RU2773261C1 - Method and device for measuring rhythmic frequencies, power and duration of drops in the non-stationary areas of acoustic signals - Google Patents

Abstract

FIELD: measuring technology.

SUBSTANCE: invention provides for measuring rhythmic frequencies, power and duration of acoustic signal drops. The essence of the invention lies in the fact that low-frequency components are isolated from the Hilbert amplitude envelope by filtering in the form of acoustic objects (for example, sounds, words). These acoustic objects contain the most important and informative areas of unsteadiness in the form of an increase in the leading fronts of these objects or “attacks”, as well as declines of these objects. After the allocation of non-stationary sections in the form of drops, the instantaneous power of each of these sections is measured, as well as the average power of non-stationary sections is measured over a long time period. Also, in the selected areas of unsteadiness in the form of drops, the duration of the drop of each section of unsteadiness is measured, as well as the measurement of the average duration of drops is carried out over a long time period. In addition, in the selected areas of unsteadiness with increasing steepness, the frequency of occurrence of these areas (“attacks”) is measured for a given period of time (rhythmic frequencies), and the average value of these rhythmic frequencies is measured over a long time period.

EFFECT: ensuring the ability to perform a high-precision assessment of the quality of acoustic signals.

2 cl, 12 dwg

Description

Technical field

The invention relates to communication technology, in particular to digital methods and devices for measuring rhythmic frequencies, power and duration of acoustic signal decays.

State of the art

A known method for measuring instantaneous and average values of the absolute and relative power of acoustic signals (RF Patent, No. 458340 BI No. 10 dated April 10, 2012), including input signal conversion, linear analog-to-digital signal conversion, Hilbert transformation with the formation of an orthogonal signal from a digital signal, extraction in digital form of a signal corresponding to the amplitude envelope of the measured analog signal, formation of K segments of a digital signal from N code combinations in each segment from a digitally selected signal corresponding to the amplitude envelope of the measured analog signal, digital squaring, formation in each of the K segments, by summation and averaging, a digital reading corresponding to the value of the peak power of the measured analog signal in a short time interval, after which in each of the obtained K=K ₁ digital readings, by dividing by two in digital form, a digital a couple corresponding to the value of the average power of the measured analog signal in a short time interval and the obtained K=K ₁₂ digital samples are stored, and also formed from K ₁₂ digital samples, by summing and averaging these samples, a digital sample corresponding to the value of the average power of the measured analog a signal over a long time interval, after which digital indication of K ₁₂ stored digital readings and a digital reading corresponding to the value of the average power of the measured analog signal over a long time interval is carried out.

A device for measuring instantaneous and average values of the absolute and relative power of acoustic signals is known (RF Patent, No. 458340 BI No. 10 dated April 10, 2012), containing a series-connected input unit, a linear analog-to-digital converter, a Hilbert orthogonal transform unit, and a unit for calculating the amplitude envelope, digital quadrator, adder-averaging, first memory block, second memory block and indication block with display.

The disadvantage of the known method and device is the impossibility of measuring power in the most important areas of non-stationarity of acoustic signals, as well as the impossibility of measuring other parameters of acoustic signals.

The closest method to the claimed one is the method of measuring the power and the steepness of the increase in the sections of non-stationarity of acoustic signals (RF patent No. 2731339, published on 09/01/2020), including input signal conversion, linear analog-to-digital signal conversion, Hilbert transformation with the formation of an orthogonal signal from a digital signal , extracting in digital form the signal corresponding to the Hilbert amplitude envelope of the analog signal, extracting the low-frequency components of the Hilbert amplitude envelope by filtering, extracting in this filtered signal sections of non-stationarity with increasing steepness, each of which contains N _x parallel code combinations, of which, after digital squaring, formation in each section of non-stationarity, by summing and averaging, a digital reading corresponding to the value of the instantaneous power of this section of non-stationarity with increasing steepness, and then To such digital readings are stored, and also the formation of K digital readings is carried out by summing and averaging these readings, a digital reading corresponding to the value of the average power over a long time interval, consisting of K sections of non-stationarity with increasing slope, after which a digital indication of K stored digital readings corresponding to the values of instantaneous power and a digital reading corresponding to the value of the average power of the sections of non-stationarity in a long time interval, consisting of K sections, as well as on each selected section of unsteadiness with increasing steepness, the duration of this section Δt is determined and the difference in the instantaneous values of the amplitudes ΔA between end point of the non-stationarity section and its initial point, and then by dividing the difference of instantaneous amplitude values between the end point of the non-stationarity section and its initial point ΔА by the duration of this non-stationarity section Δ t, form a digital reading corresponding to the rise slope S of the non-stationarity section with increasing slope, and then K of such digital readings are stored, and also form digital readings from K, by summing and averaging these readings, a digital reading corresponding to the value of the average rise slope S over a long time interval, consisting of K sections of non-stationarity with increasing steepness, after which a digital indication is made of K stored digital readings corresponding to the values of the slope of increase S of the sections of non-stationarity and a digital reading corresponding to the value of the average slope of the increase in sections of non-stationarity over a long time interval, consisting of K plots.

A device is known for measuring the power and steepness of the increase in the sections of non-stationarity of acoustic signals (RF patent No. 2731339, published on 09/01/2020), containing a series-connected input unit, a linear analog-to-digital converter, a Hilbert orthogonal transform unit, a unit for calculating the amplitude envelope, a low-pass filter, the output of which is connected to the first input of the block of keys and the input of the block for detecting sections of non-stationarity, the first output of which is connected to the second input of the block of keys and the first input of the block for determining the steepness of the increase in sections of non-stationarity, the second input of which is connected to the second output of the block for detecting sections of non-stationarity, and the output of the block of keys connected to the third input of the block for determining the steepness of the rise of the non-stationarity sections and the input of the digital quadrator, the output of which is connected to the first input of the adder-averaging, the output of which is connected to the first input of the first memory block, the first and second outputs which are connected, respectively, to the first and second inputs of the indication block with the display, while the first output of the block for determining the steepness of the rise of the non-stationarity sections is connected to the first input of the second memory block, and its second output is connected to the second input of the first memory block and to the second input of the second a memory block, the first and second outputs of which are connected, respectively, to the third and fourth inputs of the indication block with the display, and the third output of the block for determining the slope of the increase in the non-stationarity sections is connected to the second input of the adder-averaging.

The disadvantage of the known method and device is the impossibility of measuring the power and duration of acoustic signals in such significant areas of non-stationarity as the decline sections that determine the quality of these signals, and the impossibility of measuring such an important parameter as the rhythmic frequencies of acoustic signals.

The essence of the invention

The objective of the invention is:

Expanding the functionality for measuring the rhythmic frequencies of acoustic signals, the power and duration of the decay of the sections of non-stationarity of these signals, as well as improving the accuracy of assessing the quality of these acoustic signals.

The problem is solved by using the Hilbert amplitude envelope extracted from the acoustic signal. From the Hilbert amplitude envelope, by filtering, the low-frequency components of this envelope are selected in the form of acoustic objects (for example, sounds, words). These low-frequency components or acoustic objects contain the most important and informative areas of non-stationarity in the form of rising fronts of these objects or "attacks", as well as recessions of these objects. The selection of sections of non-stationarity with increasing steepness and the selection of sections of non-stationarity in the form of recessions are carried out. After selecting sections of non-stationarity in the form of recessions containing N _x code combinations in each section, the instantaneous power of each of these sections is measured, as well as the average power of sections of non-stationarity is measured over a long time interval consisting of K such sections. Also, in the areas of non-stationarity in the form of recessions, the duration of the decline of each segment of non-stationarity is measured, as well as the measurement of the average duration of recessions over a long time interval consisting of K such sections. In addition, in areas of non-stationarity with increasing steepness, the frequency of occurrence of these sections (“attacks”) is measured at a given time interval (rhythmic frequencies), and the average value of these rhythmic frequencies is also measured over a long time interval consisting of K such time intervals. As a result of such measurements, it is possible to assess the quality of acoustic signals with great accuracy, since it is the areas of non-stationarity in the form of attacks and recessions that contain the greatest amount of information, and their distortion during transmission and processing significantly reduces the quality of these acoustic signals. These measurements will make it possible to carry out measures to reduce the distortion of non-stationary sections in acoustic signals and thereby improve their quality.

The proposed method for measuring rhythmic frequencies, power and duration of decays of sections of non-stationarity of acoustic signals, including input signal conversion, linear analog-to-digital signal conversion, Hilbert transformation with the formation of an orthogonal signal from a digital signal, digital extraction of a signal corresponding to the Hilbert amplitude envelope of the analog signal, selection by filtering the low-frequency components of the Hilbert amplitude envelope, as well as digital squaring, summation and averaging, the first storage with summation and averaging, the second storage with summation and averaging, a digital indication, characterized in that after the selection of the low-frequency components of the Hilbert amplitude envelope, they determine and extract the most important and informative segments of non-stationarity with increasing steepness and decay segments, and in the decay segments each of which contains N _x parallel code combinations of which, after digital squaring, in each section of non-stationarity, by summing and averaging, a digital sample is formed, corresponding to the value of the instantaneous power of this segment of non-stationarity with a decline, and then K of such digital samples are stored, and also digital samples are formed from K, by summing and averaging these readings, a digital reading corresponding to the value of the average power over a long time interval, consisting of K sections of non-stationarity with a decline, after which a digital indication is made of K stored digital readings corresponding to the values of the instantaneous power and a digital reading corresponding to the value of the average power of the sections of unsteadiness with a decline in a long time interval, consisting of K sections, as well as in each selected section of non-stationarity with a decline, the duration of this decline Δt is determined in the form of a digital reading, and then K of such digital readings are stored, and also carry out the formation of K digital readings, by summing and averaging these readings, a digital reading corresponding to the value of the average duration of the recessions in a long time interval, consisting of K sections of non-stationarity with recessions, after which they carry out a digital indication of K stored digital readings corresponding to the duration values decays of non-stationarity sections and a digital readout corresponding to the average value of the duration of the declines of non-stationarity sections over a long time interval, consisting of K sections, and, in addition, on the selected sections of non-stationarity with increasing steepness, the number of these sections (attacks) is determined in a given period of time in the form of the value rhythmic frequency readings, and then K of such digital readings are stored, and also digital readings are formed from K digital readings, by summing and averaging these readings, a digital reading corresponding to the average value of rhythmic frequencies for a duration long time interval, consisting of K time intervals with rhythmic frequencies, after which a digital indication is made of K stored digital readings corresponding to the values of rhythmic frequencies and a digital reading corresponding to the average value of rhythmic frequencies over a long time interval, consisting of K specified time intervals with rhythmic frequencies .

And in a device for measuring rhythmic frequencies, power and duration of the decays of the sections of non-stationarity of acoustic signals, containing a series-connected input unit, a linear analog-to-digital converter, a Hilbert orthogonal transformation unit, an amplitude envelope calculation unit, a low-frequency filter, as well as a key unit, a digital quadrator, an adder-averaging unit, a first memory block, a second memory block and an indication block with a display, a block for detecting two sections of non-stationarity, a block for determining rhythmic frequencies, a block for determining the duration of recessions and a third memory block are additionally introduced, while the first and second outputs of the Hilbert orthogonal transformation block are connected , respectively, with the first and second inputs of the block for calculating the amplitude envelope, the output of which is connected to the input of the low-pass filter, the output of which is connected to the first input of the block of keys and the input of the block for detecting two sections of non-stationarity, the first output of which connected to the input of the block for determining rhythmic frequencies, and its second output is connected to the second input of the block of keys and to the first input of the block for determining the duration of the recessions, the second input of which is connected to the third output of the block for detecting two sections of non-stationarity, and the output of the block of keys is connected to the input of the digital quadrator, the output of which is connected to the first input of the adder-averaging, the second input of which is connected to the third output of the block for determining the duration of the recessions, and the output of the adder-averaging is connected to the first input of the first memory block, the first and second outputs of which are connected, respectively, to the first and second inputs of the block indication with a display, wherein the first output of the block for determining the duration of the recessions is connected to the first input of the second memory block, and the second output of the block for determining the durations of the recessions is connected to the second input of the first memory block and the second input of the second memory block, the first and second outputs of which are connected, respectively, with the third and solid inputs of the indication unit with a display, wherein the first and second outputs of the rhythmic frequency determination unit are connected, respectively, with the first and second inputs of the third memory unit, the first and second outputs of which are connected, respectively, with the fifth and sixth inputs of the indication unit with the display.

