Facility for processing the signals with the controlled retardation

BACKGROUND OF THE INVENTION 1. Field of the Invention The invention relates to an apparatus for processing delayed signal signals for reproducing sound from random, continuous analog electrical signals, such as coded audible speech sounds and the like, wherein the electrical signals are constructed as analogue representations of audible sounds with to the frequency components of audible sounds.

The present invention is useful for processing speech signals in human speech or similar signals into a final, human-readable form when transmitted with components having a substantially natural or normal frequency, but at time intervals that typically differ from the original speech intervals.

Sound compression and expansion systems that use relative motion between the magnetic band as the recording environment and the air gap of the sensor head that senses the signal recorded on the magnetic environment are well known, such as described in U.S. Patent 2,352,023. However, devices of this type are subject to the usual limitations resulting from the operation, cost and weight of devices which use mainly mechanical moving parts. A version of time compression and expansion with a delay line for real-time signals is also well known from U.S. Patent No. 1,611,151. In this device, the speech signal propagates over the delay line and the movable transducer traverses over the delay line and senses the propagation signal. in it; the relative velocity between the scan head and the propagation speed in the environment indicates compression and bandwidth expansion to transmit a signal over a narrowband telephone line. Later, as is. as described in U.S. Patent No. 2,545,831, mechanical moving parts of known systems, such as that described in U.S. Patent No. 1,611,151, have been eliminated by using sequential electronic switching on the taps of the electrical delay line, thereby creating compression or expansion of the signal frequency , for transmission over narrowband lines. The use of a sequentially tapped delay line to change the duration of recorded speech signals without changing their frequency components is described in U.S. Pat. No. 3,480,737. By referring to the delay line scan rate to the propagation rate in the delay environment and the relative reproduction rate of the recorded call, compared to the delivered call from which it is derived, the recorded speech signal was temporally expanded or compressed without changing its frequency components.

Another known frequency-time transformation method uses a signal-controlled variable delay line to correct errors. Systems of this type detect undesirable frequency phenomena caused by temporal irregularities in the sequence of periodic pulses of the repetitive signal or such changes in the sound system where the speed of the recording environment passing around the sensing head is subject to periodic changes resulting in sound irregularities known as 'slow variation'. When reproducing the original signal, these systems eliminate speed errors by servo control of the time delay of the delay line connected to the signal channel. Sound systems that use speed-compensated reference or time signal tracks. reproductions by means of a variable delay line are shown, for example, in Figure 9 of U.S. Patent No. 3,202,768, where an open loop servo mechanism is shown. U.S. Pat. 3 347 997 illustrates a closed loop servomechanism that controls the reproduction rate to compensate for the relatively low frequency component of the & quot; slow variation & quot; while high frequency origin defects, i.e., & quot; variation & quot; are compensated for by variable delay lines. The function of such compensation systems depends on the repetitive nature of the error signals. By providing a delay line with a maximum time delay, proportional to the compensation of the maximum expected error, these systems do not encounter a problem that would arise in a system requiring an unlimited increase in the amount of delay in order to continuously modify the time-frequency nature of the signal.

Known systems, when used to reduce the frequency of a speech signal while compressing the time or bandwidth of a given call segment, inevitably result in the deletion of part of the original speech waveform. The ratio of the discarded speech signal to the total signal is proportional to the compression ratio, and the discard loss is naturally and basically proportional to this process of reducing the frequency and compression time for processing that portion of the call. Since the part of the call that is being transmitted alternates with the parts that have been dropped, the question of connection to transmit parts at continuous time intervals creates a particular problem for which various solutions have already been proposed.

For example, an inclined air gap in a rotating magnetic sensor head or an inclined tape approach to the point of contact with the rotating air gap has been designed so that the inlet and outlet of the recording on the magnetic tape to the air gap occur gradually because the inclined direction creates zero to maximum contact air gaps with recording environment.

In addition, various variations have been proposed, including the use of two spaced sensors rotating simultaneously with respect to the delay environment, so that a call reproduced with compressed bandwidth in one sensor is superimposed on the call from the other sensor, which would usually be omitted the first sensor.

These efforts. after compensating for discontinuities caused by periodically omitted portions of the continuous speech wave, it was essentially possible to realize only because the output signal is derived from sampling on the path over which the speech signal is propagated. Indeed, such sampling can ensure that the signal in the line is correctly accessible at any point in the line. In the aforementioned correction systems with variable delay lines, the omission of part of the call or its compensation was not taken into account. .

The present invention provides compression-expansion systems for speech or other coded signals, wherein the active element of the frequency-time conversion is a delay line responsive to the control signal and connected directly to the path between the signal source and the last loudspeaker or other device receiving the converted speech wave. Such a system is not without the problem of time compression caused by discontinuity as a result of the alternate discharge of portions of the speech wave. The discharged portion of the speech wave may be stored and then removed in the delay line or removed from the input input. delay terminal equipment. wires connected directly in the signal channel. In each case, this signal and the line switching induced transients must be omitted because the variable delay line is repeatedly controlled by the delay signal between the minimum and maximum delay values. Line switching and draining occur simultaneously with the requirement that the output signal be made up of two adjacent,. . originally separated parts of the speech wave. Therefore, both of these functions are realized by several different embodiments described herein, acting in a manner consistent with the requirements imposed by the parameters of the actual speech signal.

The main purpose of the invention is to provide a speech compression-expansion system that uses a signal-controlled variable delay line located directly in the signal channel between the signal source and the sound speaker. The variable delay line is repeatedly operated with delay values ​​lying between the maximum and minimum to change the frequency-time characteristic of the sound reproduced from the original signal.

135258

The invention provides a first controlled delay circuit with a first input terminal connected through an amplifier to an electrical signal reproducing apparatus and a second input terminal connected to a control block, the output terminal for outputting signals from the first a delay circuit with a controlled, periodically linearly increasing time delay for z. the following values ​​of electrical signals are. connected to a selection circuit, a manually operated control device for controlling the speed of the reproduction device and the frequency conversion are connected to the reproduction device and the amplifier, and the selection circuit is provided with an output filter.

The selection circuit may be connected to the first lubrication circuit as well as to a padding signal source.

In another embodiment, the selection circuit is coupled. with a controlled delay circuit to produce composite output signals derived from electrical signals coupled to the inputs of the first controlled delay circuit and the second controlled delay circuit.

A first linearly increasing pulse generator may also be connected to the first controlled delay circuit, and a second linearly increasing pulse generator may be connected to the second controlled delay circuit, the linearly increasing pulse generators being coupled to the pulse train generator.

According to a preferred embodiment, the amplifier is provided with a bandpass filter and may also be provided with a variable cut-off filter operable by the control device.

The first controlled delay circuit may advantageously be provided with a delay line with an analog representation of the electrical signals.

In a further embodiment, the first controlled delay circuit is provided with an analog sliding segment to which a control block shift frequency generator is connected.

A first lubrication circuit may be connected to the output of the first controlled circuit, and a second lubrication circuit may be connected to the output of the second delay circuit, wherein a second linearly increasing pulse generator is connected to alternately energize the first lubrication circuit and the second lubrication circuit.

Preferably, the shift frequency generator is provided with a fixed frequency source connected to the gates and another source of a different fixed frequency for any value of said factor connected to the gates, with the gate connected in the control block. connected to an analog shift register and a gate connected to another analog shift register for alternately clocking these analog shift registers at said different frequencies to shift input signals to one of said analog shift registers at one clock frequency and to shift signals from the other analog shift register at a second clock frequency, while other gates are connected to the outputs of the analog shift registers and the outputs of said other gates. are connected to alternate combining the outputs of the analog shift registers when the signals are clocked at the second clock frequency.

The device according to the invention is further characterized in that an analog-to-digital converter, a serial shift register having a controllable clock and read pulse rate, a digital-to-analog converter and a variable frequency generator for controlling the serial shift are connected in series in the first controlled delay circuit. registry.

The serial shift register can also advantageously be provided with a digital serial shift register.

