrix addressing at which the analog storage elements are located. Typically, an analog memory array includes a capacitive load storage device at each intersection of an X and a V line which define the matrix to store an analog value represented by the charge on the capacitance. Each such memory location is made accessible by a plurality of X read lines 164 and a corresponding number of Y read lines 165, with the intersection of lines 164 and 165 corresponding to the location of the load memory elements at the intersections between the write lines 162 with 163 are located.

To store an analog signal in a load memory 161, an analog signal input on line 166 is applied and the current one thereof is stored in the load memory element associated with the simultaneously energized X and K write line intersection, which is terminated by a Y counter counter enable stage 167 and a K counter read word enable stage 168 are typically energized. Typically, the A and V counters 167 and 168 operate at a particular pulse rate that is one second

t&gt; 5 pulse generator 169 is derived, with the number of pulses on the, V-line sequencing the, Y-view lines 162 sequentially, whereupon the V-counter 16i is advanced and the next line dei

A 'intersections with the then active Y line is excited by the next train of pulses from the pulse generator 169. Accordingly, the memory 161 has a storage capacity of X.Y., number of storage elements corresponding to the number of intersections between X and Y lines. When the pulse generator 169 operates at a constant frequency, the input of the analog signal on the line occurs 166 at a particular sequence or rate, and the memory capacity is selected to store a signal, which was sampled in accordance with the general requirements described above.

A frequency-converted output signal is derived from the output line 171 which, as standard, receives the analog values ​​stored in the charge storage elements of the memory 161 as the matrix intersections in a regular scheme sequentially through the operation of the X-count read enable control 172 and K counters Read-only control 173. The pulse train for counters 172 and 173 is determined by a sawtooth generator 174 which controls a voltage controlled oscillator 175, the variable sequence being selected to produce the desired signal compression or strain according to the principles of the invention. For this purpose, a sequential control potentiometer 176 is provided to select the slope of the sawtooth voltage in the generator 174, and this slope is adjusted by means of a multi-gear setting in conjunction with the tape playback speed setting 52, the latter being made with reference to the embodiment of FIG F i g. 7 has been described. This dual control function is indicated by line 177. Another control signal from the sequential control potentiometer 176 is applied to the line 178 and to the pulse generator 169 to control the pulse repetition rate thereof in relation to the maximum readout speed established by controlling the ramp voltage generator 174 and oscillator 175. In particular, the write pulse train must be kept higher than the maximum readout pulse train to avoid the readout sequencer overriding the input instruction. Once any storage element has been read, it is available to store the value of the next burst and it can be reset either after reading or after the input of the next write signal. Sawtooth generator 174 provides a reset signal on line 179 to reset the count at the end of each sawtooth voltage period for the start of a next sample signal storage sequence.

Fig. 22 shows a system in which a direct access memory 181 is used and in which write and read control units 182 and 183 are provided, and which operates in a manner analogous to that according to the system of Fig. 21. Since the memory 181 stores binary information, the input signal on the line 184 in an A / D converter 185 must be lost and the corresponding output must be converted in a D / A converter! 86. The sequential control of the input and output or read-out operation for the memory array generally corresponds to that described earlier in connection with FIG. 21.

Where a gap filling is used, a special means for minimizing the disturbance caused by the irregularity at the beginnings and / or ends of the signal samples is used, and this is in Fig. 23, the gate signal control being illustrated in FIG. The logic control and sequencing elements are here arranged and configured such that the primary signal prone 191 terminates at zero zero crossing portions

undο and the auxiliary fill signal 192 at the next zero crossing point starts in the same direction and then ends at the end of a blanking period for the primary signal for the auxiliary signal in a zero crossing point, whereupon the new primary signal sample 191 at its next zero crossing in same direction takes place. Thus, after passing the low pass filters 195 and 196, the source signals 193 and 194 are directed to their respective gates 197 and 198 and voltage comparison stages 199 and 200 to remove processing and interfering high frequency components, the latter units being grounded by directional circuits 201 and 202. to trigger the pulse generator (PG) 203 or 204 whenever a positive-going NuII pass occurs at the respective signal 193 or 194. The gates 197 and 198 are operated so that the signals 193 and 194 lead to the reset output lines 205 and 206 of the flip-flops 207 and 208. The flip-flop 207 is brought into a state by a pulse on the line 209 from the gate 211, as determined by the reset output line 213 from the flip-flop 208 and by the inverted output 216 of the sampling period pulse train generator 219. The flip-flop 207 is reset by the pulse output of the gate 217, which is due to the direct output 215 of PTG 219. Similarly, the flip-flop 208 is brought into the one state by the pulse 210 from the gate 212, which is represented by the reset output on the