Thanks to this solution of the problem, the proposed method and device for measuring rhythmic frequencies, power and duration of the recessions of the non-stationary sections of acoustic signals, unlike the prototype, allows you to expand the functionality for measuring rhythmic frequencies, instantaneous and average power and durations of the recessions of the sections of non-stationary acoustic signals, as well as increase accuracy of assessing the quality of acoustic signals.

List of figures

The proposed method and device are illustrated by the figures, which show:

Fig. Fig. 1. Structural diagram of the device for measuring rhythmic frequencies, power and duration of the decays of the sections of non-stationarity of acoustic signals.

Fig. 2. Block of the Hilbert orthogonal transformation.

Fig. 3. Block for calculating the amplitude envelope.

Fig. 4. Block for detecting two sections of non-stationarity.

Fig. 5. Block for determining rhythmic frequencies.

Fig. 6. Block for determining the duration of recessions.

Fig. 7. Adder-averaging.

Fig. 8. First, second and third memory blocks.

Fig. 9. Segmentation and overlay scheme of the Nuttall window function included in the block of the Hilbert orthogonal transformation.

Fig. Fig. 10. Timing diagrams of the operation of the segmentation scheme and the overlay of the Nuttall window function included in the block of the Hilbert orthogonal transformation.

Fig. 11. Scheme of overlapping segments and compensating for the unevenness of the Nuttall window function included in the block of the Hilbert orthogonal transformation.

Fig. Fig. 12. Timing diagrams of the operation of the scheme for overlapping segments and compensating for the unevenness of the Nuttall window function included in the block of the Hilbert orthogonal transformation.

Implementation of the invention

A feature of the proposed method for measuring the rhythmic frequencies, power and duration of the recessions of the sections of non-stationarity of acoustic signals, in contrast to the prototype, is the expansion of functionality that allows you to measure the rhythmic frequencies, power and duration of the declines of the sections of non-stationarity of acoustic signals, and also allows you to improve the accuracy of assessing the quality of acoustic signals.

The proposed method is based on the use of the Hilbert amplitude envelope extracted from the acoustic signal. From the Hilbert amplitude envelope, by filtering, the low-frequency components of this envelope are selected in the form of acoustic objects (for example, sounds, words). These low-frequency components or acoustic objects contain the most important and informative areas of non-stationarity in the form of rising leading edges of these objects or "attacks", as well as recessions of these objects. The selection of sections of non-stationarity with increasing steepness and the selection of sections of non-stationarity in the form of recessions are carried out. After selecting sections of non-stationarity in the form of recessions containing N _x code combinations in each section, the instantaneous power of each of these sections with a decrease is measured, as well as the average power of the sections of non-stationarity is measured over a long time interval consisting of K such sections with a decrease. Also, in the areas of non-stationarity in the form of recessions, the duration of the decline of each segment of non-stationarity is measured, as well as the measurement of the average duration of recessions over a long time interval consisting of K such sections. In addition, in each section of non-stationarity with increasing steepness, the frequency of occurrence of these sections (“attacks”) is measured at a given time interval (rhythmic frequencies), and the average value of these rhythmic frequencies is also measured over a long time interval consisting of K such time intervals . As a result of such measurements, it is possible to assess the quality of acoustic signals with great accuracy, since it is the areas of non-stationarity in the form of attacks and recessions that contain the greatest amount of information, and their distortion during transmission and processing significantly reduces the quality of these acoustic signals. These measurements will make it possible to carry out measures to reduce the distortion of non-stationary sections in acoustic signals and thereby improve their quality.

The method for measuring the rhythmic frequencies, power and duration of the recessions of the sections of non-stationarity of acoustic signals is implemented as follows. The input acoustic signal, depending on its amplitude, is subjected to transformation in the form of amplification or attenuation. Next, a linear analog-to-digital signal conversion is carried out, and the resulting signal is digitally subjected to a Hilbert transform to form an orthogonal signal. This transformation corresponds to the fact that all spectral components included in the measured acoustic signal are shifted in phase by 90°. After that, the selection in digital form of the signal corresponding to the Hilbert amplitude envelope of the measured analog signal A(t) is carried out. To do this, use a digital signal corresponding to the original acoustic signal u(t) and a digital signal after the Hilbert transformation, corresponding to the original acoustic signal, but with spectral components shifted by 90° u ₁ (t). The selection of the amplitude envelope is carried out in accordance with the formula

Further, from the signal selected in digital form, corresponding to the Hilbert amplitude envelope of the original analog signal, the low-frequency components of acoustic objects (for example, sounds or words) are isolated by filtering. These acoustic objects contain the most important and informative two areas of non-stationarity: 1) in the form of the steepness of the rise of the leading edges of acoustic objects or "attacks"; 2) in the form of recessions of these acoustic objects. How important these parameters are is shown by the following facts - it is known that the elimination of attacks from a speech signal makes it completely unintelligible, while at the same time, keeping only attacks that make up 10-15% of the duration of acoustic objects allows maintaining verbal intelligibility at the level of 85%. The elimination of attacks from the musical signal makes it impossible to identify the instrument, even for musicians. A very important parameter is the number of attacks per unit of time, which determines the rhythmic structure of speech or musical signals in the form of their rhythmic frequencies. It is the structure of rhythmic frequencies that largely determines the quality of transmitted acoustic information and the degree of its impact on listeners. For example, the same phrase, uttered with different rhythmic frequencies, can evoke completely different emotional states in listeners. The decays of acoustic objects are also a very important parameter. The decays in the signals characterize the duration of the decay of musical instruments, as well as the duration of the decay of sounds (reverberation time) in different rooms. Reverberations give the sounds volume, juiciness, richness of the timbre composition, the voices of the singers become melodious. On the other hand, at long reverberation times (long decays), a strong echo occurs, making it difficult to perceive information. Therefore, the assessment of the duration and power of the decays of acoustic objects is of great importance.

During the processing and transmission of acoustic signals, there are distortions of both attacks and recessions contained in the non-stationary areas, which significantly degrades the quality of perception of sound information. These two areas of non-stationarity of acoustic objects are determined and isolated. Moreover, each such section of non-stationarity with a decline contains N _x code combinations (discrete samples), since these sections are not identical in duration and amplitude. After that, squaring is carried out, and then summation and averaging in digital form and a value is obtained, a digital reading corresponding to the value of the instantaneous power in each section of unsteadiness with a decline, according to the formula [Mirsky G.Ya. Electronic measurements: M.; Radio and communications, 1986]

where n _i is the numerical equivalent of the instantaneous signal amplitude at the i-th sample,

N _x - the number of discrete readings in a given section of non-stationarity with a decline.

Further, K of such digital samples corresponding to the values of the instantaneous power in each section of non-stationarity with a decline, are stored, and they are also formed from these K digital samples, by summing and averaging these samples, a digital sample corresponding to the value of the average power of the sections of non-stationarity with a decline in a long time a segment consisting of K such segments. And, finally, a digital indication is made of K stored digital readings corresponding to the values of the instantaneous power in each section of non-stationarity with a decline, and a digital reading corresponding to the value of the average power of the sections of unsteadiness with a decline in a long time interval, consisting of K such sections.

Such an indication makes it possible to estimate with great accuracy both the individual values of the instantaneous power in each section of non-stationarity with a decline, and the nature of the change in the values of the average power of the sections of non-stationarity with a decline over a long time interval consisting of K such sections.

In addition, the duration of this segment Δt is determined in each section of non-stationarity with a decline. And then K of such digital readings is remembered. And also carry out the formation of K digital readings, by summing and averaging these readings, a digital reading corresponding to the value of the average duration of the recessions on a long time interval, consisting of K sections of non-stationarity with recessions. After that, a digital indication of K of the stored digital readings is carried out, corresponding to the values of the durations of the declines of the non-stationarity sections and a digital reading corresponding to the average value of the duration of the declines of the sections of non-stationarity over a long time interval, consisting of K sections.

Such an indication makes it possible to estimate with great accuracy both the individual values of the durations of the declines in each section of non-stationarity, and the nature of the change in the average values of the durations of the declines of the sections of non-stationarity over a long time interval consisting of K such sections.

In addition, the number of these sections (attacks) in a given period of time is determined in the selected sections of non-stationarity with increasing steepness in the form of a reading of the value of the rhythmic frequency. And then K of such digital readings is remembered. And also carry out the formation of K digital readings, by summing and averaging these readings, a digital reading corresponding to the average value of rhythmic frequencies over a long time interval, consisting of K time intervals with rhythmic frequencies. After that, a digital indication is made of K stored digital readings corresponding to the values of rhythmic frequencies and a digital reading corresponding to the average value of rhythmic frequencies over a long time interval, consisting of K given time intervals with rhythmic frequencies.

Such an indication makes it possible to estimate with great accuracy both the individual values of rhythmic frequencies at given time intervals, and the nature of the change in the average values of rhythmic frequencies over a long time interval, consisting of K such given time intervals.

The described method of measuring two sections of non-stationarity with increasing steepness and decay makes it possible to estimate the quality of acoustic signals with great accuracy, since it is the sections of non-stationarity that contain the greatest amount of information and their distortion during transmission and processing significantly reduce the quality of these acoustic signals. These measurements will make it possible to carry out measures to reduce the distortion of non-stationary sections in acoustic signals and thereby improve the quality of these signals.

The method is carried out using a device for measuring rhythmic frequencies, power and duration of the recessions of the sections of non-stationarity of acoustic signals (Fig. 1), which contains: a series-connected input unit 1, a linear analog-to-digital converter (LADC) 2, a Hilbert orthogonal transformation unit (BGOP) 3 , a block for calculating the amplitude envelope (BAO) 4, a low-pass filter (LPF) 5, as well as a block of keys 6 digital quadrator 7, an adder-averaging 8, the first memory block 9, the second memory block 10 and the indication block with a display (BISD) 11 The device additionally includes a block for detecting two sections of non-stationarity (BODUN) 12, a block for determining rhythmic frequencies (BFR) 13, a block for determining the duration of recessions (BODS) 14 and a third memory block 15.

In this case, the first and second outputs of the BGOP 3 are connected, respectively, to the first and second inputs of the BVAO 4, the output of which is connected to the input of the low-pass filter 5, the output of which is connected to the first input of the block of keys 6 and the input of the BODUN 12, the first output of which is connected to the input of the BORC 13 And the second output of the BODUN 12 is connected to the second input of the block of keys bis the first input of the BODS 14, the second input of which is connected to the third output of the BODUN 12. Moreover, the output of the block of keys 6 is connected to the input of the digital quadrator 7, the output of which is connected to the first input of the adder-averaging 8 , the second input of which is connected to the third output of the BODS 14. And the output of the adder-averaging 8 is connected to the first input of the first memory block 9, the first and second outputs of which are connected, respectively, to the first and second inputs of the BISD 11. In this case, the first output of the BODS 14 is connected with the first input of the second memory block 10, and the second output of the BODS 14 is connected to the second input of the first memory block 9 and the second input of the second memory block 10, the first and second outputs of the cat the first and second outputs of the BISD 11 are connected, respectively, to the third and fourth inputs of the BISD 11. Moreover, the first and second outputs of the BORC 13 are connected, respectively, to the first and second inputs of the third memory block 15, the first and second outputs of which are connected, respectively, to the fifth and sixth inputs of the BISD 11 .

The proposed method is carried out using the proposed device as follows (Fig. 1). An acoustic analog signal is fed to the input of the device and then goes to the input of the input block 1, where, depending on its amplitude, it undergoes a transformation in the form of amplification or attenuation. Then the acoustic signal from the output of input block 1 is fed to the input of LADC 2. In this block, a linear analog-to-digital signal conversion is carried out. The digital signal in the parallel code is supplied from the output of the LADC 2 to the input of the BGOP 3. In the BGOP 3, the Hilbert transformation is carried out in digital form with the formation of an orthogonal signal. This transformation corresponds to the fact that all spectral components included in the original acoustic signal are shifted in phase by 90°.

Next, the digital signal from the first and second outputs of the BGOP 3 in parallel codes is fed, respectively, to the first and second inputs of the BHAO 4. In block 4, the signal corresponding to the Hilbert amplitude envelope of the measured analog signal A(t) is extracted in digital form. For this, a digital signal is used, from the first output of the BGOP 3, corresponding to the measured acoustic signal

u(t) and a digital signal from the second output of the BGOP 3 corresponding to the measured acoustic signal, but with spectral components shifted by 90° u ₁ (t).

