Classification method for hydroacoustic noise emission signals of marine object

Abstract

FIELD: hydroacoustics.SUBSTANCE: invention can be used in tasks of determining the class of an object when developing hydroacoustic systems. A method for identifying between the CPS of a detected object and the CPS of a portrait from the database of CPS portraits using the correlation coefficient is proposed. Based on n readings of the spectrum levels of the detected object and by searching through n readings of the levels of k portraits of the DB, the correlation coefficient CC is calculated. The maximum CC is selected, compared with the threshold CC, and if the threshold is exceeded, a decision is made about the class of the object.EFFECT: increase in the reliability of classification by creating a database of portraits based on the continuous part of the spectrum (CPS) in one-third octave frequency bands.1 cl, 1 dwg

Description

[0001] The invention relates to the field of hydroacoustics and can be used in the tasks of determining the class of an object when developing hydroacoustic systems.

[0002] The acoustic field of an object can be represented as the sum of broadband noise with a continuous spectrum and narrowband discrete components (DC). Broadband noise is caused mainly by the object's propulsion devices (in particular, cavitation on propellers), as well as by hydrodynamic processes during flow past the object's hull, wheelhouse, and the oncoming flow of the steering system, and sharply increases with increasing speed. It is known that at high speeds, propeller cavitation noise, which has a continuous spectrum, suppresses most of the DC, and in the overall spectrum this noise becomes predominant. A decrease in immersion depth at a constant speed has the same overall effect on the propeller noise spectrum as an increase in speed at a constant depth. In this case, maxima are formed in the continuous noise spectrum. The position of these maxima depends on the specific object, its speed, and immersion depth. [Robert J. Urick Fundamentals of Hydroacoustics. - Shipbuilding. - 1978. - P. 350-359].

[0003] When observing an object, simultaneously with the noise signal emitted by the object, a signal from interference is received, which includes components of noise interference, sea noise and local interference of shipping. The interference level will be added to the level of the detected target, which will distort the actual ratios of the levels of the continuous part of the spectrum (CPS) of the object's signal, which will lead to distortion of the classification results. Therefore, in the proposed classification method for reducing the influence of the spectrum of interfering targets (interference) received by the side field of the antenna: the method is used [Afanasyev AN, Znamenskaya TK; OJSC "Concern "Oceanpribor". Method for classifying hydroacoustic noise signals of a marine object. Patent No. 2546851 RF, IPC G01S 3 / 80. No. 2013155104 / 28; Cl. 11.12.2013; Published. 10.04.2015, Bulletin No. 10], which contains the following operations:

[0004] • noise emission signals from a marine object in an additive mixture with interference are received by two semi-antennas of a hydroacoustic antenna,

[0005] • perform spectral processing of received signals at the outputs of the semi-antennas,

[0006] • sum the power spectra from the outputs of two half-antennas ,

[0007] • find the difference power spectra from the outputs of two semi-antennas,

[0008] • accumulate and smooth the total power spectrum and the difference power spectrum by frequency,

[0009] • determine the noise emission power spectrum of a marine object as a difference power spectrum ,

[0010] • define the detection threshold for the difference power spectrum,

[0011] • decide on the presence of discrete components by which to classify a marine object.

[0012] A method for classifying hydroacoustic noise signals of a marine object is also known [Znamenskaya TK; JSC Concern Okeanpribor. Method for classifying hydroacoustic noise signals of a marine object. Patent No. 2801677 of the Russian Federation, IPC G01S 3 / 80. No. 2022123375; Cl. 31.08.2022; Published 14.08.2023, Bulletin No. 23], which is closest to the proposed one in terms of technical essence and the number of common features.

