Single Ping shallow sea reverberation signal processing method

Abstract

The invention belongs to the technical field of marine acoustic signal processing, and provides a single-Ping shallow sea reverberation signal processing method, which comprises the following steps: obtaining an original time-frequency spectrum matrix of a single-Ping shallow sea reverberation signal, and carrying out interference structure strengthening operation based on local similarity on the original time-frequency spectrum matrix, the interference structure strengthening operation based on local similarity comprises the following steps: performing matrix expansion operation on an original time-frequency spectrum matrix based on interference structure local similarity of shallow sea reverberation to obtain an expansion matrix of the original time-frequency spectrum matrix; performing robust principal component analysis operation on the expansion matrix to obtain a low-rank approximate matrix of the expansion matrix; and performing matrix inverse expansion operation based on sliding window averaging on the low-rank approximate matrix to obtain a primary reinforcement matrix of the original time-frequency spectrum matrix. According to the method provided by the invention, characteristic enhancement is carried out on the time-frequency spectrum by utilizing the inherent characteristics of shallow sea reverberation, and obvious interference structure characteristics can be obtained without multi-Ping data.

Description

Technical Field

[0001] The present application belongs to the technical field of ocean acoustic signal processing, and specifically relates to a method for processing a single Ping shallow sea reverberation signal. Background Art

[0002] Shallow sea reverberation is mainly caused by the scattering of sound waves by the sea surface, seabed and inhomogeneous bodies inside the seawater medium. In the shallow sea environment, due to the mutual interference of sound waves of different modes when propagating in the seawater, the shallow sea reverberation signal shows a specific interference structure pattern (reverberation interference fringes) in the time-frequency domain or the distance-frequency domain. The existence of the interference structure plays an important role in the signal processing and target detection of the sonar system. For example, it can be used to estimate the reverberation level of active sonar, or detect and track shallow sea underwater targets. However, due to the influence of factors such as random scattering on the seabed, the time-varying environment of the ocean and noise, the time-frequency spectrum of the original shallow sea reverberation signal has problems such as high noise level and insufficient clarity, making it difficult to obtain a significant interference fringe structure from it.

[0003] At present, there are many methods for interferometrically strengthening the shallow sea reverberation signal, for example, processing the reverberation signal from the wavelet domain and the fractional Fourier domain, performing singular value decomposition, modal decomposition and principal component analysis on the signals at different stages of the processing, etc.; for example, the Chinese invention patent CN117250604A previously applied for by the applicant uses multiple echo signals to construct an echo signal spectrum enhancement matrix to enhance the low-rank characteristics of the echo signal.

[0004] The above existing shallow sea reverberation signal processing methods all adopt the method of processing with the joint Ping signal to ensure the requirement of the data volume carrying the reverberation interference structure information. However, considering the influence of the shallow sea time-varying environment on the underwater acoustic signal during the propagation process, it is difficult to ensure the consistency of the interference structure information in each Ping signal, making it difficult to ensure the stability and clarity of the reverberation interference structure separated from multiple Ping signals. Therefore, there is an urgent need for a shallow sea reverberation signal processing method that can obtain significant interference structure characteristics through a single Ping signal. Summary of the invention

[0005] The purpose of this application is to provide a processing method for a single Ping shallow sea reverberation signal. The method utilizes the interference structure distribution characteristics of the shallow sea reverberation to perform structural enhancement processing on the Ping signal, effectively improving the accuracy of signal separation and feature extraction, and obtaining significant reverberation interference structure characteristics using only the Ping signal.

[0006] The embodiments of the present application can be implemented through the following technical solutions: A method for processing a single-ping shallow sea reverberation signal comprises the following operations: obtaining an original time-frequency spectrum matrix of the single-ping shallow sea reverberation signal, and performing an interference structure enhancement operation based on local similarity on the original time-frequency spectrum matrix, wherein the interference structure enhancement operation based on local similarity comprises steps A1 to A3: Step A1, based on the local similarity of the interference structure of shallow sea reverberation, performing a matrix expansion operation on the original time-frequency spectrum matrix to obtain an expanded matrix of the original time-frequency spectrum matrix; Step A2, performing a robust principal component analysis operation on the extended matrix to obtain a low-rank approximate matrix of the extended matrix; Step A3: performing a matrix inverse expansion operation based on sliding window averaging on the low-rank approximate matrix to obtain a primary enhancement matrix of the original time-frequency spectrum matrix.

