RU2690027C1 - Method for simultaneous measurement of speech intelligibility of several sources - Google Patents

Abstract

FIELD: optics.SUBSTANCE: invention relates to acoustic metrology, particularly to instruments for determining speech intelligibility. According to the method, optical fibres are laid, divided into measuring sections, connected to a measuring module, a signal is emitted, the maximum length of return of the reflected signal is used to determine the length of the fibre, and the period T of measuring pulse signals is calculated. Receiver receives an acoustically modulated sequence of reflected measuring pulses, demodulates the acoustic component of the received reflected measuring pulse signals, and measures the level of natural noise. M sources of signals are then placed, and N frequencies of acoustic signals are sequentially transmitted by each of the M sources. Radiation frequencies are distributed over the average frequencies N of the bands, receiving the signals, calculating the signal/noise ratio, constructing a variation series of signal/noise ratio values, selecting a leading term of the variation series, characterizing variation of individual features of a source of acoustic test signals is picked up and memorized. Then, acoustic signals are emitted simultaneously, signal-to-noise ratio is measured, a row is constructed. Level of speech intelligibility is calculated from the senior member of the variation series of each of M sources of acoustic test signals.EFFECT: technical result is shorter time and high accuracy of simultaneous measurement of speech intelligibility.1 cl, 3 dwg

Description

The invention relates to a measuring technique and can be used for simultaneous measurement of speech intelligibility and is intended to assess the protection of objects from unauthorized leakage of acoustic speech information (ARI) in real conditions.

The “Method of measuring speech intelligibility” is known for RF Patent No. 2284585 dated September 27, 2007, consisting in that in order to improve the accuracy and reliability of speech intelligibility, measurement is carried out with an objective correction of the measurement path to compensate for the difference between the specified and measured test signals in the place of their real radiation [1].

The disadvantage of the prototype is that it does not allow for speech recognition with the stated technical result in a real signal and noise environment due to the rigid one-time fixation of radiation points and the reception of test signals.

Known instrumental and calculation formant method for assessing speech intelligibility, based on the results of experimental studies conducted and published in 1962 by N.B. Pokrovsky. The essence of the method consists in decomposing speech into elementary signals, taking into account the losses of these elementary signals as they pass through the path and find the relative number of them that have reached the ear of the listener without distortion. Formants can be considered as such elementary signals forming speech sounds. Estimating formant intelligibility of speech, the entire analyzed frequency range is divided into 20 adjacent bands, with central frequencies and boundary frequencies, within each of which the spectra of speech and noise, as well as the formant density of formants, can be considered practically unchanged. Formant legibility is calculated as the sum of the formant clearances in each of the bands. To assess speech intelligibility, it is necessary to measure the levels of the hidden speech signal and noise (interference) at the place of possible placement of acoustic receivers or at the place of possible listening to speech without the use of technical means. It is considered that speech recognition is possible if the value of verbal speech intelligibility calculated by the measurement results exceeds the established norms [2].

The fundamentally difficult task for the formant approach is to evaluate speech intelligibility under conditions of small signal-to-noise ratios. Partially, this problem can be overcome with the help of test signals of a higher level. However, at the next stage, when converting signal-to-noise ratios to speech intelligibility, the problem of low accuracy of estimation arises again due to small absolute values of the coefficient of perception.

The closest analogue to the technical nature of the claimed method and adopted for the prototype is "Method of measuring speech intelligibility" for the RF patent №2620569 from 05.26.2017, consisting in the fact that in it to reduce time and improve the accuracy of calculating the level of intelligibility of speech used a spatial-distributed acoustic signal transducer (PRP AC) based on an optical fiber (OM) and uses a real-time point-to-multipoint measurement circuit (consisting of a single source of acoustic signals and multiple receivers) [3].

The disadvantage of the prototype is that it does not allow simultaneous speech recognition of multiple sources with the stated technical result in real signal and noise environment due to the lack of implementation of the multipoint-to-multipoint measurement circuit (multiple sources - multiple receivers) in real time.