BRIEF DESCRIPTION OF THE DRAWINGS The invention will now be described in more detail with reference to the drawings, in which: FIGS. / bj are compression diagrams and. Expansion showing the time relationships of the input and output signals, FIG. 3 / a) shows a set of curves representing different parameters occurring in call processing at different compression ratios greater than one, while Fig. 3 / bj shows similar relationships for ratios less than one, ie expansion, Fig. 4 / a] Figures 4 to 4 (f) show waveforms suitable for describing transition processing patterns between adjacent reproduced call patterns; Figures 5 / a) to 5 / d) illustrate a set of curves representing active transition processing between adjacent patterns; 6 (a) to 6 (e) illustrating waveforms used in describing a system that uses two delay lines to induce a transition between adjacent speech samples; Fig. 7 is a block diagram of a speech compressor-expander speech system according to the invention; Figures 8 (a) to 8 (d) show waveforms suitable for describing the operation of the system of Figure 7, Figure 9 is a block diagram of a dual delay line system of the invention; 10 (a) to 10 (d) show waveforms suitable for describing the operation of the system of FIG. 9 for compression; FIGS. 11 / aj to 11 / d) show waveforms suitable for describing the operation of the system of FIG. 9 for expansion; Fig. 12 is a partial block diagram of one of the possible modifications; Figs. 13 (j) to 13 (c) show waveforms suitable for describing the operation of the modified system of Fig. 2 for compression; Figs. 14 (a) and 14 (cj); waveforms used to describe the operation of the mo195258 diffused system of FIG. 12 for expansion, Fig. 15 is a partial block diagram of a dual delay line binaural system; Figs. 16 / a) to 16 / f) show waveforms useful for describing the operation of the system of Fig. 15; Fig. 17 is a partial block diagram; The speech signal processing apparatus of the present invention using an analog shift register as a variable delay element. FIG. 18 is a block diagram illustrating the filling of a gap with a continuous signal in a system similar to FIG. 17; 19 Dec Fig. 20 illustrates a portion of the variable delay embodiment of the invention consisting of a r-bit parallel digital shift register; Fig. 20 illustrates the variable delay embodiment of the serial digital shift register; Fig. 21 illustrates an arrangement using an analog memory matrix to generate variable delay; 22 illustrates an arrangement using a random-selection r-bit memory; FIG. 23 Fig. 24 shows the waveforms used to describe the circuit operation of Fig. 23, Fig. 25 graphically shows the clock frequency and the maximum signal frequency for the system of Fig. 17, Fig. 26 is a block diagram of a dual delay memory line system using an analog shift register with separate, read and write clock signals, and finally FIG. 27 / a) and b) show a waveform of the gating control signals for the system of FIG. 3.

Before describing the preferred embodiments of the invention, the speech signal parameters relating in particular to call compression for reproducing the call in a shorter time interval will be described. Given the basic and unavoidable limitations accompanying the compression of the speech signal, the following discussion will be carried out primarily taking into account the method and apparatus used in compression. As a result of the compression process over time, a portion of the original information is dropped, directly proportional to the compression factor, which is also a factor reducing the time required for the given speech sequence. However, the method and apparatus can also be applied to the expansion process, and considerations associated with the use of reproduced signals that take a longer time span than the original speech will be described in detail below. . The system is also able to perform a frequency transformation without corresponding time change, in order to achieve the desired frequency signal, which can be used to establish a call in an environment whose rate of propagation is different from that of airborne propagation. .

Giant. 1 illustrates a particular system using a delay line that produces a maximum time delay of 6 ms for the end portion of the sample. Provided that the processed signal is limited by frequency components between 333 and 5000 Hz, some parameters of the reproduction system for compression may be defined. The magnetic tape 21 on which the speech signal of which the lowest frequency component is 333 Hz is recorded is represented by a sine wave 22, the tape passing through the transducer 23 and being wound on the take-up reel 24 at a speed S. Electrically. the signal produced by the transducer 23 passes through the compression circuit 25 and is reproduced in the form of an audible signal and a loudspeaker 26.

The system consisting of a pick-up transducer 23, a winding head 24, a forming circuit 25 and a loudspeaker 26, shown in Fig. 1 (a), reproduces the signal recorded on the tape 21 without frequency or time change if the winding coil 24 pulls the tape through the transducer 23 at a speed. In this case, the transducer 25 would introduce a fixed, constant time delay of any value. So, in the case of the fall of FIG. 1 / b), where c = 1, the reproduction of a sine wave at a frequency of 333 Hz is shown only with a change consisting in a fixed phase delay that has been neglected.

For call compression, the tape speed increases by a factor c and the forming circuit 25 varies the delay linearly from a minimum to a maximum value. As shown in the curves of FIGS. 3 / c), 3 / d) and 3 / e), it restores a compression ratio of c = 2 for a final signal delay of 6 ms, requiring a 8 ms delay line, 12 ms of the original recorded waveform 22, which now retains half and emits half of the signal originally taking 24 ms of recorded time. This retained portion is referred to as the "remainder" and is shown in the form of the pre-processing rate, in the case of Fig. 1 / c), including the cycles numbered 1, 2, 3, and 4. is maintained at a value of 6 ms, corresponding to a delay line of 8 ms at the end of the sample, the 6 ms portion of the original information of 12 ms at the recording rate is omitted and in the case of Fig. 1 / c) is marked "omitted". This omitted portion includes cycles 5, 6, 7 and 8 of the original wave 22 and represents a gap in the information content between successive "residues" that are reproduced as audible signals. This audible output is shown in FIG. 1 / d), wherein the "rest" is shown as a portion of the tape 31 at a playback speed of 2 S and containing, cycles 1 to 4, which after processing is effectively stretched to the portion of the tape 32 occupying the original 12 ms recording time and containing cycles 1 to 4, with their original recorded frequency. In the case of Fig. 1 / d), it is indicated that the next reproduced cycle is the original wave cycle numbered 9, since cycles 5 to 8 inclusive have been omitted. The representation of a smooth transition between the end of cycle 4 and the beginning of cycle 9 in the case of Fig. 1 / d) should not be considered. a representation of the real state of the signal as would be seen from observing the real signal as compared to the idealized signal shown in FIG. 1.

The cases in Figs. 1 / f), 1 / g) and 1 / h) illustrate the situation that occurs when the compression ratio is five, that is, the tape speed drawn by the transducer 23 is five times the recording speed S At a maximum final signal delay of 6 ms, this compression ratio results in a 1.5 ms 'residual' length containing a 2½ cycle of wave 22 with a frequency component of 333 Hz in Fig. 1. 1 / a) and again a 6 ms interval, equal to the final signal delay and corresponding to a 10 ms delay line at the end of the sample. However, the information gap increased until the last half of Cycle 3 and the first half of Cycle 13 and all intermediate information of the original recorded wave were lost by the drop and this gap in the call represents 30 ms of the original recorded call.

The relationships between the parameters of the call compression system and the relationships relating to the content of the coded call information give a specification in relation to each other. optimal conditions and provide maximum limitation of the operating mode of the system. according to the invention, a given factor of clarity. These parameters can be investigated in relation to a particular system for different compression ratios, and for this reason, the parameters of a system using a delay line with a maximum final delay AT_max = - 6 ms, as shown in the following table.

Table I

Typical speech compression parameters

<td>Ratio</td><td>Kom-</td><td>Length</td><td>Cas repro-</td><td>Time</td><td>Sample-</td><td>Speed</td><td>Number</td>

<td></td><td>prese</td><td>knowledge</td><td>d ,, ,,</td><td>record</td><td>vací</td><td>repetition</td><td>cycles</td>

<td>C</td><td>d</td><td>d</td><td>Tvss / AT_max</td><td>^Tvyst / C^ATmax</td><td>period</td><td>1 / P</td><td>sample</td>

<td></td><td></td><td>T out</td><td></td><td>(ms)</td><td>T (ms)</td><td>1 / T</td><td>(f min =</td>

<td></td><td></td><td>(ms)</td><td></td><td></td><td></td><td></td><td>= 333 Hz)</td>

<td> 1,25</td><td> 2/9</td><td> 6 2/3</td><td> 24/6</td>

<td> 1,5</td><td> 2/5</td><td> 7 1/5</td><td> 12/6</td>

<td> 2</td><td> • 2/3</td><td> 8</td><td> 6/6</td>

<td> 3</td><td> 1</td><td> . 9</td><td> 3/6</td>

<td> 4</td><td> 6/5</td><td> 9 3/5</td><td> 2/6</td>

<td> 5</td><td> . 4/3</td><td> 10</td><td> 1,5/6</td>

The basis for the frequency-time transformation used in the present invention can be derived as follows. We consider the sine wave V = = E. sinwt, recorded by tape recorder. If the tape is played c times the speed of the original recording, the result is

V = E. sin c. ω t ,. (1), where c denotes the compression ratio. If c&gt; 1, time compression for any call trap occurs and if c &lt;1, the time expansion occurs by the factor e = 1 / c.

If the signal is then applied to a delay line whose delay increases linearly with time at a speed d, so that an average delay of signal c ', which represents the delay of any point of the waveguide when passing through the line, results in a signal

V - E sin (c — c'j ω t (2)).