[0007] Furthermore, the local similarity of the interference structure is specifically: in the time-frequency spectrum of the shallow sea reverberation signal, local areas that are similar to each other are located on both sides of the diagonal of the time-frequency spectrum and can be mapped to each other through translation.

[0008] Further, the local similarity of the interference structure is analyzed and determined by the following steps: Establish a shallow sea reverberation model; Determine the ideal time-frequency spectrum matrix of a single-ping shallow sea reverberation signal based on the shallow sea reverberation model; Local regions with similar interference structures are extracted from the ideal time-frequency spectrum matrix, and their local similarity of interference structures is determined.

[0009] Furthermore, the expansion matrix is ​​obtained by the following steps: Step A11, based on the original time-frequency spectrum matrix of the single Ping shallow sea reverberation signal The size is determined by A sliding window, where , are the minimum and maximum frequency sampling points respectively, , are the minimum and maximum time sampling points respectively, , are the number of rows and columns of the sliding window respectively; Step A12, making the sliding window take 1 as the step length, Slide in the order of rows first and columns later, and intercept the elements in the window one by one, and get the following formula Sub-matrices : , in, for OK A two-dimensional matrix of columns, , They are the serial numbers of the frequency sampling points and time sampling points in the original time-frequency spectrum matrix respectively; Step A13: According to the window sliding order, each sub-matrix Reshape it into a length of One-dimensional column vector of : , in, () means extracting all elements of a two-dimensional matrix in the order of rows first and columns and arranging them into column vectors; Step A14: slide each window in the order of the window sliding. Arrange them row by row and get the expanded matrix as shown below : , in, for OK A two-dimensional matrix of columns.

[0010] Further, step A2 includes the following steps: Step A21, initialize the low rank matrix , sparse matrix And set , ,in, , The size and Consistent, for The expected value of the rank, for The number of non-zero entries in ; Step A22: Iteratively execute steps A221 to A222 until the number of iterations reaches a preset upper limit, or the result of the robust principal component analysis reaches a preset convergence target: Step A221, and difference Rank The low-rank approximation is processed and the low-rank matrix is ​​updated according to the low-rank approximation result. ; Step A222, from the residual matrix The one with the largest absolute value is retained elements, and the remaining elements are set to zero, thus obtaining the updated sparse matrix ; Step A23, the low-rank matrix at the end of the iteration As an expansion matrix The low-rank approximation matrix of .

[0011] Further, step A3 includes the following steps: Step A31, split each column of the low-rank approximation matrix and number them in the same order as in step A14, to obtain the following formula: Column vectors: ; Step A32, for each column vector , by dividing each element into The way to convert the elements into a row vector is to Convert to OK Submatrix of columns ; Step A33, placing each sub-matrix in the same window sliding order as in step A12 , the elements at the same position are superimposed and averaged, and finally a primary enhancement matrix of the original time-frequency spectrum matrix is ​​obtained .

[0012] Preferably, when When is an odd number, for The rounded value of When is an even number, for ;when When is an odd number, for The rounded value of When is an even number, for .

[0013] Preferably, the method for processing a single Ping shallow sea reverberation signal further comprises performing an interference structure enhancement operation based on a waveguide invariant on the primary enhancement matrix.

[0014] Further, performing an interference structure enhancement operation based on a waveguide invariant on the primary enhancement matrix comprises the following steps: Step B1, a primary enhancement matrix of the original time spectrum Each grid point Slope Path interpolation to obtain the values ​​at several interpolation points ,in, , , , are the frequency resolution and time resolution, respectively. is the preset waveguide invariant, is the interpolation point; Step B2, The window is drawn around : , in, is the window size parameter; Step B3, determining the secondary enhancement matrix of the original time-frequency spectrum matrix based on the following formula: , in, For the secondary reinforcement matrix The value at is a counting function.