The prior art has not identified solutions for methods of simultaneous measurement of speech intelligibility of several sources, characterized by the claimed combination of features, therefore, the present invention meets the patentability criterion - "novelty."

Search results known solutions in this and related areas of technology in order to identify signs that match the distinctive features of the analogues of the claimed invention, have shown that they do not follow explicitly from the prior art. From the prior art defined by the applicant, the influence of the envisaged essential features of the claimed invention on the achievement of this technical result is not revealed. Therefore, the claimed invention meets the condition of patentability "inventive step".

The objective of the invention is to develop a method that reduces the time and improves the accuracy of simultaneously measuring the level of speech intelligibility of several sources of acoustic speech information (ARI) through the use of a spatially distributed acoustic signal transducer (PRP AC) based on optical fiber (OM) as K combinations microphone -receiver, which allows you to implement a multi-point-to-multipoint measurement scheme in real time.

The technical result is achieved by the fact that in a known method of measuring speech intelligibility, which consists in laying a spatial-distributed acoustic signal transducer represented by an optical fiber at given points of the allocated room, programmatically breaking it into K measuring sections and specifying K measurement points, each which acts as a separate receiver of acoustic signals, connect the spatially distributed transducer of the acoustic signal to the measuring module located in the place of unauthorized removal of information, emit a test pulse signal with the optical emitter of the measuring module, determine the length L of the spatially distributed acoustic signal transducer by the maximum return time of the reflected test pulse signal; so that each subsequent measuring pulse signal is radiated after receiving the reflected signal of the previous measurement pulse signal, emit measuring pulse signals with a period T, the receiver of the measuring module receives an acoustically modulated sequence of reflected measuring pulse signals from different predetermined K points of the allocated room, identified by the return time of the reflected measuring pulse signals t _k , and demodulate the acoustic component of the received reflected measuring pulse signals from different K points of the allocated room, measure the level of natural noise in the absence of acoustic test signals in different K points of the allocated room in the audible speech spectrum Δƒ _a , M sources of acoustic test signals of a given power are also placed in the allocated room, N frequencies of test acoustic signals of a given level are emitted each by M sources of frequencies The radiations are distributed over the mid frequencies of N bands, into which the audible spectrum of speech is divided, receive and measure the level of acoustic a signal with noise at each frequency of acoustic test signals at different K points of the allocated room, calculating the signal-to-noise ratio at different K points of the allocated room, constructing a variational series of signal-to-noise ratio values, choosing the senior member of the variational series to emphasize the nature of changes in the individual characteristics of the acoustic source test signals, select and remember the nature of changes in the individual characteristics of the source of acoustic test signals, while radiating M sources of acoustic test signals at N frequencies, acoustic signals, receive and measure the level of acoustic signals with noise at each frequency of acoustic signals at different K points of the allocated room, calculate the signal-to-noise ratio at different K points of the selected room, build a variational series using individual characteristics values of the signal-to-noise ratio for each of the M sources of acoustic test signals, by the senior member of the variational series of each of the M sources of acoustic tests Signal signals compute the level of speech intelligibility.

Thanks to a new set of essential features, the claimed method reduces time and improves the accuracy of simultaneously measuring the level of speech intelligibility of several sources by using a spatially distributed acoustic signal transducer and implementing a multipoint-to-multipoint measurement circuit in real time. As PRP AC is used OV, which replaces any necessary number of point combinations of microphone-receiver at various points of the allocated room (VP) of any shape.

The claimed method is illustrated by drawings, which show:

FIG. 1 - Block diagram of the device that implements the proposed method and mechanism of the acoustic signal effect on the OV, where:

1 - measuring module;

2 - optical emitter;

3 - optical power divider;

4 - optical signal receiver;

5 - device analysis and control of the measuring module;

6 - demodulator;

7 - a spatially distributed acoustic signal transducer (PRP AC) based on optical fiber (OM) in a dedicated room (VP);

8 - impurities in the optical fiber;

9 - modulation of the optical signal by an external influence (sound wave);

10 - source of acoustic test signals (AIS);

11 - measuring optical radiation;

12 - acoustically modulated optical signal reflected from impurities of the optical fiber;

FIG. 2 - Location of AIS sources and given optimal points of reception of the acoustic signal in the VI, where:

13 is a section of optical fiber outside the designated room;

14 - measuring points (acoustic sensors) along a spatially distributed acoustic signal transducer based on optical fiber;

FIG. 3 - Correction of the optimal points of reception of the acoustic signal along the laid OV according to the measurement results, where:

15 - correction of measurement points along a spatially distributed transducer of an acoustic signal based on optical fiber.

