A target depth discrimination method based on negative eikonal waveguide horizontal array wave number difference domain feature extraction

Abstract

This invention provides a target depth discrimination method based on wavenumber difference domain feature extraction of a horizontal array of negative-gradient waveguides, belonging to the field of target depth discrimination technology. This invention uses a horizontal linear array deployed on the seabed to receive broadband acoustic pressure signals from shallow-sea negative-gradient waveguides. After performing Fourier transform on the received signal, wavenumber domain beamforming is performed in the frequency domain to obtain the beam output. The beam output is scanned on the horizontal wavenumber domain to obtain the wavenumber spectrum. A two-dimensional distribution matrix of horizontal wavenumber difference versus frequency is obtained through coordinate transformation, and a two-layer game model is used to optimize the sampling accuracy. Energy accumulation is performed along the full bandwidth of the signal at each sampling point in the horizontal wavenumber difference domain to construct a cumulative energy versus horizontal wavenumber difference curve. The coordinates corresponding to the peak value of the cumulative energy versus horizontal wavenumber difference curve are calculated and compared with the criteria to complete the depth discrimination, thus solving the technical problem of insufficient accuracy in sound source depth discrimination in shallow-sea negative-gradient waveguides.

Description

Technical Field

[0001] This invention belongs to the field of target depth discrimination technology, and specifically relates to a target depth discrimination method based on wavenumber difference domain feature extraction of a negative gradient waveguide horizontal array. Background Technology

[0002] In the field of shallow-sea sonar detection, depth discrimination of sound sources based on signals received by a horizontal array is a key technology for underwater acoustic target identification. Traditional depth discrimination methods mainly rely on normal mode analysis and beamforming techniques to distinguish between surface and underwater sound sources by extracting frequency-wavenumber domain features of the sound field. In the current negative-sound-step waveguide environment, due to the refraction effect of the step layer on the sound propagation path, the normal mode structures excited by sound sources at different depths exhibit significant differences. Existing methods perform depth discrimination by analyzing the mode energy distribution. However, in wavenumber domain feature extraction, existing technologies struggle to simultaneously optimize sampling accuracy and energy resolution, leading to fuzzy peak features in the horizontal wavenumber difference domain and decreased discrimination accuracy in complex marine environments. In other words, existing technologies suffer from insufficient accuracy in sound source depth discrimination in shallow-sea negative-sound-step waveguides. Summary of the Invention

[0003] In view of this, the present invention provides a target depth discrimination method based on wavenumber difference domain feature extraction of horizontal array of negative gradient waveguide, which can solve the technical problem of insufficient accuracy of sound source depth discrimination in shallow sea negative gradient waveguides in the prior art.

[0004] This invention is implemented as follows: This invention provides a target depth discrimination method based on wavenumber difference domain feature extraction of a horizontal array of negative-sound-gradient waveguides, comprising the following steps: A horizontal linear array deployed on the seabed receives broadband acoustic pressure signals emitted by a sound source in a shallow-sea negative-sound-gradient waveguide; the received broadband acoustic pressure signals are Fourier transformed to obtain a frequency domain signal; wavenumber domain beamforming is performed on each frequency point within the range from the start frequency to the cutoff frequency in the frequency domain signal to obtain the beam output for each frequency point; the beam output is scanned on the horizontal wavenumber to obtain the wavenumber spectrum; the horizontal wavenumber is determined by the wavenumber corresponding to the average sound velocity on the transect and the wavenumber at the seabed. Equal-interval sampling is performed between wavenumbers corresponding to low sound speeds, with N samples. The wavenumber spectrum is transformed to obtain a two-dimensional distribution matrix of horizontal wavenumber difference-frequency. The sampling accuracy of the two-dimensional distribution matrix of horizontal wavenumber difference-frequency is optimized through a two-layer game model. For each sampling point in the horizontal wavenumber difference domain, energy accumulation is performed along the full bandwidth of the signal to construct a cumulative energy-horizontal wavenumber difference curve. The coordinates corresponding to the peak value of the cumulative energy-horizontal wavenumber difference curve are calculated and compared with the criteria to complete the depth discrimination. When the coordinates corresponding to the peak value are less than the criteria, it is determined to be a surface sound source. When the coordinates corresponding to the peak value are greater than the criteria, it is determined to be an underwater sound source.

[0005] Specifically, the wavenumber domain beamforming is a signal processing method based on normal mode theory, located at depth... and distance from the origin The sound pressure generated by a point sound source at depth z and distance r is calculated by summing the normal mode function and the horizontal wavenumber.

[0006] Specifically, the output of the wavenumber domain beamforming is obtained by integrating the received data along the array aperture and introducing a steering coefficient and a compensation coefficient. The compensation coefficient is used to compensate for the energy attenuation during the propagation of the sound wave.

[0007] Specifically, the wavenumber spectrum is obtained by taking the modulus of the beam output corresponding to each horizontal wavenumber to obtain the energy response, scanning the horizontal wavenumber to obtain the wavenumber spectrum, and sampling the horizontal wavenumber at equal intervals between the wavenumber corresponding to the average sound speed on the stratum and the wavenumber corresponding to the minimum sound speed on the seabed.

[0008] Specifically, the horizontal wavenumber-frequency two-dimensional distribution matrix is ​​obtained by selecting all array element sound pressure signals at each frequency point within the spectrum from the start frequency to the cutoff frequency, performing wavenumber domain beamforming, and obtaining energy distribution matrices in the frequency domain and horizontal wavenumber domain.

[0009] Specifically, the coordinate transformation is calculated by subtracting the wave number corresponding to the frequency from the horizontal wave number. The wave number corresponding to the frequency is determined by dividing the frequency by the average speed of sound in water and then multiplying by twice the value of pi.