[0015] Preferably, the preset waveguide invariant has multiple candidate values; after using each candidate value to perform interference structure enhancement operation based on the waveguide invariant, the operation with the most obvious enhancement effect is selected from each operation result, and its result is used as the secondary enhancement matrix of the original time-frequency spectrum matrix.

[0016] The embodiment of the present application provides a method for processing a single Ping shallow sea reverberation signal, which makes full use of the inherent characteristic of the local similarity of the interference structure in the shallow sea reverberation time spectrum, expands the effective information carried by the single Ping signal and strengthens the interference structure. It can extract significant interference structure features from the single Ping shallow sea reverberation signal. Compared with the existing shallow sea reverberation signal processing method that needs to combine multiple Ping signals to extract the interference structure, the calculation amount is small, and there is no need to consider the impact of the time-varying characteristics of the marine environment on the multiple Ping signals. It has stronger noise resistance and adaptability in complex marine environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic flow chart of a method for processing a single Ping shallow sea reverberation signal according to some embodiments of the present application; Figure 2 It is a schematic diagram of the original time-frequency spectrum matrix of a specific single-ping shallow sea reverberation signal; Figure 3 It is a schematic diagram of the shallow sea reverberation model and the propagation principle of the shallow sea reverberation signal; Figure 4 It is a schematic diagram of the ideal time-frequency spectrum matrix of a single-ping shallow sea reverberation signal obtained according to the shallow sea reverberation model; Figure 5 is a schematic diagram of performing a matrix expansion operation on an original time-frequency spectrum matrix in a specific embodiment; Figure 6 is a schematic diagram of an area with local similarity in an expansion matrix in a specific embodiment; Figure 7 is a flow chart of performing robust principal component analysis on an extended matrix in a specific embodiment; Figure 8 for Figure 2 Schematic diagram of a primary enhancement matrix of the original time-frequency spectrum matrix shown; Fig. 9 It is a schematic diagram of the original time-frequency spectrum matrix of another specific single-Ping shallow sea reverberation signal; Fig.10 for Fig. 9 Schematic diagram of a primary enhancement matrix of the original time-frequency spectrum matrix shown; Fig.11 A schematic flow chart of a method for processing a single Ping shallow sea reverberation signal according to some embodiments of the present application; Fig.12 1 is a schematic diagram of the principle of performing an interference structure enhancement operation based on a waveguide invariant on a primary enhancement matrix in a specific embodiment; Fig.13 for Fig. 9 Schematic diagram of the secondary enhancement matrix of the original time-frequency spectrum matrix shown. DETAILED DESCRIPTION

[0018] Hereinafter, the present application will be further described based on preferred embodiments with reference to the accompanying drawings.

[0019] Figure 1 This is a flowchart of a method for processing a single Ping shallow sea reverberation signal according to some embodiments of the present application, with reference to Figure 1 , the method includes the following operations: Operation 1: Obtain the original time-frequency spectrum matrix of a single Ping shallow sea reverberation signal; Operation 2: performing an interference structure enhancement operation based on local similarity on the original time-frequency spectrum matrix.

[0020] Specifically, operation 1 can be implemented in a manner known to those skilled in the art. For example, a single Ping shallow sea reverberation signal (time domain) can be subjected to short-time Fourier transform processing to obtain its original time-frequency spectrum matrix: ,in, , are the minimum and maximum frequency sampling points respectively, , are the minimum and maximum time sampling points respectively. Obviously, the original time-frequency spectrum matrix for OK A matrix of columns, where the elements of each grid point are the sound pressure or sound intensity of the reverberation signal corresponding to each sampling time and sampling frequency.