[0013] The prototype method contains the following operations:

[0014] • reception of noise signals from a marine object by the antenna;

[0015] • in a given frequency range (FR), a 1 / 3-octave analysis of the continuous part of the spectrum (CPSS) of the object at the current moment in time is performed using 1 / 3-octave filters;

[0016] • in each passband of 1 / 3 octave filters, after the procedure of accumulation and smoothing of spectra, the spectral level of the signal of the current spectrum is calculated by summing the frequency samples;

[0017] • determine the position of the maxima in the spectrum of a given frequency response by exceeding the detection threshold by signal levels in 1 / 3 octave filters, determine the average frequencies f cp.1 / 3i detected maxima and their width ΔF i , and the class of the object is determined by the detected DS together with the parameters f cp.1 / 3i and ΔF i SChS.

[0018] The emergence of new classes of marine objects (MO), the increasing diversity of MO within the identified classes, and changes in the acoustic characteristics of known MOs lead to a decrease in the probability of correctly classifying marine objects. This, in modern conditions, is a significant drawback of the considered classification method based on generalized classification parameters. Involving additional customization of the MO classification method complicates the operation of the hydroacoustic system, but cannot always eliminate this drawback.

[0019] The objective of the present invention is to increase the classification reliability of the method for classifying noise emission of a MO.

[0020] The technical result of the invention consists in increasing the reliability of classification by creating a database of portraits based on the SCS in one-third octave frequency bands and the proposed method for identifying between the SCS of a detected object and the SCS of a portrait from the SCS portrait database.

[0021] To ensure the specified technical result in a method for classifying hydroacoustic noise emission signals of a marine object, in which a noise emission signal of a marine object is received in an additive mixture with interference by two semi-antennas of a hydroacoustic antenna, spectral processing of the received signals is performed at the outputs of the semi-antennas, the power spectrum of the noise emission of the marine object is determined as a difference power spectrum The obtained power spectra are accumulated, the spectrum is smoothed by frequency, and a 1 / 3-octave analysis of the continuous part of the spectrum (CFS) of the object at the current moment in time is performed using 1 / 3-octave filters in a given frequency range (FR). To do this, in each passband of 1 / 3-octave filters, after the procedure of accumulation and smoothing of the spectra, the spectral level of the signal (SL) is calculated. 1 / 3 ) the current spectrum by summing the frequency counts and determine the class of the marine object, new features are introduced, namely: a database (DB) of “spectral portraits of objects according to the SCS” is created, where each k-th portrait represents n samples of the UR levels b1 / 3i in 1 / 3 octave filters in a given frequency range, a given region, and a given motion mode. Based on a sample of n UR readings i1 / 3, the difference spectrum of the signal power, recalculated at the antenna input using the transfer characteristic of the receiving path at the current moment in time, the correlation coefficient (CC) is calculated between the array of n UR readings 1 / 3 at the input of the antenna with arrays of n UR readings b1 / 3 for each portrait. To do this, the CC is calculated for each portrait using the formula

[0022]

[0023] where The maximum correlation coefficient value is selected and compared with the correlation coefficient threshold. If the correlation coefficient value is greater than or equal to the threshold, a flag is generated indicating that the detected object belongs to the class of the corresponding portrait in the database.

[0024] The essence of the invention is explained in Fig. 1, which shows a block diagram of a device that implements the method taking into account the possibilities of constructing modern hydroacoustic systems.

[0025] The signal S(t) (Fig. 1) from the outputs of block 1 of the half-antennas A1 and A2 is fed respectively to the input of block 2 of the analog-to-digital converter (ADC), which is connected to the universal microprocessor 3. The universal microprocessor includes a series-connected unit for calculating the difference power spectrum 4, an accumulation unit 5, a unit for forming n frequency bands of 1 / 3 octave standard filters by the power spectrum 6, a unit for calculating signal levels in n 1 / 3 octave frequency bands at the antenna input 7, a unit for converting signal levels to the radiation point 8, a unit for calculating the correlation coefficient KK 11, a classification unit 12, a display and control unit 13. Also, the microprocessor 3 includes a series-connected unit for controlling the filling mode of the DB of "spectral portraits by SCS" 9, the output of which is connected to the input of the DB block of "spectral portraits by SCS" 10.The second output of block 5 is connected to the second input of block 13, the output of which is connected to the input of block 9, the output of which is connected to the second input of block 6. The second outputs of blocks 7 and 8 are connected to the second and third inputs of block 10, the output of which is connected to the second input of block 11.