The selection of the amplitude envelope in BVAO 4 is carried out in accordance with

The digital signal from the output of the BVAO 4 in a parallel code is fed to the input of the LPF 5, in which the low-frequency components of the envelope in the form of acoustic objects (for example, sounds, words) are selected digitally. These low-frequency acoustic objects contain the most important and informative 2 areas of non-stationarity: the first in the form of rising fronts of these objects or “attacks”, and the second in the form of recessions. LPF 5 bandwidth from 0 to 200 Hz (in some cases up to 500 Hz). Further, the digital information signal in the form of parallel code combinations from the output of the LPF 5 is fed to the first input of the block of keys 6 and to the input of the BODUN 12, in which two sections of non-stationarity are determined: with increasing steepness and recessions. With such a definition, a signal (log. 1) appears at the first output of the HOODUN 12, corresponding to the beginning of the section of non-stationarity with increasing steepness, and at the second output of the HOODUN 12, after a certain period of time associated with the durations of acoustic objects, a signal (log.1) appears, corresponding to the beginning of the decline section, and at the third output of the HOODUN 12 signal (log. 1) appears at the end of the decline section.

The log.1 signal, corresponding to the beginning of the non-stationarity section with a decline, from the second output of the HOODUN 12 is fed to the first input of the BODS 14 and to the second input of the key block 6, as a result of which this block of keys 6 opens and a digital information signal begins to pass to its output in the form of parallel code combinations from the output of the low-pass filter 5. This digital information signal from the output of the block of keys 6 is fed to the input of the digital quadrator 7, in which the squaring of the digital information signal received in the form of a section of non-stationarity with a decline is carried out. Further, this section of the digital information signal containing N _x digital samples (parallel code combinations), from the output of the digital quadrator 7, enters the first input of the adder-averaging 8, the second input of which, from the third output of the BODS 14, receives a digital signal corresponding to the number of digital samples N _x contained in the given section of non-stationarity with decay. In the adder-averaging 8, the summation and averaging of the digital samples N _x of this section of non-stationarity is carried out and the value of the digital sample corresponding to the value of the instantaneous power in this section of non-stationarity with a decline is obtained. After the end of this section of non-stationarity with a decline in the second output BODU 12, a log signal appears. 0, which is fed to the first input of the BODS 14 and to the second input of the block of keys 6, under the action of which the block of keys 6 is closed and the digital information signal in the form of parallel code combinations ceases to be fed to the input of the digital quadrator 7. recession, the operation of the key block 6, the digital quadrator 7 and the adder-averaging 8 occurs in a similar way.

Digital samples in the form of parallel code combinations corresponding to the values of the instantaneous power of the sections of non-stationarity with a decline come from the output of the adder-averaging 8 to the first input of the first memory block 9, the second input of which receives short pulses from the second output of the BODS 14, corresponding to the ends of these sections of non-stationarity with recession. In the first memory block 9 K of such digital samples corresponding to the instantaneous power values in each section of non-stationarity with a decline, they are stored, and they are also formed from these K digital samples, by summing and averaging these samples, a digital sample corresponding to the value of the average power of the non-stationarity sections with decline over a long time interval, consisting of K such sections.

After that, K stored digital readings corresponding to the values of instantaneous power in each section of unsteadiness with a decline, from the first output of the first memory block 9, as well as a digital reading corresponding to the value of the average power of the sections of unsteadiness with a decline over a long time interval, consisting of K such sections , from the second output of the first memory block 9 are fed, respectively, to the first and second inputs of the BISD 11.

In this case, the signal is log. 1 from the second output of the BODU 12, corresponding to the beginning of the non-stationarity section with a decline, also enters the first input of the BODS 14, as a result of which the duration of this section Δt is determined in this block.

Digital readings in the form of parallel code combinations corresponding to the values of the duration of the non-stationarity sections with a decline come from the first output of the BODS 14 to the first input of the second memory block 10, the second input of which receives short pulses from the second output of the BODS 14, corresponding to the ends of these sections of non-stationarity with a decline. In the second memory block 10 K of such digital samples, corresponding to the duration values of the non-stationarity sections with a decline, are stored, and these K digital samples are formed from these K digital samples, by summing and averaging these samples, a digital sample corresponding to the average value of the duration of the non-stationarity sections with a decline on a long time a segment consisting of K such segments.

After that, K stored digital samples corresponding to the values of the instantaneous duration of the sections of non-stationarity with a decline, from the first output of the second memory block 10, as well as a digital reading corresponding to the average value of the duration of the declines of the sections of non-stationarity on a long time interval, consisting of K sections, from the second output of the second memory block 10 are fed, respectively, to the third and fourth inputs of the BISD 11.

In this case, the log.1 signal from the first output of the HOODUN 12, corresponding to the beginning of the section of non-stationarity with increasing steepness, is fed to the input of BORC 13. In this box 13, in the selected sections of non-stationarity with increasing steepness, the number of these sections (attacks) is determined in a given period of time in the form counting the value of the rhythmic frequency.

Digital readings in the form of parallel code combinations corresponding to the values of rhythmic frequencies at given time intervals are received from the first output of the BORC 13 to the first input of the third memory block 15, the second input of which is fed with short pulses from the second output of the BORC 13, corresponding to the end of the specified time intervals. In the third memory block, 15 K of such digital readings corresponding to the values of rhythmic frequencies at given time intervals are stored, and these K digital readings are formed from these K digital readings, by summing and averaging these readings, a digital reading corresponding to the average value of rhythmic frequencies over a long time interval, consisting of K such given time intervals.

After that, K stored digital readings corresponding to the values of rhythmic frequencies at given time intervals, from the first output of the third memory block 15, as well as a digital reading corresponding to the average value of rhythmic frequencies over a long time interval, consisting of K such given time intervals, from the second the outputs of the third memory block 15 are fed, respectively, to the fifth and sixth inputs of the BISD 11.

The proposed device for measuring rhythmic frequencies, power and duration of the recessions of the sections of non-stationary acoustic signals, in contrast to the prototype, allows you to expand the functionality for measuring rhythmic frequencies, instantaneous and average power and duration of the decays of the sections of non-stationary acoustic signals, as well as improve the accuracy of assessing the quality of acoustic signals.

A feature of the proposed device for measuring rhythmic frequencies, power and duration of the decays of the sections of non-stationarity of acoustic signals is that non-standard in it are: the block of the Hilbert orthogonal transformation BGOP 3, the block for calculating the amplitude envelope BVAO 4, the adder-averaging 8, the first, second and third memory blocks 9, 10, 15, as well as a block for detecting two sections of non-stationarity BODUN 12, a block for determining the rhythmic frequencies of BORCH 13 and a block for determining the duration of recessions BODS 14.

An example implementation of the block Hilbert orthogonal transform (HOP) 3 is shown in Fig2. This block contains the following connected in series: a Nuttall window function segmentation and overlay circuit (SSNOFN), a direct discrete Fourier transform (SDFT) circuit, a phase rotation circuit of the transform coefficients (SPFKP), an inverse discrete Fourier transform (IDFTF) circuit, a circuit for overlapping segments and compensating for the unevenness of the window Nuttall functions (SPSKNOFN). In addition, BGOP 3 contains a sampling pulse frequency doubling circuit (SUCHID) and a delay line. The first (code) input of SSNOFN is connected to the input (code) of BGOP 3 and the first (code) input of the delay line, and the code output of SSNOFN is connected through serially connected SPDPF, SPFKP, SODPF to the code input of SPSKNOFN, the code output of which is connected to the second code output of BGOP 3. The second input of SSNOFN is connected to the second input of SPSKNOFN, the second input of the delay line and the input of SUCHID, the output of which is connected to the third input of SSNOFN, the third input of SPSKNOFN, the second input of SPDPF, the second input of SPFKP and the second input of SODPF. The code output of the delay line is connected to the first code output of the BGOP 3

The operation of the Hilbert Orthogonal Transform (HOP) block 3 is based on the expression for the direct and inverse discrete Fourier transform (DFT)

where x(n) is a sequence of B time samples, X(k) is a sequence of B frequency samples.

Block BGOP 3 operates as follows (Fig. 2). The input (code) BGOP 3 receives parallel code combinations from the output of LADC 2 (Fig. 1). These code combinations inside the BGOP 3 (Fig. 2) are fed to the first (code) input of the delay line and to the first (code) input of the SSNOFN, the second and third inputs of which receive, respectively, the sampling frequency pulses from the LADC 2 (in Fig. 1 circuit not shown) and double-sampled pulses from the SUCHID output. In SSNOFN, segments are formed, consisting of B parallel code combinations in each segment, corresponding to B time discrete samples of the audio signal. A Nuttall window function is then applied to each segment. The digital signal in the form of segments of B parallel code combinations in each segment from the code output of the SSNOFN is fed to the code input of the SPDPF, where the B point direct discrete Fourier transform of these B parallel code combinations is performed in each segment.

The need to impose the Nuttall window function is due to the fact that the discrete Fourier transform (DFT) uses a rectangular window without overlap, which leads to discontinuities in the analyzed functions. The resulting window transformation sidelobes in the spectrum, called leakage, will distort the amplitudes of neighboring spectral components. To reduce the level of distortion and interference, it is necessary to minimize such leakage of side-lobe energy into the main components of the signal. Obviously, the lower the level of side lobes of the window function in the frequency domain, the higher the accuracy of the direct discrete Fourier transform. The Nuttall window has the lowest level of side lobes among the existing window functions.

As a result of the B point direct discrete Fourier transform B of the code combinations in the SPDF, B pairs of coefficients are formed corresponding to the representation of the digital acoustic signal in the spectral domain. Next, the digital signal from the code output of the SPDT is fed to the code input of the SPDT, where the phase of the conversion coefficients is rotated by changing the sign of the coefficient at jsin 2πnk/V in each pair of coefficients, which corresponds to a 90° phase rotation of all spectral components in the time domain in the original analog signal .

Then the digital signal from the code output of the SPFKP is fed to the code input of the SODFT, where the B point inverse discrete Fourier transform is carried out from B pairs of coefficients to B code combinations in each segment.

After that, the digital signal from the code output of the SODPF is fed to the code input of the SPSKNOFN. This scheme is necessary for better signal recovery in the case of using the Nuttall window, for which addition is additionally carried out with 50% overlap. For this purpose, in SPSKNOFN, addition is carried out with 50% overlap of each segment with the previous segment, delayed by a duration equal to half the duration of the segment. Since the Nuttall window is not one of the windows that provides unity gain when using 50% overlap, an additional increase in the accuracy of the reconstructed digital acoustic signal is carried out by compensating for the unevenness of the Nuttall window function. Such compensation makes it possible to increase the protection ratio, which characterizes the level of noise and distortion in the signal, up to 92 dB, which is essential for improving the accuracy of the measuring device.

The digital signal from the code output of the SPSKNOFN is then fed to the second code output of the BGOP 3.

Thus, Hilbert orthogonal transformation of the digital signal was carried out in BGOP 3, corresponding to the phase rotation of all spectral components of the analog signal by 90°. However, this digital signal, after passing through the SSNOFN, SPDPF, SPFCP, SODPF and SPSKNOFN, acquired a time delay. For the normal operation of the block for calculating the amplitude envelope (BAO) 4, it is necessary that the original digital signal received at the code input of the BGOP 3 would have exactly the same time delay at the first code output of this block as the digital signal at its second code output. For this purpose, a delay line is used in BGOP 3.

A feature of BGOP 3 is that non-standard in it are SSNOFN and SPSKNOFN, which require additional disclosure. These blocks and timing diagrams of their operation are shown in Fig. 9 - fig. 12.

Scheme of doubling the frequency of the sampling pulses (SUCHID), included in the BGOP 3, can be made in the form of connected in series: a meander shaper, a differential circuit, a full-wave rectifier and a short pulse shaper.

An example of the implementation of the block for calculating the amplitude envelope (BVAO) 4 is shown in Fig.3. BVAO 4 consists of the first and second squaring circuits, an adder and a square root extraction circuit. The code input of the first squaring circuit is connected to the first input of the BVAO 4, and the code input of the second squaring circuit (SVK) is connected to the second input of the BVAO 4, and the code outputs of these circuits are connected, respectively, to the first and second code inputs of the adder. The code output of the adder is connected to the code input of the square root extraction circuit, the code output of which is connected to the code output of BVAO 4.