[0021] Figure 2 The figure shows a two-dimensional color map of the original time-frequency spectrum matrix of a specific single Ping shallow sea reverberation signal after the above processing. The Ping signal comes from the measured data set of a shallow sea sound field propagation experiment. The shallow sea water depth of the experiment is 88 meters, the sound source is an explosion sound source, the depth is 50 meters, the receiver depth is 65 meters, and the distance between the sound source and the receiver is 200 meters. A total of 31 Ping shallow sea reverberation signals were collected. Figure 2 This is the processing result of the shallow sea reverberation signal of the first Ping. The size of its original time-frequency spectrum matrix is ​​256×46, that is, it includes 256 frequency sampling points and 46 time sampling points. In this time-frequency spectrum matrix and the related time-frequency spectrum matrix described later, the values ​​of each grid point have been normalized.

[0022] pass Figure 2 It can be seen that the original time-frequency spectrum matrix of a single Ping shallow sea reverberation signal contains a large amount of noise interference, from which it is difficult to obtain significant interference structure features. Therefore, in an embodiment of the present application, an interference structure enhancement operation is performed on the original time-frequency spectrum matrix through operation 2 to suppress noise, so that the interference fringes are significantly enhanced in the processed time-frequency spectrum matrix.

[0023] Obviously, due to the limited amount of information carried by a single Ping shallow sea reverberation signal, it is impossible to use multiple independent signals for noise suppression and feature enhancement like the multi-Ping joint processing scheme. Therefore, only by strengthening the effective information in it according to the inherent characteristics of the shallow sea reverberation signal can the effect of the single Ping signal processing method be guaranteed. To this end, it is necessary to first analyze the mechanism and characteristics of the shallow sea reverberation interference structure in order to select the appropriate matrix enhancement operation.

[0024] In some preferred embodiments, the local similarity characteristics of the interference structure of shallow sea reverberation can be analyzed and determined by the following steps: The first step is to establish a shallow sea reverberation model.

[0025] Specifically, Figure 3 A specific shallow ocean bistatic reverberation model is shown, consisting of Figure 3 It can be seen that the dual-base reverberation at a certain moment can be interpreted as the sum of the echo signals of the scatterers whose sum of distances from the sound source and the receiver is a fixed value, that is, the echo signal generated by the elliptical ring with the sound source and the receiver as the focus, where is the sound source position, is the receiver position, is the distance between the sound source and the receiver, is the distance from the sound source to the scatterer, is the distance from the scatterer to the receiver, is the scattering element on the elliptical ring.

[0026] The second step is to determine the ideal time-frequency spectrum matrix of a single Ping shallow sea reverberation signal based on the shallow sea reverberation model.

[0027] The ideal time-frequency spectrum matrix is ​​the result obtained by considering only the ideal propagation of the underwater acoustic signal in the shallow sea sound channel and the seabed scattering. Since it does not alias the various noises in the actual propagation process, it can clearly display the interference fringes in the time-frequency spectrum, which is conducive to analyzing the similarity of the interference structure.

[0028] The ideal time-frequency spectrum matrix of a single Ping shallow sea reverberation signal can be established by a method known to those skilled in the art. For example, in some optional embodiments, the reverberation sound pressure spectrum can be established by referring to the Chinese invention patent CN117250604A previously applied by the applicant. or Reverberation Intensity Spectrum The ideal expression of , are continuous frequency variables and time variables respectively. The above reverberation sound pressure spectrum or Reverberation Intensity Spectrum The ideal expressions of can be used as the ideal time-frequency spectrum of a single-ping shallow sea reverberation signal. By discretely sampling it, the ideal time-frequency spectrum matrix can be obtained.

[0029] Alternatively, in some other optional embodiments, the Helmholtz equation can be solved by using the separation of variables method to convert the sound source The far-field sound pressure of one-way propagation is expressed as the sum of a series of simple normal waves. Then, the sound pressure expression of reverberation is obtained using the principle of sound field reciprocity. The time domain echo signal of a single scatterer is calculated using the inverse Fourier transform. All scatterers are then integrated to obtain the time domain expression of a single Ping reverberation signal. By performing a short-time Fourier transform on the time domain expression, the ideal time-frequency spectrum matrix of a single Ping reverberation signal can also be obtained (similarly, the values ​​of each element in the matrix can be sound pressure or sound intensity).