[0010] Specifically, the cumulative energy-horizontal wavenumber difference curve is obtained by integrating the energy corresponding to each sampling point in the horizontal wavenumber difference domain along the full bandwidth from the starting frequency to the cutoff frequency.

[0011] The two-layer game model includes an upper-layer model aimed at optimizing sampling accuracy and a lower-layer model aimed at improving energy resolution. The upper-layer model and the lower-layer model are coupled and optimized through the energy resolution index and the sampling standard deviation.

[0012] Specifically, the upper-level model takes the sampling standard deviation, signal bandwidth, and energy resolution indices of the horizontal wavenumber difference domain as inputs, and outputs sampling accuracy optimization indices to adjust the number of samples and optimize sampling accuracy.

[0013] Specifically, the lower-level model takes the peak energy, peak half-width at half-maximum, and sampling standard deviation of the horizontal wavenumber difference domain of the cumulative energy-horizontal wavenumber difference curve as input, and outputs the energy resolution index as the input parameters of the upper-level model.

[0014] Specifically, the constraints of the upper-level model are that the number of samples is between 50 and 2000, and the standard deviation of the sampling in the horizontal wavenumber difference domain is 0.01. Up to 0.2 .

[0015] Specifically, the constraints of the lower-level model are that the peak energy of the cumulative energy-horizontal wavenumber difference curve is greater than or equal to 0.5 times the total energy, and the peak half-width at half-maximum (FWHM) satisfies 0.05. Up to 0.5 .

[0016] Specifically, the criterion is based on the wavenumber difference between the average sound velocity on the stratum and the average sound velocity in the water, and is calculated by weighted averaging the spectral distribution of the broadband sound pressure signal to obtain the critical threshold for distinguishing between surface sound sources and underwater sound sources.

[0017] Specifically, the coordinates corresponding to the peak value are the horizontal wavenumber differences corresponding to the maximum energy value in the cumulative energy-horizontal wavenumber difference curve.

[0018] The physical mechanism of depth discrimination specifically utilizes the difference between the energy dispersion of high-order normal modes excited by a surface sound source and the energy coupling of low-order normal modes excited by an underwater sound source. The determination of shallow and deep sources is achieved by analyzing the peak position distribution of the cumulative energy-horizontal wavenumber difference curve.

[0019] Wherein, the aperture L of the horizontal linear array is the total length of the horizontal linear array, and the initial position of the array is the distance from the first reference element. This is the distance from the sound source to the first element of the horizontal array.

[0020] This invention employs a depth discrimination method based on horizontal wavenumber difference domain feature extraction. It transforms the frequency-wavenumber domain distribution into a horizontal wavenumber difference-frequency domain distribution through coordinate transformation and utilizes a two-layer game theory model to collaboratively optimize sampling accuracy and energy resolution, thus solving the technical problem of insufficient accuracy in sound source depth discrimination in shallow-sea negative-gradient waveguides. The invention optimizes the sampling standard deviation and number of sampling points in the horizontal wavenumber difference domain through an upper-layer model, while the lower-layer model improves the peak energy and peak width characteristics of the cumulative energy curve. The two models achieve coupling feedback through energy resolution indicators and sampling standard deviation, avoiding the mutual constraint between sampling accuracy and energy resolution. The horizontal wavenumber difference curve constructed along the full bandwidth of the cumulative energy clearly distinguishes the high-order mode energy concentration characteristics of surface sound sources from the low-order mode energy coupling characteristics of underwater sound sources. Comparison of the peak position with the criteria achieves reliable depth discrimination. In summary, this invention solves the technical problem of insufficient accuracy in sound source depth discrimination in shallow-sea negative-gradient waveguides mentioned in the background art. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the main process of an embodiment.

[0022] Figure 2 This is a schematic diagram showing the positional relationship of the depth discrimination method under the negative gradient waveguide condition proposed in this paper.

[0023] Figure 3 This is a detailed process flowchart of the present invention. Figure 4 This is a schematic diagram of the sound velocity profile used in the simulation of this invention.

[0024] Figure 5 The diagram shows the two-dimensional distribution matrix of horizontal wavenumber difference-frequency of a water surface sound source and the cumulative energy-horizontal wavenumber difference curve after processing by this invention.

[0025] Figure 6 The diagram shows the two-dimensional distribution matrix of horizontal wavenumber difference-frequency of an underwater sound source and the cumulative energy-horizontal wavenumber difference curve after processing by this invention.

[0026] Figure 7 A comparison chart of the peak value and criteria of the cumulative energy-horizontal wavenumber difference curve obtained from the depth simulation of the sound source. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0028] This invention provides a target depth discrimination method based on wavenumber difference domain feature extraction of a negative gradient waveguide horizontal array, comprising the following steps:

[0029] S01. A horizontal linear array deployed on the seabed is used to receive broadband acoustic pressure signals emitted by a sound source in a shallow-sea negative gradient waveguide. The aperture of the horizontal linear array is L, and the initial position of the array is at a distance of [missing information]. The starting frequency of the broadband sound pressure signal is The cutoff frequency is ;

[0030] S02. Perform Fourier transform on the received broadband sound pressure signal to obtain the frequency domain signal, and perform wavenumber domain beamforming on each frequency point in the frequency domain signal from the start frequency to the cutoff frequency to obtain the beam output of each frequency point.

[0031] S03. The beam output is scanned on the horizontal wavenumber to obtain the wavenumber spectrum. The horizontal wavenumber is sampled at equal intervals between the wavenumber corresponding to the average sound speed on the stratum and the wavenumber corresponding to the minimum sound speed on the seabed. The number of samples is N.

[0032] S04. Perform coordinate transformation on the wavenumber spectrum to obtain the horizontal wavenumber difference-frequency two-dimensional distribution matrix, and optimize the sampling accuracy of the horizontal wavenumber difference-frequency two-dimensional distribution matrix through a two-level game model.