The original signal is restored if the delay is cť = cť (for recovery) = c - 1 / t (3),

<td> 30/7,5</td><td> 30</td><td> 33,3</td><td> 10</td>

<td> 18/9</td><td> 18</td><td> 55,6</td><td> 6</td>

<td> 12/12</td><td> 12</td><td> 83,3</td><td> 4</td>

<td> 9/18</td><td> 9</td><td> 111</td><td> 3</td>

<td> 8/24</td><td> 8</td><td> 125</td><td> 2 2/3</td>

<td> 7,5/30</td><td> 7½</td><td> 133</td><td> 2 1/2</td>

<td>so c '</td><td>d 2</td><td>- (c + 2)</td><td> (4)</td>

thus, the average line delay rate, which is defined as half of the sum of the final and initial value of the delay line delay, multiplied by time t forms the delay c't.

Giant. 2 / a) shows a diagram of the relationship between the output time / t of the signal sample and the corresponding input time t_vst. Thus, the slope 4 of line I represents a signal four times the original frequency or rate of occurrence and a periodicity of 1/4, while the slope 1 of line 1 represents the resulting restored or unchanged signal. In order to convert this signal, as shown by line I with a slope c = 4, to the signal shown by line II with a slope of las corresponding to the reduced frequency, it is necessary to increase the delay of the input signal ct / v_St o the magnitude of c't, or c-1 (t) as shown in line III. So the rest of the signal, T_vst, has a ordinate that intersects line III at the ordinal value of T_vsthis value, added to the T-bar_INwith_t in point c Tv_St on line I, delaying the signal to Tout on line III. The delay dt introduced by the delay line 19 52 S8 is shown by line IV. Such a delay line causes the instantaneous signal t1 to be delayed by a linearly increasing value d. T for the interval from t_vst do t_inyst as shown by line IV. Thus, as in the case of the end signal at time t = T, it is half the sum of the initial delay dT_vst and the terminal delay dTst, gives an average value on line IV equal to c'T_vst, the value needed for recovery.

The following applies: c'T_vst = dT_INst Η ”dTout (С-l) Tvst = —I^-(Tvst + CTvst)

More generally, recovery can be achieved by a cumulative delay of the input signal Tvst ⁰ value

<img file="CS195258B2_D0001.tif" />

T output f (t)

Tvýst T_inst T_inst * dt - Ct_vst - (C 1) t (5).

For a linearly variable delay line with a delay rate d, f (t) * = '' = dt and

<img file="CS195258B2_D0002.tif" />

(C2-1) T²_vstTvst

CT_inst - (C ljTvst -

<img file="CS195258B2_D0003.tif" />

<img file="CS195258B2_D0004.tif" />

t²

Tvýst - T_inst (6), of which we receive (4)

C 'Tvst = - (C + 1) T_vst.

So d - 2 c + 1 is true

I (7) for c&gt; 1 for c &lt;1 holds 0 - d &lt;2 holds —2 &lt;d &lt;0.

In Fig. 2 / b) the corresponding relations for signal expansion are shown. Line I_Ewith a slope of 1/4 represents a signal whose frequency is a quarter of the frequency of the original rate of occurrence. To convert such a signal to a signal corresponding to line II with a slope las corresponding to a higher frequency, it is necessary to reduce the delay of the input signal CTvst (= --1— Tvst = —v- Tvst) by the value cť =

<img file="CS195258B2_D0005.tif" />

from the initial delay value c'T_INýst. This delay value, c '', shifts the signal at any point to the corresponding ordinal value on line II. The delay d 'introduced by the delay line is shown by line IVe. Such a delay line causes the instantaneous signal to be delayed by a linearly decreasing value d_vst to t, as shown by line IVe. Thus, as in the case of the initial signal at time t - o, it is half the sum of the initial delay —dT_INst and its end delay —dT increase the average delay value on line IVe straight - c'Tvs_t.

So,_T __ dTvst - dTvst

C Ivst - o

<img file="CS195258B2_D0006.tif" />

The process of linear increase in time delay cannot continue indefinitely. From time to time the delay line must be returned to its original length. If this process is repeated at periodic intervals that are longer than the lowest frequency component of the signal, the remainder of the original signal will be reproduced at an angular frequency (d — c ') ω and the remainder will be omitted. When the relationship (3) is satisfied, the system operates as if parts of the original tape were cut, glued together and reproduced at normal speed. Parts of the signal can be heard at the correct frequency, but the information is transmitted in a shorter time if c&gt; 1. The call was pressed 1 / c times its original length.

The values ​​shown in Table I were plotted in Fig. 3 / a). For any

X.9 5 2 5'8 the given compression ratio is the sampling time given by the curve Tout and · the length of the “residue” is represented by the curve L,_st. The difference between the two curves is the deletion, which is equal to the final signal delay at the end of the sampling period, ie 6 ms in the case shown in Fig. 3 / a). if the curve is considered at any compression ratio, such as c = 5 in Fig. 3 / a), the "rest" and ejection durations for the tape, running at a rate c times greater than the recording speed, are obtained , projected on the timeline, show the actual original recording time of the corresponding parts of the “rest” and the drop. As indicated for · c = 5, the residue length is 1.5 ms and the discard is 6 ms, which represents 7.5 ms of recorded and reproduced information and 30 ms of the deleted information. This last value is represented by the size c AT_maxwhich is also plotted in FIG. 3 / a).

For a speech signal whose lowest frequency of 333 Hz has a period of 3 ms, the residual length of 1.5 ms for c = 5, corresponding to a recording time of 7.5 ms, contains 2.5 cycles of the 333 Hz signal. For all higher frequency components in the speech signal, more cycles will be included in the rest of 1.5 ms. The length of the residue should exceed the lowest frequency period that occurred, ie it should cover at least the entire cycle, otherwise satisfactory compression will not be achieved. As shown in FIG. 3 / a) below the timeline at 3 ms, the 333 Hz signal would produce a low quality compressed output when processed with a sampling period close to 3 ms, with its re-collected samples, as the sampling would then cause disturbing discontinuity for almost all cycles of the 333 Hz signal being processed. Sampling periods of less than 3 ms would not allow one cycle to be completed, so that the resulting re-assembled output would not only contain these discontinuities, but would also show a fundamental change in its frequency response in the form of wave compression due to peak cuts. Incorrect frequency. Although this condition is not a real condition. For a speech wave, due to the complex nature of waveforms, the principle of control remains that sampling periods smaller than the lowest wavelength period in the speech signal do not provide proper compression.

Sampling periods greater than the lowest wavelength period produce compression and an interruption interval exists from an area where the sampling period is only slightly greater than the lowest wavelength period as indicated in Figure 3 / a) on a timeline between 3 ms and 6 ms. The result of this discontinuity period is a deformed expanded wave in which the discontinuity phenomenon between the samples manifests extremely near the point of one cycle and decreases as the number of cycles in the sample increases. The practical limit in FIG. 3 (a) is an example of two and a half cycles per sample, but generally the more cycles in the sample, the lower the failure factor.

In order to avoid extreme disturbances of waves whose wavelengths are greater than the sampling period, these lower frequencies should be filtered out before the speech signal enters the delay line, otherwise these discontinuous and highly distorted waves propagate along the line and occur. to intermodulate with the desired signal, which can seriously reduce system performance.

For values ​​of compression ratio lower than c = 5 and keeping Δ T_mex - 6 ms, the length of the residue increases with the result that the actual sample time increases above 7.5 ms and therefore more than the minimum number of cycles of the lowest frequency components will be present in the residue. Thus, it would be up to the user to work with a line generating a delay less than the indicated 6 ms for the AT_max, in order to reduce the amount of discharged parts.

Consider a dropped portion of a sample with a constant length of 6 ms in compression-reproduction, with the actual loss of information being a compression ratio of 6 ms, so at c = 5, the dropped portion of information for each sample is 30 ms of recorded time.

As shown in the timeline of FIG. 2, this portion of the interval is from 7.5 ms to 37.5 ms · and the relationship between loss of information and the intelligibility of the reproduced speech signal must be investigated.

Generally, human speech is an extremely complex coding of a relatively limited set of sounds, called phonemes, which, along with various call code attributes, such as sound-non-sound components, pitch, pitch and fluency of the sound template represented by sound energy and its absence; connected by very important transitions between their time components, they create an acoustic flow with infinite variability and versatility. The ability of the human ear to receive this acoustic information and the ear-brain information decoding system is not entirely clear, since it seems that the easily and quickly understood information far exceeds the ear acoustic response characteristics of the receiver itself.