The claimed method is implemented as follows.

The use of PRP AC based on OB 7 using an acousto-optic effect for speech recognition outside the EP boundaries through standard fiber-optic information telecommunications for various purposes located in them is a consequence of the emergence and development of fiber-optic technologies.

Sources AIS 10 affect the PDP AS based on S 7 of the standard fiber-optic transmission systems of the EAP and cause the optical signal to be modulated by 9 acoustic frequencies [4]. The reflected optical signal 12 modulated by the acoustic component of the external action extends beyond the EF and is received by measuring module 1 connected along the ground 13, with further demodulation of the acoustic component and measurement of speech intelligibility by known methods based on the methods described in [2].

The measuring module 1 consists (see Fig. 1) of the optical emitter 2, the optical power divider 3, which provides the input of the reflected optical signal into the receiving path of the measuring module, the successor of the optical signal 4 and the analysis and control device 5 of the measuring module 1, which has the demodulator 6, which provides demodulation of the acoustic component (Δfa) of the optical signal 4 of the acoustically modulated optical return signal 12 converted by the receiver, while the analysis and control device 5 and 5 are connected in series with the optical emitter 2, the output of which is connected to the input of the optical power divider 3, the output of which is connected to the input of the receiver of the optical signal 4 connected to the demodulator 6.

The principle of operation of the measuring module 1 with OVP APS based on S is based on the principle of operation of an optical reflectometer: a powerful measuring optical radiation 11 (impulse) is introduced into the tested EGS 2 based on S 7 with an optical radiating element 2 and the characteristics scattered on impurities 8 distributed throughout the length of the PDP AC on the basis of OB 7, reflected back optical radiation 12. Due to the sensitivity of the receiving part of the measuring module 4 to phase (amplitude, frequency, polarization) modulation (for example, When using a Mach-Zehnder interferometer [5]) in the PDP AU based OB possible as the measurement of acoustic effects across the test fiber length and localization measurement at any portion thereof due to the different return times reflected from the optical signals impurities t _k (FIG. one). The change in acoustic pressure on short sections of the PDP AC based on OB 7 is determined by the difference in reflectograms in time and analyzed either by a computer using special software or by an operator visually and by ear. Thus, the FPP AC based on S is used as a system of K distributed measurement points (acoustic sensors) 14.

The claimed method can be used in the EAP of any form, while the sources of AIS 10 can also take any place in it. To do this, inside the VP, measurement points 14 are set, through which a PDP AC is built based on OB 7. Later, according to the measurement results, the position of these points can be corrected along the located PDP AC based on OB 15. It is also possible to use regular fiber optic telecommunications information services. In this case, use any, determined from the results of measurements, the optimal point of reception of the acoustic signal along the existing OB. The measuring module (IM) 1 is connected to the OV abroad VP 13.

Next, conduct simultaneous measurement of speech intelligibility of several sources, for which:

1) emit by the optical emitter 2 of the measuring module 1 a test pulse signal to determine the maximum return time of the reflected test pulse signal t _{max of the} length L of the RLP AC on the basis of the OB.