[0033] S05. For each sampling point in the horizontal wavenumber difference domain, perform energy accumulation along the full bandwidth of the signal to construct the cumulative energy-horizontal wavenumber difference curve;

[0034] S06. Calculate the coordinates corresponding to the peak value of the cumulative energy-horizontal wavenumber difference curve and compare them with the criteria to complete the depth discrimination. When the coordinates corresponding to the peak value are less than the criteria, it is determined to be a surface sound source. When the coordinates corresponding to the peak value are greater than the criteria, it is determined to be an underwater sound source.

[0035] Wherein, the aperture L of the horizontal linear array is the total length of the horizontal linear array, in meters, and the initial position of the array is the distance from the first reference element. The distance from the sound source to the first element of the horizontal linear array, in meters. The starting frequency. The cutoff frequency is the lowest frequency of the broadband sound pressure signal, measured in Hz. The highest frequency of the broadband sound pressure signal is expressed in Hz.

[0036] The wavenumber domain beamforming described herein is a signal processing method based on normal mode theory. Located at depth and distance from the origin The sound pressure produced by a point sound source at depth z and distance r is expressed as follows: ,in , and These are the mode function and attenuation coefficient corresponding to the horizontal wavenumber of the m-th mode, respectively, where M is the total order of the excitation modes of the sound source. The depth... The depth z is the water depth of the sound source, in meters (m). The distance r is the horizontal distance from the receiving point to the sound source, in meters. The horizontal wavenumber... The horizontal wavenumber of the m-th mode is given by the given value. The wave number k is the wave number of the sound wave in water, and its unit is 1000 kJ / m². The glancing angle Let be the angle between the sound ray of the m-th mode and the horizontal plane, in rad. The mode function... Let m be the normalized amplitude distribution function of the m-th mode, and let the attenuation coefficient be... The energy decay coefficient of the m-th mode is given by . .

[0037] The output expression for the wavenumber domain beamforming is: ,in , It is the guidance coefficient. It is a coefficient used to compensate for propagation loss in the received data. Confirmed. The distance. The guiding coefficient is the distance from the end element of the horizontal linear array to the sound source, in meters. The compensation coefficient is used to adjust the phase of the received signal of each array element to achieve beamforming. This is used to compensate for the energy attenuation of sound waves during propagation. Under the far-field assumption, the compensation coefficient... Approximately proportional to Substituting it into the above formula, we get... ,in The coefficient denoted as the beamforming response coefficient corresponding to the m-th mode.

[0038] Specifically, the wavenumber spectrum is calculated by taking the square of the modulus of the beam output corresponding to each level wavenumber to obtain the response. Then, in the horizontal wavenumber The wavenumber spectrum was obtained by scanning the upper part. The response horizontal wavenumber The corresponding energy response, in units of The horizontal wavenumber exist arrive Equal-interval sampling is performed within the range, with a sample size of N, where It is the average speed of sound on the mezzanine, measured in m / s. It is the minimum speed of sound on the seabed, measured in m / s. The average speed of sound on the aforementioned mezzanine. The average speed of sound in the water layer above the mezzanine is the minimum speed of sound on the seabed. The minimum speed of sound in the seabed medium. Select the starting frequency. to cutoff frequency The sound pressure signals of all array elements at each frequency point within the spectrum are processed in the same way to obtain a two-dimensional distribution matrix of horizontal wavenumber and frequency. ,in It is the frequency at the starting frequency to cutoff frequency The number of equally spaced sampling points within the range. The horizontal wavenumber-frequency two-dimensional distribution matrix. This represents the energy distribution matrix in the frequency domain and the horizontal wavenumber domain.

[0039] Wherein, the horizontal wavenumber difference-frequency two-dimensional distribution matrix is ​​denoted as It is obtained by performing a coordinate transformation on the horizontal wavenumber-frequency two-dimensional distribution matrix. The formula for calculating the coordinate transformation is as follows: ,in Horizontal wavenumber, in units of , The wave number corresponding to the frequency, in units of , ,in It is the average speed of sound in water, measured in m / s. The horizontal wavenumber difference... The difference between the horizontal wavenumber and the corresponding frequency wavenumber, expressed in units of... This is used to characterize the wavenumber differences between modes. The average sound velocity in water... Let be the average sound velocity across the entire water column. After coordinate transformation, the cumulative energy-horizontal wavenumber difference curve is constructed by accumulating energy along the full signal bandwidth at each sampling point in the horizontal wavenumber difference domain. ,in The accumulated energy Horizontal wavenumber difference The corresponding total energy across the entire frequency band, in units of .

[0040] The two-layer game model comprises an upper-layer model aimed at optimizing sampling accuracy and a lower-layer model aimed at improving energy resolution. The objective function of the upper-layer model is: ,in The sampling standard deviation of the horizontal wavenumber difference domain, in units of , The reference sampling point count is set to 1000. The bandwidth influence factor is set to 0.02. Signal bandwidth, measured in Hz. The reference bandwidth is set at 100Hz. The energy coupling coefficient is set to 0.3. The energy resolution index is output from the lower-level model. The sampling standard deviation of the horizontal wavenumber difference domain... The number of reference sampling points characterizes the dispersion of the horizontal wavenumber difference domain sampling points. The bandwidth influence coefficient serves as a reference value for the standardized number of sampling points. The signal bandwidth is used to adjust the degree of influence of signal bandwidth on sampling accuracy. Starting frequency With cutoff frequency The difference, the reference bandwidth The energy coupling coefficient is a reference value for standardized bandwidth. The energy resolution index refers to the weighting coefficients that couple the upper and lower layer models. This is an evaluation index for the energy resolution capability output by the lower-level model.