Fortunately, the ability of the ear-brain system to understand the information transmitted by human speech signals is good enough to allow the loss or omission of significant portions of the actual acoustic flow, without noticeable loss of perception and understanding of the content of the acoustic signal information. Because understanding the content of a call decreases faster than recognizing individual words when the call is brought to the listener at increasing speed, the problem of dropping part of the signal stream can be solved with respect to clarity, and near the point at which the clarity of the individual words. This point is reached where the loss or conversion of transitions or other key components representing the connection between consonant and vowel sounds results from the deletion of many or all of the given key components, so that the apparent content of the information of the interconnected, concatenated residues is changed. Even before the point of complete loss of comprehensibility is reached, there is a margin of tolerance due to the discomfort of the listener in continuous listening. It is the result of unnatural sounds and fatigue that occurs at a tense concentration necessary to attempt to release the content of the information with excessive cutting off of time.

In order to compress a call, the loss of clarity may be associated with the omission of portions of the call containing significant key components or phonemes. The length of these components varies, with the shortest component being approximately 10 to 20 msec. While these short key components do not control the content of the call, they occur with regularity sufficient to make their systematic loss undesirable, and therefore, the required upper limit for the discharge period is considered to be 30 ms and preferably closer to 15 ms. By introducing this threshold, designed for clarity of reproduced syllables and words, the speed of speech can be increased up to the level of understanding of any listener and the degree of difficulty with minimal consideration for the limitation to be taken into account if loss or distortion of word content or creating incorrect key folders in mutually concatenated pieces of call.

Giant. 3 / aj shows the relationship of the write-off time to compression as a linear function of cAT_maxwherein the range from 18 to 30 ms is for the uncertainty of the discharge. Thus, a 6ms release at c-5 translates into a real-time recording interval from t = 7.5 ms to t = 37.5 ms that is close to the allowable upper limit of the release without undue loss of clarity as needed, so that it does not contribute significantly to the loss of clarity of the perceived call. Smaller c values ​​result in less actual dropout time, and therefore the clarity of especially those key components that occur at the beginning of the timeline, i.e. in the vicinity of 10 ms, is improved.

Although Table I and Fig. 3 / a are parameters for a typical call compression system with a final signal delay of 6 ms and defining fairly narrow limits of action, it is clear that the principles used can be adapted for use over a wider range . Thus, changing the actual frequency band of the speech signal and the maximum delay line length are both important design factors that influence the selection of the residue to the omitted portions and the sampling period for a given compression ratio range c. On the other hand, the actual frequency range of the signal is important for the design of the delay line, which must adapt to the frequency spectrum contained in the signal and to such quantitative and qualitative factors as the acoustic pitch, while retaining all or some of the shaping frequencies for individual voice and signal spectrum width in which linear phase-frequency properties need to be maintained. However, the final system used will also include selected structures, with respect to the factors contained within the broad limits described below.

Giant. 3 / b) shows a corresponding signal expansion diagram showing the initial gap, the output residue, and the change in the maximum length of the delay line with the expansion ratio and for a given input sampling interval Tv_St· The output gap appears at the beginning of each sampling period and then the time-expanded remainder of the output appears. Maximum required delay dTv_WITHt is also shown as a function of the expansion ratio e.

One aspect of the call compression system described in connection with FIG. 1 has not yet been discussed, namely the behavior of the audio output of the speaker 26 when the variable delay forming circuit 25 is switched from maximum to minimum delay at the end of the period. Just before switching, there is a speech signal on the delay line to be discharged, and if the line is immediately switched to zero delay, all this information will be represented in a largely condensed form in the output signal, unless it is crossed out or preliminarily deleted. In a conventional delay line using members R and L or C, there is practically the time interval required to switch the line from maximum to minimum delay. It has been found that although the line does not contain any signal information, a significant minimum time constant is still associated with this line switching, which creates a disturbing transient in the output signal where the repetition rate of this transient is the reciprocal of the sampling period. Due to the limitations given by the system parameters, as described above, this switching frequency and spectrum components themselves are also transient in the audio domain. Thus, they will be present as very unpleasant intermodulation components in the audio output of the device. The present invention provides a variety of methods for suppressing a transient effect and arrangements for bridging a speech gap to minimize said unpleasant noise effects. In more complex systems, the introduction of false or actual call components further improves the transition from one sample to the next and can be adapted to fill the portion that is omitted during compression.

Fig. 4 shows a portion of the 333 Hz wave at the transition point shown in Fig. 1 / dj, in which cycle 4 and cycle 9 of the originally recorded 333 Hz wave are shown as a smooth uninterrupted sine wave. The connection between the end of Cycle 4 and the beginning of Cycle 9 at point 41 is shown as a continuous portion of a sine wave, but in fact, as mentioned above, it never occurs in non-selective periodic sampling of independent complex waveforms. Instead of a smooth transition point 41, a discontinuity between the end of one portion and the beginning of the next portion in successive patterns should be expected. This discontinuity would undoubtedly not lead to a loss of clarity if:. there would be no transient phenomenon of pinning the line, either loaded or unloaded, at this very time. moment. Since this transient causes a very disturbing audio output from the system, it must be eliminated and a gate signal can be applied for this purpose, see fig. 4 / b), symmetrically with respect to the transition point 41, to produce the output signal shown in FIG. 4 / c). If the gate is long enough to include the transition effect caused by the switching of the line, the noise thus generated is eliminated. Although the improvement achieved in this way is significant, it is not ideal because the introduction of the gate signal in the audio area is audible by itself as a recurring discontinuous gap that intermodulates with the audio signal. This phenomenon can be reduced by using an output filter designed for a certain repetition rate and gate signal width to smooth the sudden transition shown in Figure 4 / c), the output response being shown in Figure 4 / d).

Further improvement can be achieved by using a gate signal in the form of a gain control signal and cutting off the transitions at “off” and perhaps “on”, so that the audio output is gradually transitioned from “off” to “on”. The result is a relatively smooth transition shown in Fig. 4 / f). The purpose is to minimize the gap effect, which has an inherent sound characteristic and can act as a key element. Narrowing the trailing edge of the gate pulse greatly facilitates this, while a premature start or proportional delay of the speech signal would be advantageous for a gradual start of the leading edge. By means of these relatively simple measures, a smooth transition between adjacent parts, which are discontinuously connected during the operation of the compression discharge process, is achieved to an extent suitable for many applications.

Fig. 5 shows more complex measures for bridging the gap between adjacent samples, which will be described below. As shown in Fig. 5 (a), the discontinuous transition that is usually expected is a sharp discontinuity in colloquial. The signal transient from the switching of the line is superimposed on it as described above. By introducing a sufficiently wide gate signal according to FIG. 5 / b), which also includes a line switching effect and adjusting the gate signal to fall to zero, and if changes in adjacent processing signals have the same direction, a zero-level gate may be achieved as shown in Figure 5. /C). This noise-free transition from line switching has been found to substantially extend the existing level of the zero amplitude signal during the gate interval, to little or no disturb the average listener.

Due to the nature of the human hearing phenomenon, especially the ability of the ear to synthesize and focus on the call even in the presence of noise and noise, it may be useful in some circumstances to introduce a false or real call component into the zero level interval indicated in Fig. ). To this end, appropriately selected noise or signal components of approximately the same amplitude and frequency may be inserted into the call into an otherwise silent gap interval. This arrangement of the invention is shown in FIG. 5 / d). Where it is necessary to fill the gap with noise components, it can be easily used during the gating interval of a suitable source and symmetrical switching to introduce noise from the source to the signal channel.

Giant. 6 illustrates a preferred form of gap filling. Two signal-controlled delay lines are used. The speech signal is applied to both delay lines, designated as channel A and channel B in Figures 6 / a) and 6 / b), the two lines being controlled by a signal to exhibit symmetrical, complementary amplitude characteristics and overlapping variable delay characteristics 6 / c) and 6 / d). Here, the delay control signals shown in FIG. 6 / d), phased so as to overlap at least in such a portion corresponding to the transition portion of the gain control characteristics of FIG. 6 / c). The inputs of both delay channels A and B are combined to form a combined output,. shown in Fig. 6 / e).