For example: when the VP perimeter is 70 meters (P = 70 m) and the length L of the spatially distributed acoustic signal transducer is equal to this perimeter (L = P), the maximum return time of the reflected test pulse signal at the received speed of light in a quartz fiber (corresponding ITU-T Rec. G.652.D) with _RH = 204000 km / s, will be 686 nanoseconds (t _max = 686 ns). At the same time, by the Kotelnikov theorem, in order to guarantee the restoration of an acoustic signal in the frequency range from 20 Hz to 8 kHz, it is necessary to obtain samples with a period of 125 μs;

2) in accordance with the length L of the spatially distributed acoustic signal transducer based on the OV, calculate the period T of the sending of the measuring pulse signals so that each subsequent measuring pulse signal is emitted after receiving the reflected signal of the previous measuring pulse signal;

3) determine the area of the PDP AC, which is directly inside the EAP. To do this, any of the MIS AIS 10 sources located in the airspace emit a set acoustic signal and, according to the nature of its impact on the exposure agent, determine the area of the controlled exposure agent located within the location interface.

Because the fiber-optic telecommunications, inside the controlled object VI usually laid not only in the CAP, but also through other object space, it is necessary to determine the portion L _VP AC PDP, which is directly inside VI. This action is necessary to reduce the load on the measuring device, which will reduce the time for assessing speech intelligibility of AIS sources, as well as reducing the power of measuring optical pulse signals, which will reduce the effect of nonlinear effects inside the optical fiber on the reflected optical signals scattered on impurities [6];

4) programmatically divide the area L of the _RP URP AC on the basis of OB 7 into K measurement sites and specify K measurement points 14, each of which acts as a separate receiver of acoustic signals.

This action provides a “multi-point” from the receiving side, which ensures further selection of the optimal, in relation to the signal-to-noise ratio, source / receiver pairs for each of the M sources of AIS 10;

5) measure the level of natural noise in the absence of acoustic test signals at different K points 14 in the audible speech spectrum Δƒ _a .

For this:

- the optical emitter 2 of the measuring module 1 in the absence of AIS 10 sources emit measuring pulse signals with a period T, and the receiver 4 of the measuring module 1 receive an acoustically modulated sequence of reflected measuring pulse signals from various specified K points 14 VP identified by the return time of the acoustically modulated reflected measuring pulse signals f _k ;

- demodulate the acoustic component of the received acoustically modulated background noise reflected at impurities 8 measuring pulse signals from different K points 14 VP in the absence of AIS 10 sources;

- measured in the absence of sources of AIS 10, the level of natural noise at various K points 14 VP;

6) place the MIS sources AIS 10 in the EP and consistently, each of the M sources AIS 10:

- N radiate a predetermined frequency level, wherein the frequency of the radiation distributed over the midrange N bands to which _a speech is divided Δƒ audible range;

- receive and measure the level of the acoustic signal with noise at each frequency from each of the M sources AIS 10 at different K points 14 VP;

- calculate the signal-to-noise ratio at various K points of 14 VP;

- build a variation range of signal / noise ratio values;

- choose the senior member of the variation series to highlight the nature of changes in the individual characteristics of the source of AIS 10;

- identify and remember the nature of changes in the individual characteristics of the source of AIS 10.

The output of paragraph 5 will allow, with the simultaneous emission of M AIS sources, to distinguish any of the sources from others, taking their demodulated signals for noise. In practice, data on the individual acoustic features of the sources of acoustic speech information, determined by the unevenness of their amplitude-frequency characteristics and the spectral density of speech [2], is obtained from any available source (audio recordings, telephone conversations, corporate databases, etc.).

7) simultaneously emit acoustic signals from M AIS 10 sources at N frequencies;

8) receive and measure the level of acoustic signals with noise at each frequency of the acoustic test signal at different K points 14 VP.

At this stage, the “multipoint-multipoint” measurement scheme is implemented in real time, which distinguishes the claimed method from any known one.

9) calculate the signal-to-noise ratio at different K points 14 VP;

10) build a variation range of signal-to-noise ratio values for each of the AIS 10 M sources according to individual characteristics;

11) the senior member of the variation series of each of the M sources of AIS 10 calculate their level of speech intelligibility in a known manner [7].