[0041] Wherein, the objective function of the lower-level model is ,in The peak energy of the cumulative energy-level wavenumber difference curve, in units of , The reference energy is set to 1. , The peak width influence coefficient is set to 0.5. The peak half-width is 1 / 250 m / s, in units of 1 / 250 m / s. , For reference, the half-width and half-height value is set to 0.1. , The sampling coupling coefficient is set to 0.4. The reference sampling standard deviation is set at 0.05. The peak energy of the cumulative energy-level wavenumber difference curve. The reference energy is the maximum energy value in the cumulative energy-level wavenumber difference curve. The peak width influence coefficient serves as a reference value for standardized energy. The peak half-width is used to adjust the degree of influence of peak width on energy resolution. The reference half-width is the horizontal wavenumber difference width corresponding to half the peak energy in the cumulative energy-horizontal wavenumber difference curve. The sampling coupling coefficient is a reference value for the standardized full width at half maximum (FWHM). The reference sampling standard deviation represents the weighting coefficients that couple the lower-level model with the upper-level model. This is a reference value for the standardized sampling standard deviation. The energy coupling coefficient... and sampling coupling coefficient represents the coupling term coefficient of the two-layer game model.

[0042] The upper-level model is constrained by the number of samples N being between 50 and 2000, and the sampling standard deviation of the horizontal wavenumber difference domain being within the range of [missing information]. Satisfying 0.01 Up to 0.2 The constraint condition for the lower-level model is the peak energy of the cumulative energy-level wavenumber difference curve. Total energy greater than or equal to 0.5 times the cumulative energy-horizontal wavenumber difference curve And peak half-width Satisfying 0.05 Up to 0.5 ,in The total energy of the cumulative energy-level wavenumber difference curve, in units of The total energy of the cumulative energy-level wavenumber difference curve. It is the sum of the energy of all sampling points on the cumulative energy-level wavenumber difference curve.

[0043] The objective function of the upper-level model is used to optimize the sampling accuracy of the horizontal wavenumber difference domain to improve the robustness of depth discrimination. The input includes the sampling standard deviation of the horizontal wavenumber difference domain. Signal bandwidth and energy resolution index The output is the sampling accuracy optimization index. The sampling standard deviation of the horizontal wavenumber difference domain. The signal bandwidth is calculated from the sampling interval of the horizontal wavenumber difference domain in step S04. Cutoff frequency With the starting frequency The difference, the energy resolution index Equal to the objective function of the lower-level model The output value, the sampling accuracy optimization index This is used to adjust the number of samples N in step S03 to optimize sampling accuracy.

[0044] The objective function of the lower-level model is used to improve the energy resolution of the cumulative energy-horizontal wavenumber difference curve to enhance peak identification capability. The input includes the peak energy of the cumulative energy-horizontal wavenumber difference curve. Peak half-width Sampling standard deviation of the horizontal wavenumber difference domain The output is the energy resolution index. The peak energy of the cumulative energy-level wavenumber difference curve. The maximum value in the cumulative energy-level wavenumber difference curve constructed in step S05 is determined by the peak half-width. The energy resolution index is the width of the horizontal wavenumber difference corresponding to half the peak energy in the cumulative energy-horizontal wavenumber difference curve. The input parameters of the objective function of the upper-level model are used to achieve coupling optimization of the two-level game model.

[0045] The calculation formula for the criterion is as follows: , where the starting frequency The start frequency of the broadband sound pressure signal is in Hz, and the cutoff frequency is... This is the cutoff frequency of the broadband sound pressure signal, measured in Hz. The spectrum of a broadband sound pressure signal, and the average sound velocity on the lattice. The unit is m / s, the average speed of sound in water. The unit is m / s. The criterion Q is the critical threshold for distinguishing between surface and underwater sound sources, and its unit is m / s. The spectrum of the broadband sound pressure signal Let f be the signal energy distribution function corresponding to frequency f. The coordinates corresponding to the peak value are: The unit is The cumulative energy-level wavenumber difference curve constructed by step S05 The horizontal wavenumber difference corresponding to the maximum value is determined. The coordinates corresponding to the peak value are... The horizontal wavenumber difference is the difference corresponding to the maximum energy value on the cumulative energy-horizontal wavenumber difference curve.

[0046] The physical mechanism of depth discrimination is as follows: high-order normal modes excited by surface sound sources dominate energy. As the frequency increases, the total modal order increases, and the signal energy is concentrated in the higher orders. In contrast, low-order normal modes excited by underwater sound sources have high energy, and the energy of the higher-order modes exhibits oscillation. As the frequency increases, the modal order increases, and the horizontal wavenumber difference between adjacent modes gradually increases. The energy coupling of low-order, high-energy modes results in a larger energy level near zero horizontal wavenumber difference. Binary discrimination between shallow and deep sources is achieved through the difference in peak distribution of the cumulative energy-horizontal wavenumber difference curve. Shallow sources refer to surface sound sources near the sea surface, and deep sources refer to underwater sound sources below the sea level. The higher-order normal modes are those with higher modal orders, and the lower-order normal modes are those with lower modal orders. The modal order is the sequence number of the normal mode.

[0047] Optionally, the present invention also provides a method for implementing a target depth discrimination system based on wavenumber difference domain feature extraction of a negative gradient waveguide horizontal array by means of a computer. The computer is provided with a readable storage medium, which stores program instructions. When the program instructions are run in the computer, they execute the above-described method.

[0048] It should be noted that this invention also solves the following technical problem: In the depth discrimination of waveguide sound sources in shallow sea negative gradient layers, traditional methods struggle to obtain sufficient energy resolution under limited bandwidth conditions, leading to indistinct peak features and affecting the reliability of discrimination. This invention constructs a two-dimensional distribution matrix of horizontal wavenumber difference and frequency and accumulates energy along the entire bandwidth, mapping the frequency domain energy information to the horizontal wavenumber difference domain to form an accumulated energy curve. This amplifies the peak position difference between shallow and deep sources. Simultaneously, it utilizes the lower-level objective function of a two-layer game model to optimize the peak energy and peak half-width. Under the constraint that the peak energy is not less than half of the total energy, the contribution of the peak width to energy resolution is adjusted by the peak width influence coefficient, achieving effective aggregation of energy features under limited bandwidth. Furthermore, this invention solves the technical problem of existing depth discrimination methods lacking physical basis for their criteria. By constructing a criterion calculation formula based on the wavenumber difference between the average sound velocity on the gradient layer and the average sound velocity in the water, and combining it with the spectral distribution of the broadband sound pressure signal for weighted averaging, the criterion can reflect the physical influence of the gradient layer on sound propagation, improving the physical rationality of depth discrimination.