Generally, the length of the delay lines used for channels A and B in Fig. 6 is composed of one full length delay line and one shorter delay line for storing the signal that is used to fill the gap. This arrangement reduces the cost of equipment consisting of multiple delay lines required to achieve the maximum delay time for system performance requirements. On the other hand, for systems where the cost is not the primary factor, two full-length variable delay lines can be used, whose control signals can be supplied alternately so that the signal channel first passes through one and then the second delay line. This gives the total signal period for switching the inactive delay line back to the minimum delay condition before it is used for retransmission. For such symmetrical delay lines, it may again be expedient to apply some overlap during the transition, as shown in Fig. 6 / d) with suitable gain control signals, which are applied in the manner shown in Fig. 6 / c).

Referring now to FIG. 7, the basic call compression and expansion system of the present invention will now be described. The system comprises a variable speed reproduction apparatus 51, which is shown as a tape speed delivery device with a hand speed selector 52. The signal derived from the conveyance of the tape around the magnetic transducer is applied to an amplifier 53 with automatic gain control, in which the signal also passes through a band-pass filter having an adjustable low and high cutoff frequency. The selection of the cutoff frequencies for the filter can be made by the hand dial 52 in conjunction with the reproduction speed selection for the reproducing apparatus 51. The hand dial 52 also supplies an amplitude control signal to the control circuit 54 for fine adjusting the sound pitch. controlling the final amplitude of the linearly increasing waveform that controls the variable delay line as described below.

Upon passing the signal of the amplifier 53 and the filter, the signal enters the first controlled delay circuit 56, typically formed by a variable delay line, which may be controlled by the signal between the minimum and maximum delay limits. This control signal supplied to line 57 is derived from a control block 58, typically formed by a linearly increasing pulse amplitude converter, which at its input receives either a compression triangular waveform supplied by line 59 or an inverse waveform on line 59 that appears on line 61 after passing through inverter 62. A linear-rising waveform, depending on the setting of the switch 63, is supplied to one or the other of lines 59 and 61, which supplies a basic linearly rising waveform from the generator 64 of the linearly rising pulse train. The repeating period of the linearly rising waveform can be adjusted by a manually operated control device 65. An impulse identical to the zeroing of the linear portion of the linearly rising waveform appears on line 66 and is fed to the lubricating pulse generator 67 to produce output lubricating pulses whose width can be controlled by the hand adjuster 68 and which are synchronized with the output pulse on line 66 .

The output of the first delay delay circuit 56 is provided to a lubrication circuit 71, which typically has a gain, and which transmits or blocks a signal dependent on the lubrication pulse B supplied to the line 72 from the lubrication pulse generator 67, wherein the delay signal is applied to the selection circuit 7, which typically has a bandpass for the metal frequencies, the output of which is connected to the sound speaker 74.

In addition to the fixed amplitude deviation of the linearly increasing voltage signal from the generator 64, which is controlled by the hand control device 52, the absolute level of the applied voltage can be controlled by the level adjusting device 60. Generally, the variable delay line 56 may be of any known type, in particular 360 degrees RC-filters, where the derivative resistance is formed by a FET or other semiconductor arrangement that varies its resistance depending on the controlled voltage or current. In principle, such delay lines work best with respect to signal distortion passing through them if the phase delay of one stage is still kept below the maximum possible value of 90 °. Accordingly - the line can be designed to operate with a phase delay of a maximum of 45 to 60 ° per degree, whereby the number of degrees is then determined from inequality N -&gt; - (6 or 8] sfímax) Δ T_max. In this unevenness, the numbers 6 and 8 represent the number of steps per electrical cycle of the highest frequency to be passed for a phase delay of 60 ° or 45 ° as the maximum phase shift per stage to be used. Value c is the compression ratio, value f_mand x is the highest frequency passing through the line- and AT_№ is the maximum required signal delay, which is determined by the maximum allowable drain interval described above. However, many other delay line designs are known which can be controlled by a signal, and the present invention is not limited to any particular form of delay line.

The operation of the system of FIG. 7 will be described with reference to FIGS. 8 (a) and 8 (bj). The waveform 81 has a period adjustable by the hand controller 65 to produce an asymmetrical sawtooth waveform 82, thereby creating a relatively long negative linear voltage, followed by a short positive linear voltage. This waveform is used directly on the call compression line 59, while after inversion in the inverter 62 its inverse waveform is used in the expansion line 61. The expansion waveform 83 is indicated by the dashed line in Fig. 8 / a). For variable delay lines 56 whose delay increases as control voltage becomes more negative, waveforms 82 and 83 have the correct sense of controlling the delay interval, the amount of delay being determined by the hand-held amplitude control device 52 relative to the voltage level. set by the level adjusting device 60. Thus, the operating point in waveform deflection 82 is selected for a given compression ratio in conjunction with a sampling period that will be predetermined by the combination for any given compression ratio, provided that the maximum delay ATmax in line 56 'is given by a fixed value obtained by selecting the line length according to d. T_INyst according to FIG. 3 (a) and Table I for the desired compression ratio.

If the maximum signal delay is not kept constant, the discharge period as shown in FIG. 1 will vary accordingly, and the amplitude of the wave will have to be adjusted accordingly to achieve the slope d required for the compression ratio c. Similar considerations relate to the slope of the waveform 83, which must be given its corresponding value d for the expansion ratio e.

The operation of the lubrication pulse generator 67 is to provide a pulse 84 (see Fig. 8 (b)) having a predetermined width in response to a trigger pulse of the sampling period signal 81 received over the line 66. This pulse may be applied in the form of gain control to the lubrication circuit 92, which modifies the lead-in edge as described above, to reduce the transient signal and to produce a gradual start of voice audible signals to the loudspeaker 74. is selected by hand adjuster 68 and is normally selected with a duration sufficient to so that the short steep linear portion of the linearly increasing waveform can return the delay line 56 to its zero or minimum delay value and dissipate the signal energy therein or the transient caused by switching the line itself before opening the signal channel .

The lubrication period B and the period B for the expansion mode are shown in Fig. 8 / cj. The expanded residues with an initial output gap are shown in Fig. 8 / d).

The system of FIG. 7 can be used to provide rubber or false fill gap signals, which corresponds to the system described in connection with FIG. 5. For this purpose, a filler signal source 75 is used to supply an input signal to the selector circuit 73 during lubrication. interval. By means of the switch 76, this gap filling can optionally be engaged during the lubrication interval. The gap fill signal may be derived from the output speech signal of the amplifier 53.

Fig. 9 shows a modified embodiment of the invention, which is particularly suitable for implementing various gap filling procedures in the described call compression systems. Parts of Figure 9 that are substantially identical to those of Figure 7 are provided with the same reference numerals. Therefore, only the additional features and changes will be described below. At the same time as the variable delay line 56, the signal wave from the amplifier 53 receives the second delay line 91 s for the variable delay. The outputs of the delay lines 56 and 91 are coupled to lubrication circuits 92 and 93 which are complementary. The signals passing through these lubrication circuits 92 and 93 are amplified and filtered in the selection circuit 73 and are routed to the audio speaker 74 as described above.

The generator 94 generates the pulse train shown in FIG. 10 (a) having an optional pulse repetition rate determined by the setting of the hand control device 65. Thus, the base sampling period is fixed. The output pulse from the generator 94 is delayed in the delay unit 95 and applied to the first generator 96 of linearly increasing pulses, while in a delayed form it is fed to the second generator 97 of linearly increasing pulses. The linearly increasing pulse generators 96 and 97 are operated by controlling the waveform level from the hand-held level adjusting device 60 and controlling the linear amplitude increase from the hand-held control means 52. As mentioned above, a control circuit 54 for fine adjustment of the acoustic height can be used. small variations in the slope of the linearly increasing pulse as a sound pitch adjustment effective by varying the frequency conversion within a small range. Further, the amount of lubrication interval of each generator can be adjusted by adjusting devices 68 and 70. The outputs of the linearly increasing pulse generators 96 and 97 are individually coupled to delay lines 56 and 91 to control the time delay of signals passing through corresponding lines in accordance with the control signals supplied. . With the selection control of the sae factors, the sense of the slope of the linearly increasing waveform for compression or expansion can be selected. .

The level and amplitude control elements for adjusting the respective linearly increasing pulse generators 98 and 97 are preferably mutually adjustable to allow selection of the relationship between the two linearly increasing waveforms. If the delay and phasing of the delay unit 95 is adjustable, any desired overlap of the delay lines can also be achieved. It is also possible to rearrange the components to have complementary gating at the inputs of two delay lines 56 and 31 with outputs switched to combine in a common channel leading to the amplifier 53. This alternative deletes a portion of the speech signal that is not present in each line used before it enters the line, eliminating the need to disperse these parts when lines are switched between active periods.