[0030] The third step is to extract local regions with similar interference structures from the ideal time-frequency spectrum matrix and determine the local similarity of their interference structures.

[0031] Figure 4 The schematic diagram of the ideal time-frequency spectrum matrix of a single Ping reverberation signal generated by the above steps in a specific embodiment is shown. The environmental parameter settings are similar to those of the actual shallow sea sound field propagation experiment. The water depth is 88.75m, and the sound velocity, density and absorption coefficient of the sediment are 1664m / s and 1.9g / cm 3and 0.2dB / The sound source depth is 50m, the receiver depth is 65m, the distance between the sound source and the receiver is 200m, the reverberation time is 2-6s, the frequency range is 300-555Hz, and the seabed scattering adopts the Lambert scattering model.

[0032] like Figure 4 As shown, since the ideal time-frequency spectrum matrix does not contain various noise pollution, it presents obvious structural characteristics of interference fringes. By searching and extracting areas with similar interference structure distribution characteristics in the ideal time-frequency spectrum matrix, local similarity areas as shown in the red and black boxes in the figure can be obtained. As shown in the figure, these two similar areas are located on both sides of the diagonal of the ideal time-frequency spectrum matrix and can be mapped to each other by translation.

[0033] Figure 4 The presented interference structure characteristics and their local similarities are the result of periodic reinforcement and cancellation during the superposition of simple normal waves of various modes in the propagation of shallow sea sound fields, and are inherent characteristics of shallow sea reverberation. Obviously, the local similarity of this interference structure also exists in the time-frequency spectrum of the measured shallow sea reverberation signal. Therefore, in operation two, this characteristic can be used to enhance the characteristics of the original time-frequency spectrum matrix, thereby achieving the purpose of suppressing noise and highlighting the interference structure characteristics.

[0034] Specifically, Figure 1 As shown, operation 2 further includes the following three steps: Step A1, based on the local similarity of the interference structure of shallow sea reverberation, performing a matrix expansion operation on the original time-frequency spectrum matrix to obtain an expanded matrix of the original time-frequency spectrum matrix; Step A2, performing a robust principal component analysis operation on the extended matrix to obtain a low-rank approximate matrix of the extended matrix; Step A3: performing a matrix inverse expansion operation based on sliding window averaging on the low-rank approximate matrix to obtain a primary enhancement matrix of the original time-frequency spectrum matrix.

[0035] Among them, step A1 uses the local similarity of the interference structure of shallow sea reverberation obtained above to calculate the original time-frequency spectrum matrix of the single Ping reverberation signal Perform matrix expansion operation to enhance the effective information content of the expanded matrix, then in step A2, extract the principal component from the enhanced matrix through principal component analysis operation, and finally in step A3, perform matrix inverse expansion operation to ensure that the enhanced matrix obtained after processing is With Same matrix structure.

[0036] The specific implementation of steps A1 to A3 of operation 2 will be described in detail below with reference to the accompanying drawings.

[0037] In some specific embodiments, the matrix expansion operation of step A1 includes the following steps: Step A11, based on the original time-frequency spectrum matrix of the single Ping shallow sea reverberation signal The size is determined by A sliding window, where , are the minimum and maximum frequency sampling points respectively, , are the minimum and maximum time sampling points respectively, , are the number of rows and columns of the sliding window respectively.

[0038] when When is an odd number, it is preferred to The rounded value of ,when When is an even number, it is preferred to As ;when When is an odd number, it is preferred to The rounded value of ,when When is an even number, it is preferred to As Through the above settings, the original time-frequency spectrum matrix can be used to the maximum extent. The local similarity information of the interference structure carried in the matrix is ​​used to enhance the effect of strengthening the matrix interference structure.