Operation of BHAO 4, i.e. selection of the amplitude envelope is carried out in accordance with the expression [1]: A(t)=[u ² (t)+u ₁ ² (t)] ^1/2 . For this, a digital signal is used from the first output of the BGOP 3 (figure 1), corresponding to the original acoustic signal u(t) and a digital signal from the second output of the BGOP 3, corresponding to the original acoustic signal, but with spectral components shifted by 90° u ₁ ( t). In the first and second ICS, the squaring of the numerical values of each parallel code combination (corresponding to the readings of the instantaneous amplitudes of the original analog signal) is carried out in digital form. Further, the digital signal in the form of parallel code combinations from the code outputs of the first and second ICS are fed to the first and second code inputs of the adder, respectively. In this scheme, the summation of the numerical values of the code combinations coming to the 1st code input of the adder with the corresponding code combinations coming to the 2nd code input of the adder is carried out in digital form. This operation corresponds to the expression u ² (t)+u ₁ ² (t). After that, the digital signal in the form of parallel code combinations from the code output of the adder is fed to the code input of the square root extraction circuit (SICC). In this scheme, the operation of extracting the square root from the numerical values of the code combinations obtained after summation is carried out in digital form. This operation corresponds to the expression [u ² (t)+u ₁ ² (t)] ^1/2 . The digital signal corresponding to the selected Hilbert amplitude envelope of the analog signal is fed from the code output of the SICC to the code output of the BVAO 4.

An example of the implementation of the block detection of two sections of non-stationarity (BODUN) 12 is shown in Fig.4. HOODUN 12 consists of a delay line (LZ), a subtraction circuit (SV), a sign detection circuit (SVZ), an inverter, a decoder, the first OR ₁ circuit, the second OR ₂ circuit, the first, second, third and fourth short pulse shapers F ₁ , F ₂ , F ₃ , F ₄ , the first RS flip-flop T ₁ , the second RS flip-flop T ₂ , the first circuit AND ₁ , the second circuit And ₂ and the third circuit And ₃ . The input (code) BODUN 12 is connected inside the block with the code input LZ and the first code input of the SV, the second code input of which is connected to the output of the LZ, and the output of the SV is connected to the input of the SVZ and to the input of the decoder, the outputs of which from the first to n are connected to the inputs of the first circuit OR ₁ , and the outputs of the decoder from n+1 to m are connected to the inputs of the second circuit OR ₂ , and the outputs OR ₁ and OR ₂ are connected, respectively, to the input of the first shaper F ₁ and the input of the second shaper F ₂ , the outputs of which are connected, respectively , with the R-input and S-input of the first RS flip-flop T ₁ , the direct output of which is connected to the first input of the first circuit AND ₁ and the first input of the second circuit AND ₂ , and its inverse output is connected to the first input of the third circuit AND ₃ , while the output of the SVZ is connected to the input of the fourth shaper F ₄ and the input of the inverter, the output of which is connected to the input of the third shaper F ₃ , the output of which is connected to the S-input of the second RS flip-flop T ₂ , the R-input of which is connected to the output of the fourth shaper F ₄ , and the direct output of the second RS flip-flop T ₂ is connected to the second input of the first circuit AND ₁ , and its inverse output is connected to the second inputs of the second circuit AND ₂ and the third circuit AND ₃ , and the outputs AND ₁ , AND ₂ and AND ₃ connected, respectively, with the first, second and third outputs of the HOODUN 12.

Work BODU 12 (figure 4) is carried out as follows. In the initial state, the direct outputs of the first RS flip-flop T ₁ and the second RS flip-flop T ₂ are present log. 0. At the input HOBUN 12 receives parallel code combinations from the output of the low-pass filter 5 (figure 1), and inside the HODUN 12 these code combinations are fed to the input of the LZ and the first input of the SW. In the LZ, these parallel code combinations are delayed for the time of one discrete sample. Further, these delayed code combinations from the output of the LZ are fed to the second input of the SV. In the SV, from each code combination arriving at its first input, the code combination delayed by one discrete count, arriving at its second input from the LZ output, is subtracted. As a result, parallel code combinations appear at the output of the CB, corresponding to increments in the amplitude of the Hilbert amplitude envelope of the analog signal, with a positive or negative sign. In this case, the positive or negative sign of the amplitude increment is contained in the most significant bit of each parallel code combination, and for positive increments (attack) it is a log. 0, and for negative increments (decay) it is a log. 1. Then the parallel code combinations from the output of the CB are fed to the input of the CV and to the input of the decoder. The decoder is designed to separate sections of the Hilbert amplitude envelope with increasing (attacks) or falling steepness (falls) from sections whose slope is insignificant or equal to zero. It is known [Popov O.B., Richter S.G. Digital processing and measurements of signals in audio broadcasting paths. M. Insvyazizdat, 2010, p. 30] that the maximum duration of non-stationarity sections with increasing steepness (attacks) is approximately Δt ₁ =300 ms, and the minimum duration is 2 ms, but most often take the value Δt ₂ =5 ms. At the maximum duration of the non-stationarity section, there will be a minimum steepness of this section, which, at the maximum value of the Hilbert envelope signal amplitude, equal, for example, to 3 volts, will be S _min = ΔA / Δt ₁ = 3000 mV / 300 ms = 10 mV / ms. This maximum duration of the section of non-stationarity with increasing steepness fits 14423 discrete samples (code combinations) at a standard sampling rate in LADC 2 (Fig. 1) equal to 48 kHz. In this case, ΔA ₁ per reading will be ΔA ₁ = 3000 mV / 14423≈0.208 mV = 208 μV. This value of the amplitude increment ΔA ₁ = 208 μV is taken as the beginning and end of the non-stationarity section with increasing steepness and for the beginning and end of the decay section (with falling steepness). In this case, at the minimum duration of the non-stationarity section in Δt ₂ = 5 ms, there will be a maximum steepness of this section, which will be S _max = ΔA/At ₂ = 3000 mV / 5 ms = 600 mV / ms. When using a 16-bit code in the LADC 2 (FIG. 1), the quantization step will be approximately 45.7 μV. In this case, the magnitude of the amplitude increment ΔA ₁ = 208 μV, corresponding to the beginning and end of the non-stationarity section with increasing steepness and the beginning and end of the decay section (with falling steepness), will correspond to the code combination 0000000000000101 at the CB output (Fig. 4).

Parallel code combinations corresponding to increments in the amplitude of the Hilbert amplitude envelope of the analog signal per sample ΔA ₁ from the output of the CB are fed to the input of the decoder (Fig. 4). In the case when the amplitude increment in the non-stationarity section with increasing steepness and in the decay section is equal to 0, then the code combination 0000000000000000 will appear at the output of the CB and the input of the decoder, under the action of which a log.1 level appears at the first output of the decoder. This level jump passes through OR ₁ to the input of the first shaper F ₁ , at the output of which a short pulse appears, which is fed to the R-input of the first RS flip-flop T ₁ . However, the state of this trigger remains unchanged and its direct output still has a log level. 0. Further, in the case when the amplitude increment in the first (attack) or second (decline) section of non-stationarity begins to increase slightly, then at the output of the CB and the input of the decoder there will be code combinations 00000000000000001, 00000000000000010, , respectively, the second, third and fourth (n=4 in Fig. 4) outputs of the decoder will consistently appear log.1 levels. These level jumps pass through OR ₁ to the input of the first shaper F ₁ , at the output of which short pulses will appear, which are fed to the R-input of the first RS flip-flop T ₁ . However, the state of this trigger still remains unchanged and its direct output still has a log level. 0. And only when the code combination 0000000000000101 appears at the output of the CB and the input of the decoder, corresponding to the beginning of the non-stationarity section with increasing steepness or the beginning of the decline section, a log level appears at n + 1 = 5 (Fig. 4) of the decoder output. 1. This level jump passes through OR ₂ to the input of the second shaper F ₂ , at the output of which a short pulse appears, which is fed to the S-input of the first RS flip-flop T ₁ . This trigger fires and a log.1 level appears at its direct output, and a log level will appear at its inverse output. 0. Level log. 1 from the direct output of the trigger T ₁ arrive at the first inputs of the first circuit And ₁ and the second circuit And ₂ , and the level of the log. 0 from the inverse output T ₁ is supplied to the first input of the third circuit And ₃ . Further, in the case of a further increase in the amplitude in the section of non-stationarity with increasing steepness or in the section with a decline, increasing code combinations 0000000000000110, 0000000000000111… 6, n+3=7 …m outputs of the decoder (Fig. 4) will sequentially appear log levels. 1. These level jumps pass through OR ₂ to the input of the second shaper F ₂ , at the output of which short pulses will take place, which are fed to the S-input of the first RS flip-flop T ₁ . However, the state of this trigger remains unchanged and its direct output will still have a log level. 1, and on the inverse level log. 0.

As the amplitude of the Hilbert envelope approaches its maximum value in the non-stationarity section with increasing steepness and reaches its flat part, and as the amplitude of the envelope approaches zero in the non-stationarity section with a decrease, the increments in the amplitude of this envelope begin to decrease, and, consequently, the values of the code combinations on the output of the CB and the input of the decoder will also decrease. However, through OR ₂ and F ₂ from the output of the decoder, there will still be jumps in the log level. 1 to the S-input of the first RS-flip-flop T ₁ , and its direct and inverse output will still be, respectively, the log level. 1 and level log. 0. And only with the appearance of the code combination 0000000000000100 at the output of the CB and the input of the decoder, the log level appears on the fourth (n=4 in Fig. 4) output of the decoder. 1. This level jump passes through OR ₁ and the first shaper F ₁ to the R-input of the first RS flip-flop T ₁ . This trigger fires and a log level appears at its direct output. 0, and its inverse output will have a log level. 1. Log level. 0 from the direct output of the trigger T ₁ are fed to the first inputs of the first circuit And ₁ and the second circuit And ₂ and the level of the log. 1 from the inverse output T ₁ is supplied to the first input of the third circuit And ₃ . This operation of the first RS-trigger T ₁ indicates that the section of non-stationarity with increasing steepness, or the section of the decline was detected, determined. Further, as the amplitude increment decreases further in the first (attack) or in the second (fall) section of non-stationarity, decreasing code combinations will occur at the output of the CB and the input of the decoder, up to 0000000000000000, and on the direct and inverse outputs of the RS flip-flop T ₁ will still be, respectively, the level of the log. 0 and log level. 1. Further, the work of this part of the HOODUN 12 (Fig. 4) is similar.

Since the low-frequency components of acoustic objects extracted from the Hilbert amplitude envelope contain not only sections of non-stationarity in the form of an increase in the leading edges of these objects or "attacks", but also contain sections of declines, then for the duration of these sections of the decline, fixing sections of non-stationarity with increasing steepness (attacks) , must be stopped. And vice versa, for the duration of the sections with increasing steepness, the fixation of sections of non-stationarity with recessions should be stopped. For this purpose, another part of the HOODUN 12 (Fig. 4) is used, consisting of a sign detection circuit (SVZ), an inverter, the third and fourth shapers of short pulses F ₃ and F ₄ and the second RS flip-flop T ₂ . The work of this part of HOLDING 12 is as follows. During the operation of the subtraction circuit (SW) in each parallel code combination at its output, the most significant bit of this combination corresponds to the sign of the amplitude increment - positive (increase) or negative (decrease). For positive increments, this is a log. 0, and for negative it is log.1. This most significant bit of each parallel code combination is fed to the input of the SVZ, and from its output it is fed to the input of the inverter (NOT circuit) and to the input of the fourth shaper F4.

When parallel code combinations appear at the output of the SV, corresponding to the section of non-stationarity with increasing steepness, signals with a log level will be received from the output of the SVZ to the input of the inverter and input F ₄ . 0. The first signal with a log level. 0 is input to the inverter and its output will be a signal with a log.1 level. This signal is fed to the input F ₃ , which generates a short pulse at its output, which is fed to the S-input of the second RS flip-flop T ₂ and causes it to work. At the output (direct) T ₂ appears the level log.1, which is fed to the second input of the first circuit And ₁ . This circuit opens and the signal from the direct output of the first RS-flip-flop T ₁ , after its operation and the appearance of the log.1 level, is fed to the first output of the HOODUN 12, indicating the presence of the beginning of the non-stationarity section with increasing steepness (attack). At the same time, the inverse output of the second RS-flip-flop T ₂ appears log level. 0, which enters the second inputs of the second and third circuits AND (AND ₂ , AND ₃ ), closes these circuits, and the log levels. 0 from their outputs are fed to the second and third outputs BODU 12. This state on these two outputs BODU 12 means the absence of a section of non-stationarity with a decline.