Referring now to Fig. 10, the operation of the call compression system shown in Fig. 9 will now be described. The pulse track generator 94 generates the waveform waveform shown in Fig. 10 (a). This pulse triggers a C2 waveform transition at the linearly increasing pulse generator 97, which generates a lubrication pulse, indicated in Fig. 10 / c / m with a predetermined width V and B, which is determined by the hand width adjuster 68. lubrication pulse. After the delay indicated in FIG. ID (Ь) triggers a pulse from the generator 94 to a linearly increasing pulse generator 96 which generates the waveform C1 shown in FIG. 10 / b). With this arrangement, the control wave C1 for the delay line 56 is temporally overlapped by the waveform C2 exhibiting a slope in the same sense and spanning the waveform of the linearly rising wave C1 with a steep slope in the opposite direction. Using the asymmetric time intervals shown in FIG. 5 and 6. If the waveforms C1 and C2 are given such that they exhibit symmetrically increasing and decreasing portions, the arrangement is suitable for alternately switching lines 56 and 91, to form alternately compressed or expanded portions of the speech sample. The selection of the relative lengths of the samples passing through lines 56 and 91 will generally be determined by the manufacturing costs of the delay lines. For the main delay line 56 of adequate length, for the desired compression ratio, a relatively shorter line 91, which is only used to fill the gap, will generally be more economical. On the other hand, two full-length lines that are alternately active relative to the passage of the speech sample residues, creating reasonable times for inactive lines that return to their minimum delay state, provide any desired overlap for smooth transitions and maximum time to discharge lines to the minimum delay state before the next speech sample is processed. The operation of the system of FIG. 9 in the gap fill mode is shown in FIG. 10 / d) and essentially corresponds to the operation previously described with respect to FIG. 5.

The operation of the system of Fig. 9 for speech expansion, i.e., increasing the duration for a given speech recitation and increasing its frequency components from the loudspeaker sensed at a lower speed than the recording speed, is shown in Fig. 11. Linear waveform generators 96 and 97 have inverted outputs for generating the expansion waveforms E1 and E2 shown in FIG. 11 / a) and 11 / c), wherein the lubrication waveform has been symmetrized such that the delay lines 56 and 91 are used alternately for approximately the same periods. Due to the nature of the call expansion, there will always be a gap in the signal output because the lines are controlled to change from the maximum delay value at the start to the minimum or zero delay value at the end of the sample. Thus, if the line is switched to the maximum delay, a time gap will inevitably arise before the delayed signal from the output end of the line occurs. When the control sequence indicated in Fig. 11 is applied, the speech samples processed by lines 56 and 91 overlap to fill the gap, as indicated in Fig. 11 / d), with the full and dashed lines of the signal residues E1 and E2. The presence of a slight overlap in the reproduced signal does not greatly affect the clarity; as it usually does not appear, and at worst may result in a weak echo of the type usually encountered in a telephone conversation. The temporally expanded speech waveform obtained by the operating mode shown in FIG. 11 is useful for recognizing and understanding difficult passages and for analyzing and learning foreign languages ​​and the like.

The system shown in Fig. 12 represents a simplification of the system of Fig. 9, where a fixed delay line 101 is used instead of the second variable-delay line 91 in Fig. 9. The control of the lubrication circuits 92 ', 93' is simplified in that the variable-width lubrication pulse odvoz derived from the pulse train generator 94 generates correspondingly gaps in the output signals that have been delayed by passing through the variable delay line 56. The fixed delay of line 101 is selected to further delay any portion of the signal from the delay line 56 by a value sufficient to fill the gap caused by the lubrication pulse B. This essentially repeats some portion of each rest of the call while the variable delay line 56 is switched back to the minimum delay value. Again, this repetition is not defective and can only introduce a weak echo that is much less defective than the presence of a gap in the speech signal. This sequence of operations is shown in Figure 13, where the variable residue C is shown_in and solid residue C_F alternate at the output.

The expansion operation mode of the circuit of FIG. 12 is shown in FIG. 14, where the linearly increasing signals are inverted to produce an expansion waveform that controls the delay line 56 so that its delay varies from maximum to minimum over the linearly increasing portion. 14 (a). The erase waveform V is selected to pass through that portion of the remainder of the signal which, at a suitably selected delay, fills the gap between the residues in the output as shown in Fig. 14 / c). Thus, the output is a continuous signal consisting of alternating residues E_F and E_in.

The system of Fig. 12 could be further simplified by eliminating the delay line 101 and applying a gate 93 'to introduce any false or audio signal from a suitable source into a gap interval that simulates the frequency content of the actual speech signal. Although . would be less desirable than using the actual speech signal to fill the gap, but would be better than reproducing the speech signal in the presence of the speech gaps, since the effect of the audibility of the gaps becomes harmful for recognizing the content of the call. especially at higher compression ratios. This modification provides operational. vid, similar to filling a gap with any sound described in connection with FIG. 7.

Giant. 15 shows a modification of the invention for binaural processing. The speech signal from bandpass filter 53 is applied to symmetrical variable delay delay lines VDL1 and VDL2 controlled by waveform generator 102. The VDL1 output is applied to the gate inputs M3 and 106. The delay line VDJL1 is controlled so that its delay varies in a linear manner corresponding to the waveform of FIG. 16 / cj. The delay line VDL2 is controlled so that its delay varies linearly in a manner corresponding to the waveform in Fig. 16 / d). Each of these waveforms has its rapid reverse transition in the middle of the linear portion of the second waveform delay.

The gates 103 and 106 are controlled by the gating waves Bi and Bi shown in Fig. 16 / e). The gate 103 is open to pass the signal during Bi and closed during Bi. The gate 106 is closed during Bi and is open to pass the signal during Bi. The amplifier 107 combines the outputs of the gates 103 and 106 and provides the combined signal to the audio speaker 108.

The gates 104 and 105 are controlled by the gating waves B2 and B2 shown in Fig. 16 (f). The gate 104 is open to pass the signal during B2 and closed during B2. The gate 105 is closed during B2 and open to pass the signal during B2. The amplifier 109 combines the outputs of the gates 104 and 105 and provides the combined signal to the audio speaker 110.

System according to. FIG. 15 reproduces the entire original signal for a compression ratio of two, since each delay line processes a portion that has been omitted in the second line, as shown in FIGS. 16 / a) and 16 / bj. For compression ratios greater than two, certain portions of the call are omitted, and for ratios less than two, overlapping or duplication of output calls increases. However, in binaural listening, clarity is increased because the entire omission is eliminated or at least greatly reduced for higher compression ratios. Overlapping or repeating parts of a call does not disturb the listener when recognizing words.

A binaural system without additional gap filling, as already described, would be achieved by removing gates 105 and 166 in FIG. 15. The VDL1 and VDL2 lines would alternately feed the processed signal to the individual binaural output speakers 168 and 110.

Giant. 17 illustrates an embodiment of the invention using a delay line. capable of operating mode compression, and from frequency to frequency values ​​for operational mode expansion. a shift register, illustrated by processing speech signals in a manner that greatly reduces the problem associated with dropping information stored in the line. The system shown in Fig. 17 comprises an analog shift register 116 having a plurality of ASR1, ASR2, ... ASR stages._nwhose speech input is connected to line 111 and whose compressed or expanded speech signal is connected to line 112. The alternating stages of the delay line are clocked by the two-phase clock signals supplied by lines 113 and 114 generated by the shift frequency generator 115. The frequency change of the generator 115 is such that the inverse of the clock frequency, that is, the period from pulse to pulse, varies as a linear function of time with a frequency varying from high frequency values ​​to low frequency values ​​for .....

values ​​of low wreaths high

The analog in Fig. 17 is of the general type. Such shift registers sample the analog signal and transmit the sampled value over the line at the clock rate by storing the presence or absence of charge information that allows the signal sample to be recovered at the output of the delay line after a time delay that is proportional to the clock rate. In the present invention, where the clock rate is varied such that its inverse is a linear time function, the delay line operates by expanding or compressing the speech signal and adjusting the line length and repetition rate of the linear control function in accordance with the principles described above. continuous processing of random speech signals. In compression mode, at the end of each linear section of the control function generated by the generator 115, all delay line steps may be reset if a reset input is appropriate, or the line may be simply emptied during the erasing period, then loaded again at the beginning of the next section. high frequency clock speed. This speed can be high enough to shorten lubrication enough to make it hard to hear. Thus, in this variation of the invention, the problems associated with filling the gap or smoothing the transitions or lubricating may be minimal.

The parameters for designing the analog ASR shift register can be determined by applying the above criteria, which will now be explained.