[0049] Specifically, the principle of this invention is as follows: This invention can solve the technical problem of insufficient accuracy in sound source depth discrimination in shallow sea negative gradient waveguides. The fundamental reason is that it fully utilizes the distribution law of horizontal wavenumber difference between normal modes excited by sound sources at different depths. High-order normal modes excited by surface sound sources dominate the energy distribution. As the frequency increases, the mode order increases, leading to a larger horizontal wavenumber difference between adjacent modes, and the energy is dispersed in a larger horizontal wavenumber difference region. Low-order normal modes excited by underwater sound sources have concentrated energy, and the energy coupling of low-order high-energy modes causes energy to accumulate near zero horizontal wavenumber difference. By transforming the frequency-wavenumber domain into the horizontal wavenumber difference domain through coordinate transformation, the linear influence of frequency changes on wavenumber is eliminated, highlighting the intrinsic wavenumber differences between modes. The two-layer game model ensures a reasonable distribution of sampling points in the horizontal wavenumber difference domain by optimizing sampling accuracy in the upper layer, and enhances peak recognition capability by optimizing energy resolution in the lower layer. The two layers work together through coupling terms to avoid local optima caused by single optimization. Accumulated energy along the entire bandwidth amplifies the energy distribution difference between shallow and deep sources in the horizontal wavenumber difference domain, and the peak position is consistent with the physical laws, thus achieving accurate discrimination.

[0050] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0051] To verify the effectiveness of this invention, technicians constructed a test environment and conducted depth discrimination tests on sound sources at different depths in a shallow sea negative-gradient waveguide environment. The test environment was a typical shallow sea negative-gradient waveguide area with a water depth of 100m, a gradient depth of 20m, and a muddy sedimentary seabed. Technicians deployed a horizontal linear array on the seabed with an array aperture L of 200m, containing 64 hydrophone units with an element spacing of 3.125m, at a deployment depth of 95m. The test sound source used a broadband pulse signal with a starting frequency of... 100Hz, cutoff frequency The frequency is 500Hz, and the pulse width is 2s. The distance from the sound source to the first reference element of the horizontal linear array. It is 5000m.

[0052] During the test, technicians first obtained sound speed profile data for the test sea area. For example... Figure 4 As shown, the sound velocity profile exhibits a typical negative cascade structure. The sound velocity in the surface water is 1520 m / s, forming a cascade at a depth of 20 m. Below the cascade, the sound velocity gradually decreases to 1480 m / s at the seabed. The average sound velocity on the cascade is calculated based on the sound velocity profile. The average sound velocity across the entire water layer is 1518 m / s. The lowest speed of sound on the seabed is 1495 m / s. The speed is 1480 m / s. This negative gradient layer structure forms a surface sound channel, which has a significant impact on sound wave propagation. The distribution of normal modes excited by sound sources at different depths is significantly different.

[0053] The test was conducted in two scenarios. The first scenario involved a sound source at a depth of 5m, simulating a surface sound source. The second scenario involved a sound source at a depth of 50m, simulating an underwater sound source. Both scenarios emitted the same broadband pulse signal, which was received and processed by a horizontal line array. Figure 2 As shown in the figure, this diagram illustrates the positional relationship of the depth discrimination method under negative cascade waveguide conditions, clearly showing the spatial relationship between the sound source, the cascade, the horizontal array, and the seabed.

[0054] Technicians first performed a Fourier transform on the received broadband sound pressure signal to obtain the frequency domain signal. For each frequency point within the range of a starting frequency of 50Hz to a cutoff frequency of 500Hz (preferably 100Hz to 300Hz), wavenumber domain beamforming was used for processing. Wavenumber domain beamforming is based on normal mode theory, using the modal decomposition principle to represent the sound field as a superposition of normal modes of various orders. During the horizontal wavenumber scan, the scanning range was set between the wavenumber corresponding to the average sound velocity on the superposition layer and the wavenumber corresponding to the minimum sound velocity on the seabed, i.e., from... arrive The initial number of samples N was set to 500, and was subsequently optimized using a two-layer game model.

[0055] For a 100Hz frequency point, the horizontal wavenumber scan range is 0.414 rad / m to 0.425 rad / m. For a 500Hz frequency point, the horizontal wavenumber scan range is 2.071 rad / m to 2.126 rad / m. The response is obtained by squaring the modulus of the beam output at each frequency point, and the wavenumber spectrum is obtained by scanning the horizontal wavenumber. The same processing is performed on each frequency point within the spectrum from the start frequency to the cutoff frequency, and the number of frequency sampling points is [not specified]. Setting the value to 200 yields a two-dimensional horizontal wavenumber-frequency distribution matrix. The rows of this matrix correspond to the horizontal wavenumber dimension, the columns to the frequency dimension, and the element values ​​represent the energy response at the corresponding horizontal wavenumber and frequency.

[0056] like Figure 3 As shown in the figure, this diagram illustrates the detailed processing flowchart of the present invention, including complete processing steps such as signal reception, Fourier transform, wavenumber domain beamforming, coordinate transformation, two-layer game optimization, and depth discrimination. Technicians process the data step-by-step according to the flowchart's instructions.