When parallel code combinations appear at the output of the SV, corresponding to the section of non-stationarity with a decline, signals with a level of log.1 will be received from the output of the SVZ to the input F ₄ . From the first such signal with a log.1 level, the shaper F ₄ generates a short pulse at its output, which is fed to the R-input of the second RS flip-flop T ₂ and causes it to work. The output (direct) T ₂ appears log level. 0, which is fed to the second input of the first circuit AND ₁ . This circuit closes and the signal from the direct output of the first RS-flip-flop T ₁ ceases to flow to the first output of the BODU 12. This state at the first output of the BODU 12 means the absence of a non-stationarity section with increasing steepness. Simultaneously, the inverse output of the second RS-flip-flop T ₂ appears log.1 level, which is fed to the second inputs of the second and third circuits And (And ₂ , And ₃ ). When the first input of the second circuit AND ₂ signals from the direct output of the first RS-flip-flop T ₁ (after its operation and the appearance of the log.1 level), the log.1 level comes from the output of AND ₂ to the second output of the HOODUN 12. This means detection the beginning of the non-stationarity section with a decline. At this time, the output of the third circuit AND ₃ and the third output HOLD 12 has a log level. 0. As the increment of the amplitude of the Hilbert envelope decreases on the decline (output to a section close to zero), the first RS-trigger T ₁ fires and a log level appears at its direct output. 0, which is supplied to the first inputs of the first and second circuits AND (AND ₁ , AND ₂ ). In this case, the first circuit AND ₁ continues to remain closed, and the second circuit AND ₂ closes and a log level appears at its output and the second output of the HOODUN 12. 0. At the same time, at the inverse output of the first RS-flip-flop T ₁ , a log.1 level appears, which is fed to the first input of the third circuit I3, as a result of which this level passes through this circuit and appears at the third output of the BODUN 12. This means the detection of the end of the non-stationarity section with a decline. Further, the work of HOODUN 12 occurs in a similar way.

The sign detection circuit (SVZ) included in the HOLD 12 can be implemented, for example, on the basis of an OR circuit with two inputs, one of which is set to the log level. 0, and the second input receives the most significant bit of each parallel code combination from the output of the CB. SVZ can also be implemented on the basis of two series-connected NOT circuits.

An example of the implementation of the Rhythmic Frequency Determination Unit (BFR) 13 is shown in FIG. 5. Borch 13 consists of the shaper F, the first counter (SC ₁ ), the second counter (SC ₂ ), the delay element (ED) and the division-by-ten circuit (DDD). The input BORC 13 is connected inside the block with a shaper of short pulses F, the output of which is connected to the first input of the MF ₁ , the output (code) of which is connected to the input of the SDD, the output (code) of which is the first output of the BORC 13. The output of the MF ₂ is connected to the input of the EZ, the output of which is connected to the second input of the MF ₁ and the second output of the BORC 13.

Work BORC 13 (figure 5) is as follows. In the initial state, MF ₁ and SC ₂ are reset to zero. The input BORCH 13 receives pulses from the first output BODU 12 (figure 1), corresponding to areas of non-stationarity with increasing steepness (attack). Inside BORC 13 (figure 5), these pulses are fed to input F, in which short pulses are formed, which are then fed to the input of MF _1. In MF ₁ , the number of these short pulses is counted, corresponding to the number of attacks per unit time, which determines the rhythmic structure speech or musical signals in the form of their rhythmic frequencies. The counted number of pulses of rhythmic frequencies appear at the code output MF ₁ in the form of parallel code combinations.

It is known that rhythmic frequencies depending on the type of information signals (speech, music, noise and other signals) can vary from 0.1 Hz to 16 Hz. 0.1 to 16 pulses per second. To detect pulses in a given frequency range, it is proposed to measure them in a time range (time unit) of 10 seconds. Then in this time range of 10 sec. can detect from 1 rhythm pulse to 160 rhythm pulses. It would seem that a time range of 1 second is more convenient, because allows you to count the rhythmic frequencies in this time period directly in Hertz. However, difficulties arise when measuring the number of rhythmic impulses whose frequencies are below 1 Hz. In this case, the infra-low frequency pulses may not fall within the 1 second time range. The formation of the time range of 10 seconds is carried out using the midrange ₂ . For this, clock pulses with a frequency of 48 kHz from the LADC 2 are fed to the input of the midrange ₂ (this circuit is not shown in figure 1). To form a time range of 10 seconds (with an accuracy of 0.017 ms), it is required to count 480076 clock pulses. After the last 480076 clock pulse arrives at the input of the MF ₂ , a short pulse appears at the output of the MF ₂ , which arrives at the input of the EZ, and the MF ₂ itself is reset to zero and begins a new cycle of counting 480076 clock pulses. Under the action of a short pulse received from the output of the EZ to the second input of the MF ₁ , this counter returns to its original state and begins a new cycle of counting the number of pulses of rhythmic frequencies in a time interval of 10 seconds. EZ is necessary to supply a short pulse from the output of the MF ₂ to the second input of the MF ₁ with a small delay, so that the parallel code combination from the code output of the MF ₁ has time to arrive at the code input of the SDD. Thus, with the help of MF ₁ , the number of pulses of rhythmic frequencies was counted in a time interval (segment) of 10 seconds. In order to convert this counted number of pulses into a unit of measurement expressed in Hertz, it is necessary to divide this numerical value of the counted pulses by 10. To do this, each parallel code combination, measured over a time interval of 10 seconds, is fed from the code output MF ₁ to the code SDD input. In DDS, each such parallel code combination is divided by ten, as a result of which the numerical values of such code combinations decrease by ten times and begin to correspond to the measurement of the number of pulses of rhythmic frequencies in a time interval of 1 second, and therefore expressed in Hertz. For example, a code combination at the SDD input corresponding to 5 pulses in a time interval of 10 seconds will be converted into a code combination corresponding to 0.5 pulses in a time interval of 1 second, which corresponds to a rhythmic frequency of 0.5 Hz. Divided by 10 code combinations from the code output of the SDD are fed to the first (code) output of the BORC 13. In addition, short pulses corresponding to time intervals of 10 seconds are sent from the output of the EZ to the second output of the BORC 13.

An example of implementation of the block for determining the duration of recessions (SBDS) 14 is shown in FIG. 6. BODS 14 consists of a key (KL), a counter (MF), a multiplication circuit (CS), a memory circuit (SP), a short pulse shaper (F) and a delay element (ED). The first input of BODS 14 is connected inside the block with the first input of the CL, the second input of which is supplied with clock pulses with a frequency of 48 kHz from the LADC 2 (this circuit is not shown in Fig.1), and the output of the CL is connected to the first input of the MF, the output (coded) which is connected to the first input of the CS, the second input (code) of which is connected to the output of the SP, and the output of the CS is connected to the first (code) output of the BODS 14, and the second input of the BODS 14 is connected to the input F, the output of which is connected to the input of the EZ, the output of which connected to the second input of the midrange and the second output of the BODS 14, while the output of the midrange is connected to the third (code) output of the BODS 14.

Work BODS 14 (Fig. 6) is as follows. In the initial state, the MF is reset to zero. The first input BODS 14 receives pulses from the second output BODU 12 (figure 1), corresponding to areas of non-stationarity with a decline. Inside HODUN 12 (Fig.6) these pulses are fed to the first input of the CL, the second input of which receives clock pulses with a frequency of 48 kHz from the LADC 2 (this circuit is not shown in Fig.1). Upon receipt of the first input CL (figure 6) level log.1, corresponding to the appearance of a section of non-stationarity with a decline, this key opens and its output begins to pass clock pulses with a frequency of 48 kHz from the second input CL. These clock pulses from the output of the CL are fed to the first input of the MF, which starts counting the number of these pulses, and parallel code combinations appear on the code output of the MF, corresponding to the number of these pulses N _x . The counting of the number of pulses in the MF continues during the duration of the non-stationarity segment with a decline. At the end of this section of non-stationarity with a decline, the first input BODS 14 appears log level. 0 (from the second output HOPE 12, figure 1), under the action of which closes CL, and the second input BODS 14 appears a log.1 pulse (from the third output HOLD 12 figure 1). Inside BODS 14 (figure 6) this log.1 pulse is fed to the input F, which generates a short pulse at its output. This short pulse passes through the EZ and enters the second input of the MF, as a result of which this counter returns to its original state and is ready for a new cycle of counting the number of clock pulses that fit into the duration of the next phase of non-stationarity with a decline. EZ is necessary to supply a short pulse from the output F to the second input of the MF with a small delay, so that the parallel code combination from the code output of the MF has time to arrive at the first code input of the CS. Thus, using the MF, it turned out to count the number of clock frequency pulses that fit in the time interval of the non-stationarity section with a decline N _x . In order to convert this counted number of pulses into a unit of the duration of the decay section, expressed in milliseconds, it is necessary to multiply this numerical value of the counted pulses N _x by 0.02083 (the duration of one cycle of the clock frequency pulses, in milliseconds). This number 0.02083 is written in the form of a parallel code combination in the BP and comes from its code output to the second code input of the control system. After multiplying in the control system the code combination corresponding to the number of clock frequency pulses in the time interval of the non-stationary section with a decline N _x by the code pattern corresponding to the duration of one period of clock frequency pulses, we obtain at the code output of the control system a code pattern corresponding to the duration of the non-stationary section with a decline, expressed in milliseconds. This code combination from the code output of the control system enters the first (code) output of the BODS 14, and the second output of the BODS 14 receives short pulses from the output of the EZ, while the third (code) output of the BODS 14 receives parallel code combinations from the output of the MF, corresponding to the number impulses N _x in the section with a decline. Further work BODS 14 occurs in a similar way.

An example of the implementation of the adder-averaging 7 is shown in Fig.7. The adder-averaging 7 consists of a series-connected adder and circuit division by N _x . The first (code) input of the adder-averaging 7 is connected to the code input of the adder, the output of which is connected to the first (code) input of the division by N _x circuit, the second (code) input of which is connected to the second (code) input of the adder-averaging 7, and the code the output of the division circuit by N _x is connected to the code output of the adder-averaging 7.

The operation of the adder-averaging 7 and the formation of digital signal values corresponding to the values of the instantaneous power of the measured signal at the duration of the sections of non-stationarity with a decline, is carried out in accordance with the expression:

where n _i ² is the numerical value of the i codeword (corresponding to the i sample of the amplitude envelope of the original analog signal).

The code input of the adder from the first code input of the adder-averaging 7 (Fig.7) alternately receives N _x parallel code combinations included in the first section of non-stationarity with a decline (corresponding to N _x samples of the amplitude envelope of the original analog signal). In the adder, these N _x parallel code combinations included in the first section of non-stationarity with a decline are added and a parallel code combination (digital count) appears at its output, corresponding to the total value of these code combinations. This code combination is then fed to the first code input of the division by N _x , the second code input of which receives code combinations from the second code input of the adder-averaging 7 (from the third output of the BODS 14 in Fig.1). In this division scheme (Fig.7) is the operation of dividing by the number N _x equal to the number of parallel code combinations (counts) that make up the section of non-stationarity with a decline. At the code output of the division by N _x circuit, a parallel code pattern appears corresponding to the value of the instantaneous power of the analog signal envelope for the duration of the first section of non-stationarity. This code combination from the output of the division circuit by N _x is then fed to the code output of the adder-averaging 7.

After that, the code input of the adder from the first code input of the adder-averaging 7 alternately receives new N _x parallel code combinations included in the second section of non-stationarity with a decline. These N _x parallel codewords are added in the adder and divided by the number N _x in the division by N _x circuit. As a result, a parallel code pattern appears at the code output of the division by N _x circuit, corresponding to the instantaneous power value of the envelope of the measured analog signal in the second section of non-stationarity with a decline. Further work of the adder-averaging 7 occurs in a similar way.