The instantaneous delay (t) of the analog shift register at time t is r (t) - dt -jf-.₀ is the initial delay. For N stages in the analog shift register, delay e-1 1 takes d = N ---—. —7—, tt where ft is the frequency of the clock shift signal at time t. If we substitute for N

P — 1

P 'expression N1, takes the form τ (ΐ] =

<img file="CS195258B2_D0007.tif" />

where f₀ is the initial hourly scrolling frequency.

The slopes of the time delay function to restore the original call frequencies are given the same values ​​as before.

For compression:

ït - Δ Tt - total delay of the signal entering the 'compression rate' c _d - ₂ c — 1 c- + l at time t - (c — l) t for poa t · (· ¹

Ν 'ft

For expansion:

ct - ATt -

1 — et for the expansion ratio ea

1 — e l + e

Ν '/ 1 t V ft

<img file="CS195258B2_D0008.tif" />

Thus, inversion of the shift frequency as a linear function of time, multiplied by Ν ', creates the delay necessary to achieve compression or expansion of the speech waveform with the restored original speech frequencies.

Fig. 18 shows the reset time t_Nto reset N degrees, given by the relation

<img file="CS195258B2_D0009.tif" />

which represents the time required for the first N pulses of the refill lines. By suppressing sample switching transitions by filtering or deleting or by any other method described herein, and keeping the tN value below 0.2 ms, the gap intermodulation phenomenon discussed above becomes virtually indistinguishable.

There is a limit to be taken into account, for the sample clock frequency, for the location of the highest signal frequency f_mex that passes through the line. It can be seen from Fig. 25 that the frequency of the compression signals decreases linearly as the signals pass through the delay line, as indicated by waveform 201. The clock frequency varies as a hyperbolic function denoted 202 and must be equal to or greater than course 201, for which the relation is given by the formula ft š ——. F_mex, which is 4fmax for a two-phase analogue register to ensure the passage of at least two samples per cycle f_m,_x.

Giant. 18 illustrates a variation of the embodiment of the analog shift register shown in FIG. 17, but with an arrangement for accurately achieving the processed signal at the beginning of resetting the delay line. The delay line with the analog shift register 121 processes the input signals supplied by the line 122 as a function of the variable frequency of the shifting pulses generated by the rectangular waveform generator 123 described above in connection with FIG. 17. The pulse frequency is such that the distance therebetween varies linearly, as shown, from the pulse source 124 where the frequency reciprocity is indicated to be linear with respect to time. At point 125 of the analog shift register, the line branches to the double lines of stages 126 and 127 of the shift register. The number of stages required in each rectangle represented by stages 128 and 127 is sufficient to continue processing the signal even after line 121 has been reset. The outputs of stages 126 and 127 are routed to complementary control gates 128 and 129, and after gating is performed, the input signal passes from both inputs to the input of the combining amplifier 130.

In addition to controlling the main line with stage 126 by generators 123 and 124, the line branch with stage 127 is controlled by gate B131 via a pulse generator 124 that triggers second rectangular oscillator generator 132, and if gating is performed for interval B by gating unit 133 the starting speed of the generator 132 is given by the fixed pulse generator 134. The pulse generator 134 may also operate at a pulse rate.

The operation of the system of FIG. 18 can be described with reference to the waveform associated with the output line 135. For a given sampling period, the frequency change of the generator 124 of the analog shift register line 121 is triggered and controlled as described above. In this state, the gate B allows the passage of signals and the output of stage 128 is transmitted to the input of amplifier 130, thereby producing a frequency-converted output signal, indicated during the waveform sampling period. At the same time, the gate B131 will allow the same pilot pulse signal from the generator 124 to start the generator 132, thereby maintaining the state of the stages 127 of the analog shift register branch whose operation is synchronous with the corresponding stages 126. 127 reached amplifier input 130. During the lubrication or reset period for the analog shift register main line 121 and generator 124, the state of the gates B128 and B129 changes, thereby interrupting the signal flow from stage 126 to amplifier 130 while allowing the signal flow to pass from stage 125258 to amplifier input. 130. Since stages '126' and '127' worked synchronously, this switching involves identical signals and is therefore imperceptible on the output line '135' of amplifier 130. At the same time, the change in state B 'and B of the gates 133 and 131 interrupts the trigger pulses from the generator 124 and the trigger pulses from the generator 134 pass to the rectangular waveform generator 132. This switching of the control gates ensures that the generator 132 will continue to process the signals in stages 127 at a time when the generator 124 can be reset to start the next sample. At the end of the erasing period, there is some discontinuity as the gates B and B return to their original state, thereby also returning pulse generator control 124 to the extent that the beginning of the next sampling period does not produce signals that exactly match the signals ended end of the lubrication pulse, which were controlled by the pulse generator 134.

Giant. 19 illustrates a variation of the present invention wherein the variable delay line, controlled by the variable frequency generator 136, operates with the repetition frequency control function β in a manner analogous to the case described with reference to FIG. 17. 19 Dec instead of conducting the analogue signal in successive stages of the shift register, an arrangement is used in which the input signal on line 137 'is first converted to a digital word in an analogue-to-digital converter 138 whose parallel output feeds a parallel word to the first stage input registers 139 which transmits this digital value sequentially by a series of degrees, a serial shift register 150 until it reaches the output of the digital-analog 'converter' 140, where it is converted to an analog signal that appears on the output line 141. This operation is entirely analogous to the operation of the system described with reference to FIG. 17 except that the information is coded by passing it through a series of stages driven by a variable clock frequency. , to achieve the desired frequency conversion. One of the advantages of the system of FIG. 19 Dec is the generation of a reset signal by a generator 136 on the line 142, which can be applied to all registers of all stages simultaneously, thereby promptly resetting and resetting the line to its original state at the end of the sampling period.

Giant. 20 illustrates a variation of the invention which is analogous to the modification shown in FIG. 19 except that the digital signal is serially processed by the serial shift register 150 after the digital input of the analog-to-digital converter 138 has been converted to a series of signals in the respective device 141. register 150 is controlled by a shift frequency generator 136 that includes a reset input line 142. The output of the serial digital shift register 158 is provided to a device 152 that converts the serial combination to a parallel digital combination to convert the digital-to-analog converter 140 to the desired analog output signal on line 141.

Giant. 21 illustrates an embodiment of the invention in which an analog voltage matrix with address storage and signal reading is used. A memory matrix 161 having a plurality of write-downs X162 and a further number of write-downs Y163 are shown whose intersections define the addresses of the matrix in which the analog memory elements are located. Typically, the analog memory matrix will be provided with a capacitor-based storage device at each intersection of the X and Y lines defining a matrix for storing the analog value represented by the capacitor charge.

Each such memory address is also accessible via a plurality of read lines X164 and a corresponding number of read lines Y165, where the intersections of the lines 164 and 165 correspond to the address of the memory elements located at the intersections of the write lines 162 and 163.

To store the analog signal in the memory 161, the analog input signal is applied to line 168 and its instantaneous value is stored in a memory element connected to the intersection of the recording lines X 'and Y simultaneously powered by the X167 reader and the Y168 reader. . Typically, the X and Y counters 167 and 188 will operate at a predetermined pulse density given by the pulse generator 169 with such a number of pulses consecutively on the write lines X162, after which the counter Y168 is activated and the next series of intersections X with then active lines Y will be powered by the following sequence. pulses from generator 169. Accordingly, the memory 167 has a memory capacity X and Y of memory elements corresponding to the number of intersections of lines X and Y. By means of a constant frequency mipulse generator 169, an analog signal is written to line 166 at a predetermined rate, and the storage capacity is selected to store the sampled signal in accordance with the aforementioned essential requirements.

The frequency converted output signal is derived from the output line 171, which receives serial combinations of analog values, from the memory elements of the memory 161, where the matrix intersections are selected in regular order by the control reader X172 and the control reader Y173. The pulse density for the counters 172 and 173 is determined by a linearly increasing voltage generator 174, a variable speed control oscillator 175, selected to effect the desired signal compression or expansion according to the prin-chips of the present invention. To this end, it is used to select the slope of the linearly increasing voltage in the generator 174 of the pulse density potentiometer 176, which slope will essentially control the tape reproduction rate along with the hand control 52 as described with reference to the arrangement shown in FIG. 7. This dual control function is indicated by line 177. Another control function of the pulse density control potentiometer 176 is performed along line 178 and relates to the repetition rate of the pulse generator 169 with respect to the maximum read rate given by the control of the linearly increasing voltage generator 174 and the oscillator 175. In particular, the write pulse density must be maintained higher than the maximum read pulse density to prevent the phenomenon when the read pulse sequence overtaken the write pulses at the write input. Once any memory element has been read, it is desirable to store the value of the following sequence of signals, which can either be reset by reading or arriving at the next write signal. The linearly rising voltage generator 174 performs a resetting of the line 179 to reset the counters at the end of each linearly rising voltage period to initiate the following signal sampling sequence.