Simultaneous measurement of the acoustic situation of VP at K points with multiple AIS sources corresponds to a measurement scheme consisting of multiple transmitters and multiple receivers - “multipoint-multipoint”, in contrast to modern measurement methods that simultaneously work with a single acoustic signal transducer -point "(one source - one receiver) [7]," point-to-multipoint "(one source - many receivers) [3]. The “multipoint” on the measurement side allows to significantly reduce the computation time and measure the level of speech intelligibility due to the absence of the need to physically change the position of the receiver to obtain the best signal-to-noise ratio. In addition, the microphone-receiver-distributed KI combinations allow, for each of the AIS sources chaotically located in the room M, to calculate in real time the optimal receiving point with the best signal-to-noise ratio, which implements the “multipoint-to-multipoint” measurement scheme and provides an increase the accuracy of calculating the level of intelligibility of speech.

Thus, through the use of a spatially distributed transducer of an acoustic signal and the implementation of a multi-point-to-multimeter measurement circuit, the implementation of a technical result is achieved.

Information sources:

1. RU, patent No. 2284585 C1, IPC G10L 15/00, H04R 29/00, 2007, Bull. №27.

2. Pokrovsky N.B. Calculation and measurement of speech intelligibility. Moscow: Svyazyzdat, 1962

3. RU, patent No. 2620569 C1, IPC G10L 15/00, H04R 29/00, 2017 Bull. №15.

4. Grinev A.Yu., Naumov K.P., Preslenev L.N., Tigin D.V. Optical devices in radio engineering: Proc. manual for universities / ed. V.N. Ushakov. Ed. 2nd, rev. and add. - M .: Radio Engineering, 2009, pp. 86-152.

5. Bykov V.P. Laser electrodynamics. Elementary and coherent processes in the interaction of laser radiation with matter. - M .: FIZMATLIT, 2006, pp. 50-77.

6. Tsukanov V.N., Yakovlev M.Ya. Fiber optic technology. A practical guide. - M .: Infra-Engineering, 2014. - 304 p.

7. Reva I.L. Comparative analysis of objective methods for assessing speech intelligibility. Collection of scientific papers of NSTU - 2010. №1 (59). - 91-102.

Claims (1)

The method of simultaneous measurement of speech intelligibility of several sources, which consists in laying a spatial-distributed acoustic signal transducer represented by an optical fiber at given points of the allocated room, programmatically breaking it into K measurement sites and specifying K measurement points, each of which acts as a separate acoustic receiver. signals, connect the spatially distributed transducer of the acoustic signal to the measuring module, emit the measurement pulse module determines the length L of the spatially distributed acoustic signal transducer by the maximum return time of the reflected test pulse signal, according to which the period T of the pulse measurement signals is calculated so that each subsequent measurement pulse signal is emitted after receiving the reflected signal previous measuring pulse signal, emit measuring pulse signals with a period T, the meter module receives an acoustically modulated sequence of reflected measurement pulse signals from different specified K points of the highlighted room, identified by the return time of the reflected measurement pulse signals _tk , demodulates the acoustic component of the received reflected measurement pulse signals from different K points of the highlighted room, measure the level of natural noise in the absence acoustic test signals at various K points of the assigned room I in the audible spectrum of speech Δƒ _a, characterized in that the selected room is placed the M source of acoustic test given power signals successively each of M acoustic test signal sources emit N frequency test acoustic signals a predetermined level, wherein the frequency of the radiation distributed over the midrange N bands , which is divided into an audible spectrum of speech, receive and measure the level of the acoustic signal with noise at each frequency of acoustic test signals in different K points of the allocated room, calculate the signal-to-noise ratio at various K points of the allocated room, build a variational series of signal-to-noise ratio values, choose the senior member of the variational series to highlight the nature of changes in the individual characteristics of the source of acoustic test signals, select and remember the changes in the individual characteristics of the acoustic source test signals, simultaneously emit M sources of acoustic test signals at N frequencies acoustic signals aly, receive and measure the level of acoustic signals with noise at each frequency of acoustic signals at different K points of the allocated room, calculate the signal-to-noise ratio at different K points of the selected room, build a variational series of signal-to-noise ratio values for each of the M sources according to individual characteristics acoustic test signals, the senior member of the variation series of each of the M sources of acoustic test signals calculate the level of speech intelligibility.

Images