An example of the implementation of the first, second and third memory blocks 9, 10, 15 is shown in Fig.8. The first 9, the second 10 and the third 15 memory blocks consist of the first and second buffer memories, a K pulse counter, an adder and a divide-by-K circuit. The first (code) input of the first buffer memory is connected to the first (code) input of the first, second and third memory blocks 9, 10, 15, and the code output of the first buffer memory is connected to the first (code) input of the second buffer memory. The second input of the first, second and third memory block 9, 10, 15 is connected to the second input of the first buffer memory and to the input of the pulse counter K, the output of which is connected to the third input of the first buffer memory and to the second input of the second buffer memory. The code output of the second buffer memory is connected to the first code output of the first, second and third memory blocks 9, 10, 15 and to the code input of the adder, the code output of which is connected to the code input of the division by K circuit, the code output of which is connected to the second code output of the first, the second and third memory blocks 9, 10, 15.

The operation of the first 9, second 10 and third 15 memory blocks is as follows. In the initial state, the first and second buffer memories, as well as the counter K of pulses, are reset to zero. At 1 code input of the first buffer memory from 1 input of the first, second and third memory blocks 9, 10, 15, parallel code combinations are received (from the output of the adder-averaging 8, respectively, from the first output of the BODS 14 and the first output of the BORC 13, Fig.1 ). At the same time, short pulses are received at the 2nd input of the first buffer memory and the counter input K of pulses from the 2nd input of the first and second memory blocks 9,10 (Fig.8), corresponding to the ends of the non-stationarity sections with a decline (from the second output of the BODS 14, Fig.1). As for the third memory block 15, its second input receives short pulses corresponding to time intervals of 10 seconds (from the second output BORC 13, Fig.1). Under the action of these short pulses, parallel code combinations from the first input are written to the first buffer memory (Fig. 8) and appear at its code output, but are not written to the second buffer memory.

At the same time, the pulse counter K starts counting short pulses corresponding to the ends of the non-stationarity sections with a decline (first and second memory blocks 9, 10), as well as counting short pulses corresponding to time intervals of 10 seconds (third memory block 15) and after K pulse at the output of this counter appears the first short pulse. Under the action of the leading edge of this pulse, K parallel code combinations that were recorded and were present at the code output of the first buffer memory are written to the second buffer memory and appear at its code output (1 output of the first, second and third memory blocks 9, 10, 15) .

After that, under the influence of the decay of the pulse from the output of the counter, the first buffer memory of the first, second and third memory blocks 9,10,15 is reset to zero and is ready to form the second set of K parallel code combinations (from the output of the adder-averaging 8, respectively, from the first output BODS 14 and the first output BORC 13, figure 1).

Thus, it turned out to be formed (stored) in the first memory block 9 the first set of K numerical values of digital samples corresponding to the instantaneous power values in K sections of non-stationarity with a decline, and also formed (stored) in the second memory block 10 the first set of K numerical values of digital samples corresponding to the duration of the non-stationary sections with a decline, expressed in milliseconds, and finally formed (stored) in the third memory block 15 the first set of K numerical values of digital samples corresponding to the number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals in 1 sec. This set is present at the code output of the second buffer memory (1 output of the first, second and third memory blocks 9, 10, 15) until the second short pulse appears at the output of the counter. Further, the work of the first and second buffer memories occurs in a similar way.

The operation of the adder in conjunction with the division circuit by K for the formation in the first memory block 9 of a digital readout corresponding to the value of the average power of the sections of non-stationarity with a decline over a long time interval, consisting of K such sections, as well as for the formation in the second memory block 10 of a digital readout corresponding to the value of the average duration of sections of non-stationarity with a decline, over a long time interval consisting of K such sections, and finally for the formation in the third memory block 15 of a digital readout corresponding to the value of the average number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 sec, on a long time interval, consisting of K such sections, is carried out in accordance with the expression

where m _i is the numerical value of the i parallel code combination (corresponding to the i digital sample of a set of K numerical values of digital samples corresponding to the instantaneous power values in K sections of non-stationarity or corresponding to the i digital sample of a set of K numerical values of digital samples corresponding to the values of the duration of sections of non-stationarity with decline in K sections, or a set of K numerical values of digital samples corresponding to the i digital count, corresponding to the values of the number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 second at K sections).

In the adder K of parallel code combinations in the form of the first set of numerical values of digital samples corresponding to the instantaneous power values in K sections of non-stationarity with a decrease, or numerical values of digital samples corresponding to the duration values in K sections of non-stationarity with a decrease, or numerical values of digital samples corresponding to the values of the number of impulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 second at K sections are added up and a code combination appears at its output, corresponding to the total value of these code combinations. This code combination then goes to the division by K scheme. In this scheme, the division operation is carried out by the number K, which is equal to the number of code combinations that make up the set of sections. At the code output of the division by K circuit, i.e. at the second output of the first memory block 9, a parallel code combination (digital reading) appears corresponding to the value of the average power of the non-stationarity sections with a decline over a long time interval consisting of K such sections, and at the second output of the second memory block 10, a parallel code combination (digital reading ), corresponding to the value of the average duration of sections of non-stationarity with a decline, on a long time interval, consisting of K such sections, and at the second output of the third memory block 15, a parallel code combination (digital readout) appears, corresponding to the value of the average number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 second, over a long time interval consisting of K such sections.

After that, under the action of the second short pulse from the output of the counter, the code input of the adder from the code output of the second buffer memory receives parallel code combinations in the form of a second set of K numerical values of digital samples corresponding to the values of the instantaneous power in the K sections of non-stationarity with a decline, or numerical values of digital samples corresponding to the values of durations in K sections of non-stationarity with a decline or numerical values of digital samples corresponding to the values of the number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 second in K sections. These parallel codewords appear at the 1st output of the first, second and third memory blocks 9, 10, 15.

Data K of parallel code combinations from the code output of the second buffer memory are added in an adder and divided by K in the division by K circuit. As a result, a parallel code combination (digital readout) corresponding to the value of the average power of sections of non-stationarity with a decline in a long time interval, consisting of K such sections, and at the 2nd output of the second memory block 10, a parallel code combination (digital readout) appears, corresponding to the value of the average duration of sections of non-stationarity with a decline in a long time interval, consisting of To such sections, and at the 2nd output of the third memory block 15, a parallel code combination (digital reading) appears, corresponding to the value of the average number of pulses of rhythmic frequencies at time intervals of 10 seconds and recalculated into time intervals of 1 second over a long time interval, consisting of K such areas.

Further, the operation of the adder together with the division circuit by K, included in the first, second and third memory blocks 9, 10, 15, occurs in a similar way.

An example of the implementation of the scheme of segmentation and overlay window function Nuttall (SSNOFN), included in the block Hilbert orthogonal transform (HOP) 3 is shown in Fig.9. This circuit includes first and second buffer memories, a multiplication circuit, a counter, and a memory circuit. The first (code) input of SSNOFN is connected to the first (code) input of the first buffer memory, the code output of which is connected through the second buffer memory to the code input of the multiplication circuit, the second (code) input of which is connected to the code output of the memory circuit, and the output is connected to the code output SSNOFN. The second input of SSNOFN is connected to the second input of the first buffer memory and to the input of the counter, the output of which is connected to the third input of the first buffer memory, to the second input of the second buffer memory and to the first input of the memory circuit. The third input of SSNOFN is connected to the third input of the second buffer memory and to the second input of the memory circuit.

Segmentation and overlay window function scheme Nuttall (Fig.9) works as follows. In the initial state, the first and second buffer memories and the counter are reset to zero. The memory circuit is also in its initial state when a codeword corresponding to the Nutall window transmission coefficient for the first of the B codewords (discrete samples) of the digital signal in the segment is present at its code output.

The first (code) input SSNOFN from the first (code) input BGOP 3 receives parallel code combinations that are fed to the first (code) input of the first buffer memory (Fig.9). Simultaneously to the second input SSNOFN from the second input BGOP 3 receive sampling frequency pulses, which are fed to the input of the counter and the second input of the first buffer memory (Fig.9). To the third input of the SSNOFN from the output of the sampling pulse frequency doubling circuit, which is part of the BGOP 3, pulses with a double sampling frequency are received, which are fed to the third input of the second buffer memory and the second input of the memory circuit (Fig.9). In this case, the counter in the SSNOFN is designed to count the number of code combinations equal to half the duration of the segment (half-segment), which is then superimposed with the Nutoll window function. For example, a sequence of half-segments should be formed from a digital signal with a sampling frequency of 48 kHz, each of which should contain B/2=480 discrete samples (code combinations). In this case, each discrete sample is, for example, a 16-bit code combination. Then the duration of each half-segment will fit 480 sixteen-bit code combinations. It is after a given number of sampling rate pulses that a short pulse appears at the output of the counter, indicating the end of this half-segment and the beginning of the next one (Fig. 10 a, b). Pulses from the output of the counter are fed to the third input of the first buffer memory, to the second input of the second buffer memory and to the first input of the memory circuit.

The first buffer memory in SSNOFN contains B/2=480 code combinations (half-segment), and the second buffer memory consists of two halves and contains B=960 code combinations (two half-segments of 480 code combinations).

As parallel code combinations arrive at 1 code input of the first buffer memory, they are written into it under the action of pulses with a sampling frequency. These code combinations appear at the code output of the first buffer memory and are applied to the code input of the second buffer memory, but are not written to it.

At the same time, B=960 zero code combinations are read from the second buffer memory under the action of pulses with a double sampling frequency. These zero 16-bit code combinations are sequentially fed to the first code input of the multiplication circuit. At this time, the second code input of this circuit is fed with 16-bit codewords corresponding to the Nutall window transmission coefficients. After multiplying the code combinations applied to the 1 and 2 code inputs of the multiplication circuit, its output will also be zero 16-bit code combinations.

Thus, during the period of filling the first buffer memory with code combinations corresponding to the first half-segment (1 ps in Fig. 10a), the output of the multiplication circuit is the formation of the first segment in a row (0 ₁ -0 ₀ segments in Fig. 10 c) from zero code combinations.

After filling 480 with sixteen-bit code combinations of the first buffer memory, the first short pulse appears at the output of the counter (Fig. 10 b) under the action of the leading edge of which these code combinations from the first buffer memory are written to the first half of the second buffer memory (1 p. s. in Fig. 10a). Under the action of the same short pulse, 480 zero code combinations from the first half of the second buffer memory are shifted and written to the second half of this buffer memory (0 ps in Fig. 10a). Thus, the first segment is formed from the zero and first half-segments (1 segment in Fig. 10a).

Under the action of the decay of the same short pulse, the first buffer memory and the memory circuit are set to their initial state. At the same time, a codeword appears at the code output of the memory circuit, corresponding to the Nutall window transmission coefficient for the first of B=960 codewords in the first segment (1 segment in Fig. 10a). It should be noted that the Nutall window gains (and their corresponding codewords) for the first half of the segment (e.g. 0 p.s. in 1 segment in Fig. 10a) are increasing, and for the second half of the segment (e.g. 1 p.s. in 1 segment in Fig. Yua) are decreasing.

Parallel code combinations that continue to arrive at 1 code input of the first buffer memory are recorded in this memory under the action of pulses with a sampling frequency. At the same time, under the action of pulses with a double sampling rate at the third input of the second buffer memory and the second input of the memory circuit, 16-bit code combinations from their code outputs are fed to the first and second code inputs of the multiplication circuit, respectively. Zero code combinations (from the second half of the second buffer memory) of the zero half-segment of the first segment (1 segment in Fig. 10a) are multiplied first, so only zero 16-bit code combinations appear at the code output of the multiplication circuit.

Further, information code combinations (from the first half of the second buffer memory) of the first half-segment of the first segment (1 segment in Fig. 10a) begin to be multiplied, therefore, multiplied 16-bit code combinations appear at the code output of the multiplication circuit, corresponding to the original code combinations, but with superimposed on them by the Nutoll window transmission coefficients.

That. during the period of filling the first buffer memory with code combinations corresponding to the second half-segment in a row (1 p.s. in Fig. 10a), the second segment is formed at the output of the multiplication circuit (1 ₁ -0 ₂ segments in Fig. 10c), consisting of a second use of the zero half-segment and a first use of the first half-segment (in which the Nutall window gains are decreasing).