to line 214 from the flip-flop 208 and by the direct output 215 of PTG 219 is dependent. The flip-flop 208 is reset by the pulse output of the gate 218, as dependent on the inverted output on the line 216 of PTG 219. Gate pairs 211 and 217 or 212 and 218 are each pulsed by the output of pulse generator 203 or 204 whenever a positive-going signal goes through zero, as previously described. Thus, if flip-flop 207 is in its one state so that primary signal 191 can pass, and if flip-flop 208 is in the other state, thereby blocking filler signal 192, gate 217 allows when output line 215 of PTG becomes positive as at 220, the next pulse from PG 203 may reset flip-flop 207, thereby blocking the primary signal from reaching line 191. At the same time, the gate 212 is set ready so that the next pulse of / - distance 204

can pass through, and the flip-flop 208 in the one state, passes and the auxiliary signal can get through, until the end of the splitting period. At this time, the inverted PTG output 216 is positive as at 221, so that the gate

t&gt; 5 218 the next pulse from PC₂ 204 passes and the flip-flop 208 in the other state (reset) passes, and the auxiliary signal 192 is cut off and the gate 211 is set in standby, so that the

Next pulse from PG 203 can pass. This places the flip-hop 207 in one state, and the primary signal 191 can pass to the amplifier 222 and is output at 223. This process is then repeated in the sequence described above.

With reference to F i g. Turning now to Fig. 26, a dual delay line system having analog shift registers with separate read and write or input clock sequences will be described. As shown in Fig. 26, an input line 231 receives an input sound signal from any source, such as an input signal. B. the tape recorder which is driven at a speed different from the recording speed or any other signal source may supply a loud or sound signal which is desirably converted into frequency components and whose time duration is to be changed from zero to a longer or to a shorter time as the normal period during which the sound or sound message originated. The signal on the line 231 is fed under control to an analog shift register ASR \, passing through a gate G 233 and is also controlled by an analog shift register ASR₂ fed, passing through a gate G 234. The outputs of the analog shift registers ASRi and ASR2 are combined or connected on an output line 232 by being output from the output of ASR ^ through a gate δ 235 and from the output of ASR₂ pass through a gate G 236.

The analog shift registers ASR \ and ASR2 are multilevel registers that pass appropriately a stage-level signal input to the output, the transmissions occurring at the clock rate determined by the clock pulse sequence acting on clock terminals 237 and 238, respectively. The number of stages in the array of analog shift registers to pass through these analog signal samples to transfer, is described in US-PS 37 86 195 the same applicant. An input pulse train generator 5i sends an adjustable input pulse repetition frequency through the G gate 241 to the clock input 237 and through the (j gate 242 to the clock input terminal 238. A readout or output pulse train generator S₂ sends a relatively fixed readout clock pulse repetition rate through C gate 243 to terminal 237 and through G gate 244 to terminal 238. Gates G and G are fed by a clock pulse generator S3 which may have an adjustable period and substantially symmetrical square waves Output quantities for both G and G gate functions generated.

The input pulse train generated by the generator Si is variable as shown and is generally set in relation to a variable speed controller 245 which controls the sequence on which the tape recorder or other sound source differentiates the sound message signal in time as that of the original linguistic utterance, reproduced. Thus, if the voice controller 245 is set to play a tape recorder at twice the normal speed, the input clock pulse of the generator Si may be set to twice the clock frequency or sequence of the generator S2, thereby making the input with a clock frequency or clock sequence takes place, which is twice the sequence at which the information is read, wherein the generators Si and S2 alternately control the shift registers. If desired, a feedback control may be applied at 246 to vary the clock rate of the generator S in response to an error signal to compensate for the whine and flutter characteristics of a turntable, or to compensate for other periodic variations in the signal source that eliminates it shall be.