[0026] Block 2 can be implemented as described in the Handbook [Goldenberg LM et al. Digital signal processing: Handbook. - Radio and Communications, 1985. - 91 p.]. Universal microprocessor 3 has the ability to operate in real time, the ability to switch from one task to another, flexible memory addressing, and high data processing speed. [Koryakin Yu.A., Smirnov SA, Yakovlev GV Ship hydroacoustic equipment. - St. Petersburg: Nauka. - 2004. - 284 p.]. Block 4 can be implemented as described in patent No. 2801677. In block 5 for accumulating the difference power spectrum, the averaged (accumulated) power spectrum is determined [Kharkevich AA Interference Control / AA Kharkevich. - 2nd ed., corrected. - M.: Nauka, Chief Editor of Phys.-Math. Literature, 1965. Pp. 70-71]. Block 6 for forming standard 1 / 3-octave filters and block 7 for calculating signal levels in 1 / 3-octave frequency bands. [Kolesnikov A.E. Handbook of Hydroacoustics / / L.: Shipbuilding. - 1982. - P.229], and block 8 can be implemented as described in the article [Zhumenkova S.V., Konyukhova G.V., Mashoshina A.I., Pestereva I.S. Methodology for operational measurement of the primary hydroacoustic field of underwater vehicles / / Underwater research and robotics. - No. 3 (45). - 2023. - P. 5-7]. Block 9 and block 13 can be implemented as described in the book [Koryakin Yu.A., Smirnov S.A., Yakovlev G.V. Ship hydroacoustics. - St. Petersburg: Nauka. - 2004. - P. 255-261], and block 10 of the DB of "spectral portraits by SCS" can be implemented on the basis of a universal microprocessor [Koryakin Yu.A., Smirnov S.A., Yakovlev G.V. Ship hydroacoustics. - St. Petersburg: Nauka. - 2004. - P. 284]. The correlation coefficient of the CC of block 11 may have been calculated using the formulas in the book [Pustylnik E.I. Statistical methods of analysis and processing of observations. - Nauka. Ed. in chief of physical and mathematical literature, 1968. - P. 217-219].

[0027] The implementation of the method using the device shown in the block diagram is carried out as follows.

[0028] Signals S i (t) half-antennas A1 and A2 are fed to the input of block 2 of the ADC, signal S i (k) from the ADC in the form of discrete samples is sent to the universal microprocessor 3. In block 4, to eliminate the influence of the interference spectrum, after a fast Fourier transform, a sum channel is formed for the obtained complex spectra of the two half-antennas difference channel power spectrum and difference power spectrum In block 5, the accumulated difference power spectrum is determined The accumulated power spectrum of the detected object is fed to block 6, which generates n frequency bands using 1 / 3-octave standard filters in a given frequency range. The lower and upper frequencies of the receiving path's frequency range are set by block 9, which is controlled by display and control block 13. Block 9, which controls the "spectral portraits by SFS" database filling mode, handles the generation, supplementation, and correction of the database, manages the operator's ability to select the desired section and / or its subsection, and provides for the ability to use the entire database to calculate the CC between database portraits and the current SFS power spectrum of the detected object. When generating an object portrait in real-time mode, the object class must be confirmed by the operator. Two portrait control and display modes are provided: automatic and manual. Object portraits in the "spectral portraits by SFS" database, block 10, are represented as signal levels in 1 / 3-octave frequency bands in a given frequency range.Object portraits using the SCS can be generated based on a priori known data at the emission point, based on the results of statistical processing of observations under known test conditions, enabling the recalculation of n levels of 1 / 3-octave frequency bands at the emission point (block 8). If the necessary data on the current experiment conditions are unavailable, the portrait can be entered into the database in the operating mode (block 7). The time of portrait recording and the time of the last adjustment of this portrait are automatically recorded in the portrait database. Multiple portraits for the same object can be entered into the database, depending on the object's operating mode during observation (tracking). The object class is a mandatory parameter for the database. In block 11, n US readings are stored. 1 / 3 spectrum of the detected object and by searching through n readings of the US b1 / 3The correlation coefficient (CC) is calculated for the DB portraits. The maximum CC is selected and compared with the CC threshold. A detected object and the DB object are considered to be of the same class if the CC value is ≥ the CC threshold. It is known that the noise spectra of objects of different classes (ship, aircraft, helicopter, etc.) have a certain similarity; therefore, to compare the noise characteristics of objects, the CC threshold is set sufficiently high (approximately 0.85). The signal propagation conditions from an object, which affect the profile, vary for different ocean regions. Since recalculating signal levels to the emission point, taking into account propagation characteristics, is quite complex and it is not always possible to take into account all the data necessary for recalculating the noise characteristics of an object to the emission point, it is recommended to distribute object portraits in the DB by ocean region. Furthermore, the composition of a domestic and foreign fleet depends on the combat service area.The number of portraits in the database will be limited by the corresponding tasks performed in the area. During long-term monitoring in operational mode, both the portrait itself and the class entered by the operator can be refined.