After filling the next 480 code combinations of the first buffer memory, the second short pulse appears at the output of the counter (Fig. 10b) under the action of the leading edge of which these code combinations are written to the first half of the second buffer memory. Under the action of the same short pulse, 480 previously recorded code combinations from the first half of the second buffer memory are shifted and written to the second half of this buffer memory. Thus, the second segment is formed from the first and second half-segments (2 segments in Fig. 10a).

Under the action of the decay of the same short pulse, the first buffer memory and the memory circuit are set to their initial state. At the same time, a codeword appears at the code output of the memory circuit, corresponding to the Nutall window transmission coefficient for the first of B=960 codewords in the second segment (2 segments in Fig. 10a).

Under the action of pulses at the third input of the second buffer memory and the second input of the memory circuit, 16-bit code combinations from their code outputs are fed to the first and second code inputs of the multiplication circuit, respectively. The codewords (from the second half of the second buffer memory) of the first half-segment of the second segment (2 segments in Fig. 10a) are multiplied first. These multiplied codewords appear at the output of the multiplication circuit. Next, codewords (from the first half of the second buffer memory) of the second half-segment of the second segment (2 segments in Fig. 10a) begin to be multiplied. These multiplied codewords also appear at the output of the multiplication circuit.

That. at the output of the multiplication circuit, the third segment is formed (2 ₁ -1 ₂ segments in Fig. 10c), consisting of the second time used first half-segment (in which the Nutall window gains are increasing) and the first time used second half-segment (in which the Nutoll window gains are decreasing).

While 16 bit patterns are being read from the second buffer memory, the patterns corresponding to the third half-segment (3 p.s. in FIG. 10a) are written to the first buffer memory.

After filling the next 480 code combinations of the first buffer memory, a third short pulse appears at the output of the counter (Fig. 10b), under the action of the leading edge of which these code combinations are written to the first half of the second buffer memory. Under the action of the same short pulse, 480 previously recorded code combinations from the first half of the second buffer memory are shifted and written to the second half of this buffer memory. Thus, the third segment is formed from the second and third half-segments (3 segments in Fig. 10a).

Under the action of the decay of the same short pulse, the first buffer memory and the memory circuit are set to their initial state. At the same time, a codeword appears at the code output of the memory circuit corresponding to the Nutall window transmission coefficient for the first of B=960 codewords in the third segment (3 segments in Fig. 10a).

Under the action of pulses at the third input of the second buffer memory and the second input of the memory circuit, 16-bit code combinations from their code outputs are fed to the first and second code inputs of the multiplication circuit, respectively. The codewords (from the second half of the second buffer memory) of the second half-segment of the third segment (3 segments in Fig. 10a) are multiplied first. These multiplied codewords appear at the output of the multiplication circuit. Next, codewords (from the first half of the second buffer memory) of the third half-segment of the third segment (3 segments in Fig. 10a) begin to be multiplied. These multiplied codewords also appear at the output of the multiplication circuit.

That. at the output of the multiplication circuit, the fourth segment is formed (3 ₁ -2 ₂ segments in Fig. 10c), consisting of the second half-segment used for the second time (in which the Nuttall window gains are increasing) and the third half-segment used for the first time (in which the Nutoll window gains are decreasing).

Further work of SSNOFN proceeds in a similar way.

An example of the implementation of the scheme for overlapping segments and compensating for the unevenness of the Nuttall window function (SPSKNON) included in the block of the Hilbert orthogonal transform (HOP) 3 is shown in Fig.11. This circuit contains: the first, second, third and fourth buffer memories (BP), an adder, a memory circuit (SP), a multiplication circuit (MC), a counter, a trigger, a shaper, a delay element (DE). The first (code) input of the first buffer memory (BP ₁ ) is connected to the first (code) input of SPSKNON, and its code output is connected to the first (code) input of the second buffer memory (BP ₂ ) and to the first (code) input of the third buffer memory ( BP ₃ ). The second input of BP ₁ is connected to the output of the delay element EZ, and the third input of BP ₁ is connected to the second input of SPSKNON, to which the input of the counter is also connected, the output of which is connected to the trigger input, the input of the EZ and to the second input of BP ₂ , the code output of which is connected to the first (code) input BP ₄ . The third input SPSKNON connected to the second input of the memory circuit (SP), the second input of the BP ₃ and the second input of the BP ₄ . The trigger output is connected to the input of the shaper, the output of which is connected to the first input of the joint venture, to the third input of the PSU ₃ and to the third input of the PSU ₄ . The code outputs of BP ₃ and BP ₄ are connected, respectively, to the first and second code inputs of the adder, the code output of which is connected to the first code input of the multiplication circuit (CS), the second code input of which is connected to the code output of the SP, and the code output of the CS is connected to the output SPSKNON.

SPSKNON (Fig.11) works as follows. In the initial state, BP ₁ , BP ₂ , BP ₃ , BP ₄ , the counter, as well as the trigger are reset to zero. The SP is also in the initial state when its code output has a codeword corresponding to the gain to compensate for the Nuttall window function ripple for the first of the B codewords in the first segment.

On the first (code) input SPSKNON (Fig.11) and then on the first (code) input of the PSU ₁ receive parallel code combinations. At the same time, pulses with doubled sampling frequency are fed to the second input of the SPSKNON, which are then fed to the third input of the BP ₁ . Under the action of these pulses, the code combinations received at the input of the BP ₁ are written to it and appear at the code output of the BP ₁ . These code combinations are applied to the first (code) inputs of BP ₂ and BP ₃ but are not recorded in them.

At the same time, the counter starts counting pulses at double the sampling rate. This counter is designed to count the number of code combinations equal to half the duration of a segment (half-segment). For example, a sequence of half-segments must be formed from a digital signal having a double sampling frequency, each of which must contain B/2=480 discrete samples (code combinations). In this case, each discrete sample is a 16-bit code combination. Then the duration of each half-segment will fit 480 sixteen-bit code combinations. It is after a given number of pulses with a double sampling frequency that a short pulse appears at the output of the counter, indicating the end of this half-segment and the beginning of the next one (Fig. 12a, b).

BP ₁ , BP ₂ , BP ₃ , BP ₄ in our example, each contain 480 sixteen-bit code combinations (i.e., each in a half-segment), Code combinations from the code outputs of the adder, SU and SP are also 16 bit.

SPSKNON is designed to form segments of a digital signal from B code combinations in each segment and add with 50% overlap of each segment with its previous segment. In order to avoid gaps in the sequence of the digital signal that is formed after the overlap of the segments, it is necessary that the recording of code combinations in BP ₁ be performed at a double sampling rate, and the reading of code combinations from BP ₃ and BP ₄ is performed at a sampling rate.

Simultaneously with the recording of code combinations in BP ₁ from BP ₃ and BP ₄ reads zero code combinations under the action of pulses at their second inputs. These zero 16-bit code combinations are fed to the first and second code inputs of the adder, the output of which will also be zero 16-bit code combinations, which are fed to the first code input of the control system. At this time, 16-bit code combinations are supplied to the second code input of this circuit from the code output of the SP, corresponding to the transmission coefficients to compensate for the unevenness of the Nuttall window function. After multiplying the code combinations applied to the 1st and 2nd code inputs of the control system, its code output will also have zero 16-bit code combinations

That. in the period of filling BP ₁ with code combinations corresponding to the first half-segment (0 ₀ ps in Fig.12a), a half-segment (0 _n in Fig.12d) is formed at the code output of the control system from zero code combinations.

After filling 480 with sixteen-bit zero code combinations BP ₁ corresponding to 0 ₀ - half-segment (Fig.12a), the first short pulse appears at the output of the counter (Fig.126) from which the trigger is triggered, and a short pulse also appears at the output of the shaper. Under the action of the leading edge of the pulse from the output of the shaper, zero code combinations corresponding to the 0 ₀ half-segment are recorded from the output of BP ₁ in BP ₃ , and in BP ₄ , zero code combinations that were present in BP ₂ are also recorded. Thus, the first segment is formed from 0 and 0 ₀ half-segments (Fig. 12a) (1 segment in Fig. 12a - below). At the same time, under the action of the same short pulse from the output of the shaper, the SP is reset to its initial state, when a code combination appears at its code output, corresponding to the transmission coefficient to compensate for the unevenness of the Nuttall window function for the first code combination in the segment.

After that, under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 0 ₀ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 0 ₁ half-segment (Fig.12a).

Under the action of pulses on the second inputs of BP ₃ and BP ₄ , 16 bit zero code combinations from their code outputs are fed to, respectively, the first and second code inputs of the adder. Next, zero code combinations from the code output of the adder (0 ₀ ps + 0 ps in Fig.12a) are fed to the first code input of the control system, the second code input of which receives code combinations from the output of the SP. That. at the output of the control system, the formation of the first segment is carried out (0 ₀ +0 segment in Fig. 12 d).

While BP3 and BP ₄ are slowed down by 2 times (compared to the write speed in BP ₁ ) reading 16-bit code combinations, in BP ₁ code combinations corresponding to 0 ₁ half-segment are recorded.

After filling 480 with zero code combinations of BP ₁ , a second short pulse appears at the output of the counter (Fig. 12b) under the action of which the trigger is triggered and a “logical 0” (“log. 0”) appears at its output, from which no signal, and hence writing in BP ₃ and BP ₄ parallel code combinations from BP ₁ and BP ₂ does not occur. At this time, reading, addition and multiplication of zero code combinations corresponding to 0 ₀ and 0 half-segments from BP ₃ and BP ₄ continues and a 0 ₀ - 0 segment is formed (Fig. 12 d).

Under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 0 ₁ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 0 ₂ half-segment (Fig.12a).

After filling with zero code combinations of BP ₁ (0 ₂ ps in Fig. 12a), a third short pulse appears at the output of the counter (Fig. 12b) under the action of which the trigger is triggered and “logic 1” appears at its output (“log. 1 ”), from which a second short pulse appears at the output of the shaper (Fig. 12c). Under the action of the leading edge of the pulse from the output of the shaper, zero code combinations corresponding to the 0 ₂ half-segment are recorded from the output of BP ₁ in BP ₃ , and in BP ₄ , zero code combinations are also recorded corresponding to 0 ₁ and which were present in BP ₂ . Thus, the second segment is formed from 0 ₂ and 0 ₁ half-segments (2 segments in Fig. 12a - at the bottom). At the same time, under the action of the same short pulse from the output of the shaper, the SP is reset to its initial state, when a code combination appears at its code output, corresponding to the transmission coefficient to compensate for the unevenness of the Nuttall window function for the first code combination in the segment.

After that, under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 0 ₂ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 1 ₁ half-segment (Fig. 12a).

Under the action of pulses on the second inputs of BP ₃ and BP ₄ , 16 bit zero code combinations from their code outputs are fed to, respectively, the first and second code inputs of the adder. Further, zero code combinations from the code output of the adder (0 ₂ ps + 0 ₁ ps in Fig.12a) are fed to the first code input of the control system, the second code input of which receives code combinations from the output of the SP. Zero 16-bit code combinations appear on the code output of the CU. That. at the output of the control unit, the formation of the second segment is carried out (0 ₂ +0 ₁ segment in Fig. 12 d).

While 16-bit code combinations are read from BP ₃ and BP ₄ (0 ₂ ps and 0 ₁ ps in Fig. 12a), code combinations corresponding to _{1 1} half-segment (1 ₁ ps in Fig. 12a).

After filling with code combinations (1 ₁ p.s in Fig. 12a), a fourth short pulse appears at the output of the counter (Fig. 12b), under the action of which the trigger is activated and a “log. 0”, from which no signal arises at the output of the shaper, which means that no code combinations from BP ₁ and BP ₂ are written to BP ₃ and BP ₄ . At this time, reading, addition and multiplication of zero code combinations corresponding to 0 ₂ and 0 ₁ half-segments from BP ₃ and BP ₄ continues and a 0 ₂ -0 ₁ segment is formed (Fig. 12 d).

Under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 1 ₁ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 1 ₂ half-segment (Fig.12a).

After filling with code combinations of BP ₁ (1 ₂ ps in Fig. 12a), a fifth short pulse appears at the output of the counter (Fig. 12b), under the action of which the trigger is triggered and “log.1” appears at its output, from which the output shaper, a third short pulse appears (Fig. 12c). Under the action of this pulse, code combinations from the code outputs of BP ₁ and BP ₂ . are recorded, respectively, in BP ₃ and BP ₄ . Thus, the third segment is formed from 1 ₂ and 1 ₁ half-segments (3 segments in Fig. 12a - below). At the same time, under the action of the same short pulse, the memory block is reset to its initial state, when a code combination appears at its code output, corresponding to the transmission coefficient to compensate for the uneven Nuttall window function for the first code combination in the third segment.