The frequency of the square wave generator S₃ can be adjusted by the controller 247, and their

Period T is equal to or equal to T equal to or less than

T '

where P is the phase of the analog shift register (eg, two phases per stage) and N is the total number of stages. To avoid column expansion, the period should be equal

For this purpose, the frequency control 247 of S3 m't may be coupled to the hand control for S \ during the expansion mode of operation.

In addition, trim controls 38 and 39 may be provided and may be connected to generators S2 and S3 if desired.

Now, the operation of the system of Fig. 26 will be described with reference to the waveforms of F i g. 27 (a) and (b) will be described. The general principles of operation set out in Applicant's parent application apply here, and the compression ratio C obtained is given as equal to the ratio -j-, where Λ and f₂ the

Rectangular wave frequencies are those generated by generators S \ and S₂ each generated. Obviously, for the extent C, it represents a partial size and corresponds to the expansion factor e, which is mentioned in the parent application. The input signals appearing on the line 231 pass through the gate 233 and load ASR \ during the G-tactile section, as in FIG. 27 (a) and the stages of ASR \ are filled in a sequence determined by the clock square wave on line 237 derived from generator S] via G-gate 241. During this period, there is no output from ASR \, but to ensure the absence of noise or noise signals at the output, G gate 235 blocks the signals from the output of ASR to output line 232. During G-gate operation, it sends Generator S₂ Clock pulses on the line 238 to the shift register ASR₂, and the G-gate 236 outputs signals from the output of ASR₂ to Ausgiangsleitung 232 through.

If the square wave generator S₁ changes its state, the G-gates are rendered ready to pass the signal, and the G-gates are disabled for signal passage. This way, during the G-intervals. shown in Fig. 27 (b), the signals on the input line 231 through the gate 234 to ASR₂ with the timing of the

Generator S \ sent and are sent via the G-gate 232 to the line 238 and the signals stored in the register ASR \ are output through the G-gate 235 to the Ausgangsleitujig 232 with the repetition frequency of the generator S-, which by the C Gate 242 to line 237 passes. In this way, at alternating half-cycles of G and G shown in Fig. 27, the input signal alternately in ASR ^ and ASR₂ is stored, and during the storing operation in one of the registers, the signal stored in the other register is output to the output line 232. The sequences with which this takes place are determined by the repetition frequency of the generators S \ and S₂ each determined, and for different frequencies of these generators, as already described, either a squeeze or an extension of the signal wave on the line 231 at the output or on the output line 232 can be achieved.

Therefore, according to this embodiment of the controlled signal processor in question, another form of latching delay is provided for frequency transformation using analog shift registers operated with different input and output clock sequences. This arrangement allows processing of an analog signal on line 231, including complex speech waves and similar quantities, without the requirement for digital representation of the input signal or other representation for delay and frequency transformation. Another advantage of the operation of the analog shift registers with different input and output clock sequences, which are opposite to operate as variable delay lines, is the elimination of the requirement for an inverse frequency control function gyro as in unit 115 of FIG. 17 , In the embodiment of Fig. 26, the clock sequences are fixed, but they are different for the input and output control of the analog shift registers, and the ratio of the clock sequences or frequencies directly determines the pressing ratio (or expansion ratio) that the signal undergoes this is sent through.

Although the present invention has been described with reference to frequency-time transformations of the original signal, the disclosed embodiments are also advantageous for frequency transformation due to other factors such as a change in the propagation velocity of sound waves. Human breathing in an artificial atmosphere, such as an atmosphere

high helium sphere sounds in a higher speech pitch than the normal speech pitch, but other parameters remain essentially unchanged. When using the voice press mode of operation provided by the present invention

can be achieved, the speech can be returned to its normal frequency range, without any change on a time scale.

Obviously, the method and apparatus described may be used for coded audible signals, other than speech signals such as music, by taking into consideration and considering the appropriate parameters that are important for the pressing.

For this 12 sheets of drawings