[0057] Next, a coordinate transformation is performed to convert the horizontal wavenumber-frequency two-dimensional distribution matrix into a horizontal wavenumber difference-frequency two-dimensional distribution matrix. The coordinate transformation formula is the horizontal wavenumber minus the wavenumber corresponding to the frequency, i.e. The row coordinates of the transformed matrix represent the horizontal wavenumber difference, while the column coordinates remain the frequency. The physical significance of this coordinate transformation lies in normalizing the modal wavenumbers at different frequencies to a unified reference frame, which facilitates the extraction of features related to the sound source depth.

[0058] To optimize the sampling accuracy of the horizontal wavenumber difference-frequency two-dimensional distribution matrix, engineers applied a two-layer game theory model. The upper-layer model aims to optimize sampling accuracy, with the objective function comprehensively considering the effects of sampling standard deviation, signal bandwidth, and energy resolution. The initial number of samples (500) corresponds to a sampling standard deviation of 0.068 rad / m and a signal bandwidth of 400 Hz. The lower-layer model aims to improve energy resolution, with the objective function considering the combined effects of peak energy, peak half-width at half-maximum (FWHM), and sampling standard deviation.

[0059] The two-level game model was solved using an iterative optimization method. The technicians set the iteration termination condition as follows: the relative change in the objective functions of both the upper and lower levels was less than 0.001, or the number of iterations reached 50. After 15 iterations, the model converged. The optimized number of samples, N, was adjusted to 680, corresponding to a sampling standard deviation of 0.052 rad / m. The optimized parameters ensured sufficient resolution in the horizontal wavenumber difference domain sampling while avoiding increased computational load due to oversampling.

[0060] For water surface sound source scenarios, technicians reconstructed the horizontal wavenumber difference-frequency two-dimensional distribution matrix using optimized sampling parameters. For example... Figure 5 As shown in the figure, this diagram illustrates the two-dimensional distribution matrix of the horizontal wavenumber difference versus frequency of a surface acoustic source and the cumulative energy versus horizontal wavenumber difference curve after processing according to this invention. The two-dimensional distribution matrix reveals that the energy is primarily concentrated in the region with a large horizontal wavenumber difference. This is because the higher-order normal modes excited by the surface acoustic source dominate the energy distribution. With increasing frequency, the total modal order increases, and the energy of the higher-order modes further strengthens.

[0061] Technicians accumulated energy along the entire signal bandwidth at each sampling point in the horizontal wavenumber difference domain. The accumulated energy was calculated by integrating the energy at all frequency points corresponding to each horizontal wavenumber difference. For a surface sound source, the accumulated energy-horizontal wavenumber difference curve exhibits a single-peak distribution, with the peak position corresponding to the horizontal wavenumber difference. It is 0.187 rad / m. Peak energy. It is 8.56 Peak half-width The value is 0.132 rad / m. The total energy of the curve is... It is 14.23 The peak energy accounts for 60.2% of the total energy, which satisfies the constraints of the lower-level model.

[0062] For underwater sound source scenarios, technicians use the same processing procedure for analysis. For example... Figure 6 As shown in the figure, this diagram illustrates the two-dimensional distribution matrix of the horizontal wavenumber difference-frequency of the underwater sound source and the cumulative energy-horizontal wavenumber difference curve after processing according to this invention. The two-dimensional distribution matrix reveals a significant difference in energy distribution compared to surface sound sources. The low-order normal modes excited by the underwater sound source exhibit high energy, while the high-order modes display oscillatory energy. Energy coupling between low-order, high-energy modes results in a region with relatively high energy where the horizontal wavenumber difference approaches zero.

[0063] The peak position of the cumulative energy-horizontal wavenumber difference curve of an underwater sound source corresponds to the horizontal wavenumber difference. The peak energy is 0.023 rad / m, significantly lower than the peak energy of a surface sound source. It is 11.34 Peak half-width The value is 0.089 rad / m. The total energy of the curve is... It is 18.67 The peak energy accounts for 60.7% of the total energy. The peak energy of underwater sound sources is sharper, and the energy is more concentrated in the region where the horizontal wavenumber difference is close to zero.

[0064] Technicians determine the critical threshold for depth discrimination based on the criterion calculation formula. The calculation of criterion Q requires the spectrum of the broadband sound pressure signal. The spectrum is obtained by taking the square of the modulus of the frequency domain representation of the received signal. For the broadband pulse signal tested, the spectrum is relatively uniformly distributed in the range of 100Hz to 500Hz. Substituting the spectrum, the average sound velocity on the stratum, and the average sound velocity in the water into the criterion formula, the criterion Q is calculated to be 0.096 rad / m through numerical integration. This criterion reflects the critical horizontal wavenumber difference that distinguishes surface sound sources from underwater sound sources under the current environmental conditions and signal parameters.

[0065] After calculating the criteria, the technicians compared the peak coordinates of the cumulative energy-horizontal wavenumber difference curve with the criteria. For the surface sound source scenario, the peak coordinates of 0.187 rad / m were greater than the criterion of 0.096 rad / m (the peak coordinates were calculated using absolute values), therefore it was determined to be a surface sound source, consistent with the actual set sound source depth of 5m. For the underwater sound source scenario, the peak coordinates of 0.023 rad / m were less than the criterion of 0.096 rad / m, therefore it was determined to be an underwater sound source, consistent with the actual set sound source depth of 50m. The discrimination results for both scenarios were correct, verifying the effectiveness of the method of this invention.