[0029] In block 13, in addition to the automatic decision on the value of the CC about the class of the object from the DB (block 10) and blocks 7 or 8, the operator is shown the levels of the US on the display i1 / 3 and US ib1 / 3 in one-third octave frequency bands. The abscissa axis represents intervals of 1 / 3 octave frequency bands centered on the geometric mean frequency, and the ordinate axis represents the values ​​of the σ 1 / 3 and US b1 / 3 For clarity, the obtained points of n readings of the US 1 / 3 and US b1 / 3are connected by straight line segments. Thus, using the fragment with histograms of the one-third-octave spectra of the SCS of the detected object and the portrait from the SCS database, the operator can intervene in the identification process using the SCS correlation coefficient parameter. To ensure a correct comparison of spectral portraits using 1 / 3-octave UCs 1 / 3 . the spectra of the frequency response system are normalized to the maximum level in the 1 / 3 octave filter.

[0030] As a result, the increase in the reliability of the classification occurs due to the portrait classification of newly emerged types of MO, the features of which could not be taken into account using generalized classification parameters.

Claims

A method for classifying hydroacoustic noise signals from a marine object, including receiving a noise signal from a marine object in an additive mixture with interference by two semi-antennas of a hydroacoustic antenna, spectral processing of the received signals at the outputs of the semi-antennas, and determining the power spectrum of the noise emission from the marine object as a difference power spectrum accumulation of the obtained power spectra, spectrum smoothing by frequency, 1 / 3-octave analysis of the continuous part of the spectrum (CPS) of the object at the current moment in time using 1 / 3 octave filters in a given frequency range (FR), for which in each passband of 1 / 3 octave filters after the procedure of accumulation and smoothing of the spectra the spectral level of the signal (SL) is calculated 1 / 3) the current spectrum by summing the frequency readings and determining the class of a marine object, which is distinguished by the fact that a database (DB) of “spectral portraits of objects according to the SCS” is created, where each k-th portrait represents n samples of the UR levels b1 / 3i in 1 / 3 octave filters in a given frequency range, a given area and a given movement mode, based on a sample of n UR readings i1 / 3 , the difference spectrum of the signal power, recalculated at the antenna input using the transfer characteristic of the receiving path at the current moment in time, the correlation coefficient (CC) is calculated between the array of n UR readings 1 / 3 at the input of the antenna with arrays of n UR readings b1 / 3 for each portrait, the CC is calculated for each portrait using the formula Where The maximum value of the correlation coefficient is selected and compared with the threshold value K, and if the value of the correlation coefficient is greater than or equal to the threshold, a sign of the belonging of the detected object to the class of the corresponding portrait in the database is generated.

Images