Giant. 22 illustrates a system using random access memory 181 with record control devices 182 and read control devices 183 that operate in an analogous manner to that described with respect to the system of FIG. 21. However, since memory 181 stores binary information, it is necessary so that the input signal on line 184 is converted in the A / D converter 185 and the corresponding output must be converted in the A / D converter 186. The sequence of the write and read pulses for the memory matrix essentially corresponds to the sequence described above for FIG. 21. The case where a special device to minimize the failure caused by discontinuity at the beginning and / or ends of the signal samples is shown in FIG. with the gating signal control shown in FIG. 24. In this case, the logic control and sequence elements are arranged to cause the first zero signal sample 191 to terminate and to start the additional fill gap signal sample 192, the next zero pass in the same direction, and then at the end of the erase period for the first end signal. said zero-crossing supplementary signal being followed by a new first signal sample 191 at its next zero-crossing in the same direction. Thus, the signal 193 and 194 from the source are for filtering through the low pass filters 195 and 196, k. removal of processing components and components of parasitic high frequencies, led to their corresponding gates · 197 and 198 and voltage comparators

199 and 200, which are coupled to ground by means of directional circuits 201 and 202 such that the PG203 or 204 pulse generator is triggered whenever the corresponding signals 193 or 194 pass through the zero position. The gates 197 and 198 are actuated to transmit signals 193 and 194 when pulses arrive from the reset output lines 205 and 198. · 206 flip-flops 207 and 208. Flip-flop

207 · Is pulsed on line 209 from gate 211, which is conditioned by a reset output line · 213 flip-flop

208 and an inverted output 216 of the pulse train generator 219 with a sampling period. The flip-flop 207 is zeroed by the pulse output of the gate 217, which is conditioned by the direct output of the pulse train generator 219 with a sampling period. Similarly, the flip-flop 208 is adjusted by the gate pulse 210, which is conditioned by the reset output of the flip-flop line 214 and the direct output 215 of the pulse train generator 219 with a sampling period. The flip-flop 208 is zeroed by the pulse output of gate 218, which is conditioned by the inverted output of line 216 of pulse train generator 219 with a sampling period. Gate pairs 211 and 217 or 212 and 218 are controlled by the pulse generator output 203 or 204 whenever the signal is positively passed through the zero position as described above. Thus, if the flip-flop 207 is in the logic "1" state, the primary signal may be

191 If the flip-flop 208 is in the “0” state, the fill signal is blocked

192 and if the pulses on the output line 215 are positive, as in the case of 220, the gate 217 will enable the next pulse from the pulse generator 203 to reset the flip-flop 207. This will block the primary signal on line 191. At the same time pulse from pulse generator 204 to set the flip-flop to logic "1" and allow the additional signal to pass through to the end of the gap period. At this point the pulse train output 216 becomes positive. as indicated by 221, allowing the gate 218 to pass the next pulse from the pulse generator 204 and reset the flip-flop 208, turn off the additional signal 192 and energize the gate 211 to pass the next pulse from the pulse generator 203. This sets the flip-flop 207 to "1". , Which allows the primary signal 191 to pass through the amplifier 222 and scan it at the output 223. The process is then repeated in the sequence described above.

Referring to FIG. 26, a dual delay line system using an analog shift register having special clock speeds for reading and storing pulses will be described. As shown in FIG. 26, the input line 231 receives input audio signals from any source, such as e.g. a tape recorder driven at a speed different from that of the recording or other source of the speech audio signal that needs to be converted to frequency components and also changed their duration from zero to some time longer or shorter than the normal interval during which the call was originally conducted. The signal on line 231 is controlled to feed to the analog shift register ASR1 by passing through the gate G233 and is also controlled to feed to the analog shift register ASR2 by passing through the gate G234. The outputs of the analog shift registers ASR1 and ASR2 are combined in output line 232, with pulses from ASR1 output passing through gate G235 and pulses from ASR2 output passing through gate G236.

The analog shift registers ASR1 and ASR2 are multistage registers adapted to pass the input signal step by step to output with transmissions at an hourly rate determined by the density of the clock pulses applied to the terminals of the clock pulses 237 and 238. Number of stages in the analog shift register arrangement for transmission the analog signal samples are in accordance with the same applicant &amp; apos; s parallel application description. In particular, the write pulse generator S1 sends a series of write pulses with adjustable repetition rate through clock G241 to clock input 237 and via gate G242 to clock input 238. The read pulse generator S2 feeds through clock G243 a series of clock pulses with relatively fixed frequency to input 237 and gate G244 to input 238. The gates G and G are controlled by a gating pulse generator Ss, which may have an adjustable period, and which generates a substantially symmetrical rectangular output for both gating functions G and G.

The velocity of the write pulses generated by the Si generator is as indicated by the variable, and will be substantially related to the variable speed control device 245 which controls the rate at which the tape recorder or other audio source reproduces the audio speech signal at a time different from the original speech. Thus, if the speed control device 245 is set to play the tape deck at twice the normal speed, the write clock pulse rate of the generator S1 may be set to twice the clock frequency of the generator S2, thus writing a clock frequency that is twice the frequency at which the information is read. in the case that the generators S1 and S2 are used alternately for controlling shift registers. If desired, feedback may be applied at location 246 to modify the clock frequency of the Si generator by an error signal to compensate for slow sound variation and gramochasis cross distortion characteristics or other periodic changes in the signal source to be eliminated.

The frequency of the square wave generator S5 may be set by the control device 247, and in general its period T will be given by

P — 1

Where P is the phase of the analog sliding register, e.g., two phases per stage and N is the total number of stages. For expansion, the gap period T should be given by the relation

For this purpose, the frequency generator control device 247 may be coupled to a manual generator control device S1 during the expansion operation.

Furthermore, fine tuning devices 28 and 39 for generators S2 and S3 may be provided and interconnected if necessary.

The operation of the system of FIG. 26 will now be described with reference to the waveforms shown in FIGS. 27 (a) and (b). The main principles of operation, stated in the basic application of the same applicant, can also be applied here, with the achieved compression ratio C equal to the ratio

12, where f 1 and f 2 are waveform frequencies generated by generators S1 and S2. Obviously, for expansion, C is the reciprocal of the value and corresponds to the expansion factor e given in the basic application. The input signals arriving on line 231 are gated by gate 233 and are received at an analog shift register ASR 1 for the duration of the gating pulse G shown in FIG. 27 / a), wherein the stages of the analog shift register ASR1 are filled at a rate determined by hourly rectangular pulses on line 237 that are generated by the generator Si and pass through the gate G241. During this period, there is no pulse at the ASRi output, but to ensure the absence of inappropriate or noise signals at the output, the gate G235 blocks the signals coming from the ASRi output to the output line 232. During the duration of the gating pulse G, the generator S2 provides clock pulses along line 238 to the shift register ASR2, and the gate G236 passes signals from the output ASR2 to the output line 232.

If the S3 generator changes its state, the G-gate transmits signals while the G-gate blocks their passage. Thus, during the interval G indicated in FIG. 27, signals on input line 231 pass through gate 234 to ASR2 at the clock frequency of generator S1 supplied by gate G242 over line 238, and signals stored in ASR1 register through gate G235 appear on output line 232 at frequency. S2, supplied by gate G242 to line 237. Thus, by alternating the half cycle G and G shown in Fig. 27, the input signal is alternately stored in an analog shift register ASR1 and ASR2, and while the signal is stored in one of the registers, the stored signal from the other register appears on the output line 232. The frequency of occurrence of these signals is determined by the repetition rates of the Si and Sz generators, and as described above, for the different frequencies of these generators, either the compression or expansion of the signal wave applied to the input line 231 can be achieved on the output line 232.

Thus, according to the present embodiment of the present invention, another form of delayed stored signal for frequency transformation using anologic shift registers operating at different input and output clock frequencies is applied. This arrangement allows the processing of analog signals supplied to line 231 including complex speech and the like waves without the need to convert the input signal to a digital combination or otherwise modify it for the delay process and frequency transformation. Another advantage of operating analog shift registers at different input and output clock frequencies when operating as variable delay lines is the elimination of the need for a function generator.