[0066] To further test the robustness of the method, the technicians conducted a simulation experiment traversing the depth of the sound source. The sound source depth ranged from 1m to 100m, traversing at 1m intervals, for a total of 40 depth points. The same signal parameters and processing flow were used for each depth point to obtain the corresponding cumulative energy-level wavenumber difference curve, and the coordinates corresponding to the peak value were extracted. For example... Figure 7 As shown in the figure, this diagram compares the peak values ​​of the cumulative energy-horizontal wavenumber difference curves obtained from traversing the sound source depth simulation with the criterion. It is clearly seen from the figure that when the sound source depth is less than 20m below the mezzanine depth, the coordinates corresponding to the peak values ​​are all greater than the criterion, correctly identifying it as a surface sound source. When the sound source depth is greater than the mezzanine depth, the coordinates corresponding to the peak values ​​are all less than the criterion, correctly identifying it as an underwater sound source. Near the mezzanine, there is a transition region where the coordinates corresponding to the peak values ​​are close to the criterion, yet correct identification is still possible.

[0067] like Figure 1 As shown in the figure, the main process of the present invention includes key steps such as signal reception, frequency domain conversion, beamforming, wavenumber spectrum extraction, coordinate transformation, two-layer game optimization, energy accumulation, and depth discrimination, which constitute a complete processing link.

[0068] Technicians also analyzed the impact of different signal bandwidths on depth discrimination performance. Keeping the initial frequency constant at 100Hz, the cutoff frequencies were set to 300Hz, 400Hz, and 500Hz, corresponding to signal bandwidths of 200Hz, 300Hz, and 400Hz, respectively. Test results showed that as the signal bandwidth increased, the peak energy of the cumulative energy-horizontal wavenumber difference curve increased, the peak half-width at half-maximum (FWHM) decreased, and the reliability of depth discrimination improved. This is because a wider signal bandwidth contains more frequency components, enabling the excitation of more order normal modes and providing richer depth information. The bandwidth influence coefficient in the two-layer game model effectively modulates the impact of bandwidth on sampling accuracy, ensuring good discrimination performance under different bandwidth conditions.

[0069] Technicians further tested the impact of array aperture on depth discrimination. Keeping other parameters constant, the array aperture was set to 100m, 150m, and 200m. The test results showed that a larger array aperture resulted in higher beamforming resolution, a clearer energy distribution in the horizontal wavenumber difference-frequency two-dimensional distribution matrix, and a sharper peak in the cumulative energy-horizontal wavenumber difference curve. At an array aperture of 200m, the difference in the coordinates corresponding to the peak values ​​of the surface and underwater sound sources was most significant, resulting in the highest confidence level for depth discrimination. This indicates that array aperture is a crucial factor affecting depth discrimination performance, and in practical applications, an appropriate array aperture should be selected based on the detection distance and environmental conditions.

[0070] During data processing, technicians recorded and analyzed key parameters. As shown in Table 1, this table lists the characteristic parameters of the cumulative energy-horizontal wavenumber difference curves at different sound source depths.

[0071] Table 1 Comparison of characteristic parameters for different sound source depths

[0072]

[0073] In this context, the coordinates corresponding to the peak values ​​all take absolute values ​​into account.

[0074] As shown in Table 1, with increasing sound source depth, the peak coordinates gradually decrease, the peak energy gradually increases, and the peak half-width at half-maximum (HWHM) gradually decreases. When the sound source moves from the water surface to underwater and passes through the mezzanine, the peak coordinates undergo a significant change, shifting from being greater than the criterion to being less than the criterion, thus effectively distinguishing between surface and underwater sound sources.

[0075] The technical staff also conducted a quantitative analysis of the optimization effect of the two-level game model. As shown in Table 2, this table compares the changes in key parameters before and after optimization.

[0076] Table 2. Comparison of parameters before and after optimization of the two-layer game model.

[0077]

[0078] As shown in Table 2, after optimization using the two-layer game model, the number of samples increased moderately, and the sampling standard deviation decreased significantly, indicating improved sampling accuracy. Simultaneously, the peak energy increased significantly, and the peak half-width at half-maximum (FWHM) decreased markedly, indicating an effective improvement in energy resolution. The optimized parameter configuration made the peak value of the cumulative energy-level wavenumber difference curve more prominent, enhancing both the accuracy and robustness of depth discrimination.

[0079] The main advancements of this invention compared to traditional depth discrimination methods are reflected in the following aspects. Traditional methods typically determine depth based on the time-domain or frequency-domain characteristics of the received signal, such as using multipath arrival delay difference or spectral characteristics for analysis. These methods face two main difficulties in negative-step waveguide environments. First, the negative-step structure leads to complex ray curvature, making it difficult to accurately extract the multipath arrival delay difference. Second, the differences in modal distribution excited by sound sources at different depths are not sufficiently reflected in the time-domain and frequency-domain characteristics, resulting in low discrimination reliability. This invention introduces the concept of a horizontal wavenumber difference domain, using the wavenumber difference of normal mode as the core feature for depth discrimination. This feature extraction method directly reflects the influence of sound source depth on modal distribution, with a clear physical mechanism. Surface sound sources mainly excite higher-order normal modes, and the horizontal wavenumber and frequency-corresponding wavenumber of these modes differ significantly. Underwater sound sources mainly excite lower-order normal modes, and the energy coupling between lower-order modes causes energy concentration in regions where the horizontal wavenumber difference is close to zero. Depth discrimination can be achieved by accumulating the peak position of the energy-horizontal wavenumber difference curve, making the method simple and effective.

[0080] The introduction of a two-layer game theory model is another important innovation of this invention. Traditional methods for processing wavenumber domain data typically set sampling parameters based on experience, lacking theoretical basis. This invention treats sampling accuracy optimization and energy resolution improvement as two interrelated objectives, achieving synergistic optimization through a two-layer game theory model. The upper-layer model focuses on sampling accuracy, ensuring that sampling in the horizontal wavenumber difference domain accurately captures modal characteristics. The lower-layer model focuses on energy resolution, ensuring that the peak value of the cumulative energy-horizontal wavenumber difference curve is sufficiently prominent. The two models influence each other through coupling terms, iteratively optimizing until convergence. This optimization strategy allows sampling parameters to adaptively adjust to different environmental conditions and signal parameters, improving the applicability and robustness of the method.