After that, under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 1 ₂ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse, delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 2 ₁ half-segment (Fig.12a).

Under the action of pulses with a sampling frequency at the second inputs of BP ₃ and BP ₄ , 16-bit information code combinations from their code outputs are fed to, respectively, the first and second code inputs of the adder. The summation adds the codewords included in the L half-segment (in which the Nutall window gains are increasing) with the same codewords included in the 1 ₁ half-segment (in which the Nutall window gains are decreasing), so the output of the adder is the window gains Nutoll levels flatten out (become close to 1), although some unevenness remains.

Further, after summing, the code combinations from the code output of the adder (1 ₂ ps + 1 ₁ ps in Fig. 12a) are fed to the first code input of the control system, the second code input of which receives code combinations from the output of the SP. After multiplying the code combinations, the unevenness of the Nuttall window function is compensated. If we compare the (1 ₁ -0 ₂ ) segment and (2 ₁ -1 ₂ ) segment (top of Fig. 12a) at the input of SPSKN with 3 segments (3 segments in Fig. 12a or 1 ₂ +1 ₁ segments in Fig. 12d) at the output of the adder, it can be seen that there is an addition with a 50% overlap of the segment with the previous segment.

The code output of the control unit receives 16-bit code combinations with compensated unevenness of the Nuttall window function. That. at the output of the control unit, the formation of the third segment is carried out (1 ₂ +1 ₁ segment in Fig. 12 d).

While 16-bit code combinations are read from BP ₃ and BP ₄ (1 ₂ ps and 1 ₁ ps in Fig. 12a), code combinations corresponding to 2 ₁ half-segments (2 ₁ ps in Fig. 12a) are recorded in the BP. in Fig. 12a).

After filling with code combinations (2 ₁ p.s in Fig.12a) BP ₁ at the output of the counter appears the sixth short pulse (Fig. 12b) under the action of which the trigger is triggered and at its output appears "log. 0”, from which no signal arises at the output of the shaper, and therefore there is no recording in BP ₃ and BP _{4 of} parallel code combinations from the code outputs of BP ₁ and BP ₂ .

At this time, reading, addition and multiplication of code combinations corresponding to 1 ₂ and 1 ₁ half-segments from BP ₃ and BP ₄ continues and 1 ₂ -1 ₁ segment is formed (Fig. 12 d).

Under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 2 ₁ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 2 ₂ half-segment (Fig.12a).

After filling the BP ₁ with code combinations corresponding to the 2 ₂ -semi-segment (2 ₂ ps in Fig. 12a), the seventh short pulse appears at the output of the counter (Fig. 12b), under the action of which the trigger is triggered and a "log" appears at its output. 1", from which the fourth short pulse appears at the output of the shaper (Fig. 12c).

Under the action of this pulse, code combinations from the code outputs of BP ₁ and BP ₂ are recorded in BP ₃ and BP ₄ . Thus, the fourth segment is formed from 2 ₂ and 2 ₁ half-segments (4 segments in Fig. 12a below). At the same time, under the action of the same short pulse, the memory block is reset to its initial state, when a code combination appears at its code output, corresponding to the transmission coefficient to compensate for the Nuttall window function unevenness for the first code combination in the fourth segment.

After that, under the action of the decay of the pulse from the output of the counter, the code combinations from the code output of the BP ₁ corresponding to the 2 ₂ half-segment are recorded in the BP ₂ and appear at its code output. In addition, under the action of a short pulse delayed in the EZ, BP ₁ is reset and starts recording code combinations corresponding to the next 3 ₁ half-segment (Fig.12a).

Under the action of pulses on the second inputs of BP ₃ and BP ₄ , 16-bit information code combinations from their code outputs are fed to, respectively, the first and second code inputs of the adder. When summed, the codewords included in the 2 ₂ half-segment (in which the Nutall window gains are increasing) are added to the same codewords included in the 2 ₁ half-segment (in which the Nutall window gains are decreasing), therefore, at the output of the adder, the gains the Nuttall windows flatten out (become close to 1), although some unevenness remains.

Further, after summing, the code combinations from the code output of the adder (2 ₂ ps + 2 ₁ ps in Fig. 12a) are fed to the first code input of the control system, the second code input of which receives code combinations from the output of the joint venture. After multiplying the code combinations, the unevenness of the Nuttall window function is compensated. If we compare the (2 ₁ -1 ₂ ) segment and (3 ₁ -2 ₂ ) segment (top of Fig. 12a) at the input of SPSKNON with 4 segments (4 segments in Fig. 12a at the bottom or 2 ₂ +2 ₁ segments in Fig. .12 d) at the output of the adder, it can be seen that there is an addition with a 50% overlap of the segment with the previous segment.

The code output of the control unit receives 16-bit code combinations with compensated unevenness of the Nuttall window function. That. at the output of the BU, the fourth segment is formed (2 ₂ +2 ₁ segment in Fig. 12 d).

Further work of BPSKNON proceeds in a similar way.

Due to this solution of the problem, the proposed method and device for measuring rhythmic frequencies, power and duration of the recessions of the sections of non-stationarity of acoustic signals, unlike the prototype, allows you to expand the functionality and use the proposed method and device for measuring such important parameters of acoustic signals as the values of rhythmic frequencies in a given segment time, and the nature of the change in the average values of rhythmic frequencies over a long time period. In addition, the proposed method and device make it possible to measure such important parameters as the values of the durations of the declines in each section of non-stationarity, and the nature of the change in the average values of the durations of the declines of the sections of non-stationarity over a long time interval. And also, the proposed method and device make it possible to measure such important parameters as individual values of instantaneous power in each section of non-stationarity with a decline, and the nature of the change in the values of the average power of sections of unsteadiness with a decline over a long time interval.

As a result of such measurements, it is possible to assess the quality of acoustic signals with great accuracy, since it is the areas of non-stationarity in the form of attacks and recessions that contain the greatest amount of information, and their distortion during transmission and processing significantly reduces the quality of these acoustic signals. These measurements will make it possible to carry out measures to reduce the distortion of non-stationary sections in acoustic signals and thereby improve their quality.

With the help of the proposed method and device for measuring rhythmic frequencies, power and duration of decays of sections of non-stationary acoustic signals, both audio broadcasting signals and speech and music signals, as well as any analog acoustic signals and noise, can be measured.

The proposed method and device for measuring rhythmic frequencies, power and duration of decays of non-stationary sections of acoustic signals can be used in existing analog and digital transmission channels, in analog signal processing systems, in director's systems when forming musical and speech programs, as well as for evaluating negative and positive effects of acoustic signals and noise on humans. For example, the number of attacks per unit of time determines the rhythmic structure of speech or musical signals in the form of their rhythmic frequencies. In turn, the structure of rhythmic frequencies largely determines the quality of transmitted acoustic information and the degree of its impact on listeners. Thus, the same phrase, uttered with different rhythmic frequencies, can evoke completely different emotional states in listeners. Recessions in non-stationary areas are also a very important parameter. Recessions in signals characterize the duration of the decay of sounds in musical instruments, as well as the duration of the decay of sounds (reverberation time) in different rooms. Reverberations give the sounds volume, juiciness, richness of the timbre composition, the voices of the singers become melodious. On the other hand, at long reverberation times (long decays), a strong echo occurs, making it difficult to perceive information. Therefore, the assessment of the duration and power of the decays of acoustic objects is of great importance.

The economic effect of using the proposed method and device for measuring the rhythmic frequencies, power and duration of the declines in the areas of non-stationarity of acoustic signals is supposed to be obtained by improving the accuracy of estimating the most important informational parameters of the non-stationarity of acoustic signals, which makes it possible to take measures to reduce distortions in these acoustic signals and themes. to improve the quality of these signals themselves. In addition, in cases where the parameters of the non-stationarity of acoustic signals and noise have a negative impact on a person, various protective measures can be carried out based on estimates of these parameters, which makes it possible to reduce losses in restoring the health and working capacity of people.

Claims (2)

1. A method for measuring rhythmic frequencies, power and duration of decays of acoustic signal non-stationarity sections, including input signal conversion, linear analog-to-digital signal conversion, Hilbert transformation with the formation of an orthogonal signal from a digital signal, digital extraction of a signal corresponding to the Hilbert amplitude envelope of the analog signal , selection by filtering the low-frequency components of the Hilbert amplitude envelope, as well as digital squaring, summation and averaging, the first storage with summation and averaging, the second storage with summation and averaging, a digital indication, characterized in that after the selection of the low-frequency components of the Hilbert amplitude envelope, they determine and extract the most important and informative sections of non-stationarity with increasing steepness and sections of decline, and in sections of decline, each of which contains N _x parallel code combinations, from which of which, after digital squaring, in each section of non-stationarity, by summing and averaging, a digital sample is formed, corresponding to the value of the instantaneous power of this section of non-stationarity with a decline, and then K of such digital samples are stored, and also digital samples are formed from K, by summation and averaging these readings, a digital reading corresponding to the value of the average power over a long time interval, consisting of K sections of non-stationarity with a decline, after which a digital indication is made of K stored digital readings corresponding to the values of the instantaneous power and a digital reading corresponding to the value of the average power of the sections of unsteadiness with a decrease by long time interval consisting of K sections, as well as on each selected section of non-stationarity with a decline, the duration of this decline Δt is determined in the form of a digital reading, and then K of such digital readings are stored, as well as they form from K digital readings, by summing and averaging these readings, a digital reading corresponding to the value of the average duration of recessions in a long time interval, consisting of K sections of non-stationarity with recessions, after which they carry out a digital indication of K stored digital readings corresponding to the values of the durations of the recessions of the sections of non-stationarity and a digital count corresponding to the average value of the duration of the decays of the non-stationarity sections over a long time interval, consisting of K sections, and, in addition, on the selected sections of non-stationarity with increasing steepness, the number of these sections (attacks) is determined in a given period of time in the form of a rhythmic count frequencies, and then K of such digital samples are stored, and also the formation of K digital samples is carried out, by summing and averaging these samples, a digital sample corresponding to the average value of rhythmic frequencies for a long time a specified segment consisting of K time segments with rhythmic frequencies, after which a digital indication is made of K stored digital readings corresponding to the values of rhythmic frequencies and a digital reading corresponding to the average value of rhythmic frequencies over a long time interval consisting of K specified time intervals with rhythmic frequencies. 2. A device for implementing a method for measuring rhythmic frequencies, power and duration of decays of acoustic signal non-stationarity sections, containing a series-connected input unit, a linear analog-to-digital converter, a Hilbert orthogonal transform unit, an amplitude envelope calculation unit, a low-frequency filter, and a key unit, a digital quadrator, an adder-averaging unit, a first memory block, a second memory block and an indication block with a display, characterized in that a block for detecting two sections of non-stationarity, a block for determining rhythmic frequencies, a block for determining the duration of recessions and a third memory block are additionally introduced, while the first and the second outputs of the Hilbert orthogonal transform block are connected, respectively, to the first and second inputs of the amplitude envelope calculation block, the output of which is connected to the input of the low-pass filter, the output of which is connected to the first input of the key block and the input of the two-count detection block The non-stationarity trigger, the first output of which is connected to the input of the block for determining rhythmic frequencies, and its second output is connected to the second input of the block of keys and to the first input of the block for determining the duration of the recessions, the second input of which is connected to the third output of the block for detecting two sections of non-stationarity, and the output of the block of keys connected to the input of a digital quadrator, the output of which is connected to the first input of the adder-averaging, the second input of which is connected to the third output of the block for determining the duration of the declines, and the output of the adder-averaging is connected to the first input of the first memory block, the first and second outputs of which are connected, respectively, with the first and second inputs of the indication block with a display, the first output of the block for determining the duration of the recessions is connected to the first input of the second memory block, and the second output of the block for determining the durations of the recessions is connected to the second input of the first memory block and the second input of the second memory block, the first and second exits which its connected, respectively, with the third and fourth inputs of the indication block with the display, and the first and second outputs of the block for determining rhythmic frequencies are connected, respectively, with the first and second inputs of the third memory block, the first and second outputs of which are connected, respectively, with the fifth and sixth inputs of the indication block with the display.

Images