[0081] The criterion design fully considers the influence of environmental parameters and signal characteristics. The criterion is not a fixed value, but rather dynamically calculated based on the average sound velocity in the mezzanine, the average sound velocity in the water, and the signal spectrum. This adaptive criterion design allows the depth discrimination method to adapt to different marine environments and signal conditions. When the mezzanine depth changes, the sound velocity profile changes, or the signal frequency band is adjusted, the criterion will adjust accordingly to maintain the accuracy of the discrimination. Traditional methods often have fixed discrimination thresholds or require manual adjustment, lacking flexibility. The adaptive criterion design of this invention significantly improves the practicality of the method.

[0082] This invention utilizes a horizontal linear array to receive data, offering advantages in deployment and maintenance compared to vertical arrays. Horizontal arrays can be deployed on the seabed, avoiding the complexity of anchoring or suspending vertical arrays. Furthermore, the aperture of a horizontal array can be made larger, providing higher azimuth and wavenumber resolution. These advantages make this invention more feasible in practical applications. While traditional depth discrimination methods based on vertical arrays can utilize depth dimension information, their effectiveness is limited by the layer barrier in negative layer environments. This invention utilizes a horizontal array to extract wavenumber domain features, avoiding the influence of layer barrier and making it more suitable for negative layer waveguide environments.

[0083] It should be noted that the variables involved in this invention are explained in detail in Table 3.

[0084] Table 3. Variable Explanation Table

[0085]

[0086] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A target depth discrimination method based on wavenumber difference domain feature extraction of a negative gradient waveguide horizontal array, characterized in that, Includes the following steps: A horizontal linear array deployed on the seabed is used to receive broadband acoustic pressure signals emitted by sound sources in a shallow sea negative gradient waveguide. The received broadband acoustic pressure signals are Fourier transformed to obtain frequency domain signals. Wavenumber domain beamforming is performed on each frequency point within the range from the start frequency to the cutoff frequency in the frequency domain signal to obtain the beam output of each frequency point. The beam output is scanned on the horizontal wavenumber to obtain the wavenumber spectrum. The horizontal wavenumber is sampled at equal intervals between the wavenumber corresponding to the average sound velocity on the gradient and the wavenumber corresponding to the minimum sound velocity on the seabed, with a sampling number of N. The wavenumber spectrum is transformed to obtain a two-dimensional distribution matrix of horizontal wavenumber difference-frequency. The sampling accuracy of the two-dimensional distribution matrix of horizontal wavenumber difference-frequency is optimized through a two-level game model. For each sampling point in the horizontal wavenumber difference domain, energy accumulation is performed along the full bandwidth of the signal to construct an accumulated energy-horizontal wavenumber difference curve. The coordinates corresponding to the peak of the accumulated energy-horizontal wavenumber difference curve are calculated and compared with the criteria to complete the depth discrimination. When the coordinates corresponding to the peak are less than the criteria, it is determined to be a surface sound source, and when the coordinates corresponding to the peak are greater than the criteria, it is determined to be an underwater sound source.

2. The method according to claim 1, characterized in that, The wavenumber domain beamforming, specifically a signal processing method based on normal mode theory, is located in the depth domain. and distance from the origin The sound pressure generated by a point sound source at depth z and distance r is calculated by summing the normal mode function and the horizontal wavenumber.

3. The method according to claim 2, characterized in that, The output of the wavenumber domain beamforming is specifically the integration of the received data along the array aperture and the introduction of a steering coefficient and a compensation coefficient. The compensation coefficient is used to compensate for the energy attenuation during the propagation of the sound wave.

4. The method according to claim 3, characterized in that, The wavenumber spectrum is specifically obtained by taking the modulus of the beam output corresponding to each horizontal wavenumber to obtain the energy response, and scanning the horizontal wavenumbers to obtain the wavenumber spectrum. The horizontal wavenumbers are sampled at equal intervals between the wavenumbers corresponding to the average sound speed on the mezzanine and the wavenumbers corresponding to the minimum sound speed on the seabed.

5. The method according to claim 4, characterized in that, The horizontal wavenumber-frequency two-dimensional distribution matrix is ​​specifically obtained by selecting all array element sound pressure signals at each frequency point within the spectrum from the start frequency to the cutoff frequency, performing wavenumber domain beamforming, and obtaining energy distribution matrices in the frequency domain and horizontal wavenumber domain.

6. The method according to claim 5, characterized in that, The coordinate transformation is specifically calculated by subtracting the difference between the horizontal wavenumber and the wavenumber corresponding to the frequency. The wavenumber corresponding to the frequency is determined by dividing the frequency by the average speed of sound in water and then multiplying by twice pi.

7. The method according to claim 6, characterized in that, The cumulative energy-horizontal wavenumber difference curve is specifically obtained by integrating the energy corresponding to each sampling point in the horizontal wavenumber difference domain along the full bandwidth from the starting frequency to the cutoff frequency.

8. The method according to claim 7, characterized in that, The two-layer game model includes an upper-layer model aimed at optimizing sampling accuracy and a lower-layer model aimed at improving energy resolution. The upper-layer model and the lower-layer model are coupled and optimized through the energy resolution index and the sampling standard deviation.

9. The method according to claim 8, characterized in that, The upper-level model specifically takes the sampling standard deviation, signal bandwidth, and energy resolution indices of the horizontal wavenumber difference domain as inputs, and outputs sampling accuracy optimization indices to adjust the number of samples and optimize sampling accuracy.

10. The method according to claim 9, characterized in that, The lower-level model specifically takes the peak energy, peak half-width at half-maximum, and sampling standard deviation of the horizontal wavenumber difference domain of the cumulative energy-horizontal wavenumber difference curve as input, and outputs the energy resolution index as the input parameters of the upper-level model.

Images