[0039] Step A12, making the sliding window take 1 as the step length, Slide in the order of rows first and columns later, and intercept the elements in the window one by one, and get the following formula Sub-matrices : , in, for OK A two-dimensional matrix of columns, , They are the serial numbers of the frequency sampling points and time sampling points in the original time-frequency spectrum matrix respectively; Step A13: According to the window sliding order, each sub-matrix Reshape it into a length of One-dimensional column vector of : , in, () means extracting all elements of a two-dimensional matrix in the order of rows first and columns and arranging them into column vectors; Step A14: slide each window in the order of the window sliding. Arrange them row by row and get the expanded matrix as shown below : , in, for OK A two-dimensional matrix of columns.

[0040] Figure 5 With a As an example, the specific implementation results of each step of the matrix expansion operation are shown. One dimension of the matrix is ​​used to represent time (or distance), and the other dimension is used to represent frequency. Obviously, for The matrix of , ,Right now It can be expressed as: .

[0041] In order to achieve To maximize the use of effective information in the ,Right now , , the sliding window is set to 1 as the step size, Slide in the order of rows first and columns later and intercept the elements to construct the submatrix until all the positions that can be intercepted are traversed, and finally we can get The sub-matrices, in the order of truncation, are: , , , , , , , , , where each submatrix contains elements.

[0042] Furthermore, the three row vectors in each submatrix are converted into column vectors in sequence, and then arranged and concatenated in order to obtain 9 Column vector of : , , , , , , , , Finally, arrange these 9 column vectors in rows to get Figure 6 The expanded matrix with 9 rows and 9 columns is shown , its expression is shown as follows: .

[0043] pass Figure 6 It can be seen that in the process of expanding the original time-frequency spectrum matrix based on the local similarity of the interference structure, the regions with the same element distribution (as shown in the two 1×3 square regions and the two 2×3 square regions in the figure) are placed on both sides of the matrix diagonal in a translationally mapped manner. Therefore, the local similarity of the original time-frequency spectrum matrix is ​​further enhanced.

[0044] The expanded matrix is ​​obtained through step A1 Then, in step A2, a robust principal component analysis is performed on it to obtain a low-rank approximation matrix. In the embodiment of the present application, those skilled in the art can use various robust principal component analysis methods known to them to implement step A2. For example, Figure 7 The process shown in the figure is to expand the matrix through the following steps Perform robust principal component analysis: Step A21, initialize the low rank matrix , sparse matrix And set , ,in, , The size and Consistent, for The expected value of the rank, for The number of non-zero entries in .

[0045] Step A22: Iteratively execute steps A221 to A222 until the number of iterations reaches a preset upper limit, or the result of the robust principal component analysis reaches a preset convergence target: Step A221, and difference Rank The best low-rank approximation is processed, and the low-rank matrix is ​​updated according to the processing results. ; Step A222, from the residual matrix The one with the largest absolute value is retained elements, and the remaining elements are set to zero, thus obtaining the updated sparse matrix ; Step A23, the low-rank matrix at the end of the iteration As an expansion matrix The low-rank approximation matrix of .

[0046] In the above steps, , The upper limit of the number of iterations can be set to a more appropriate value according to the processing accuracy, and the convergence target can be set to, for example, of the form, where represents the Frobenius norm, In addition, the present application does not limit the optimal low-rank approximation processing method in step A221, and those skilled in the art can implement the optimal low-rank approximation processing by using various methods including but not limited to singular value decomposition (SVD), QR decomposition, etc. according to the prior art.

[0047] The low-rank approximation matrix reflects the expansion matrix The "structure" or "background" in The original time-frequency spectrum matrix has been changed Therefore, it is necessary to reconstruct the low-rank approximation matrix through step A3 to restore it to the same level as Having the same structure ensures that the subsequent interference structure analysis can be performed in the correct time-frequency domain.

[0048] In some specific embodiments, step A3 includes the following steps: Step A31, split each column of the low-rank approximation matrix and number them in the same order as in step A14, to obtain the following formula: Column vectors: ; Step A32, for each column vector , by dividing each element into The way to convert the elements into a row vector is to Convert to OK Submatrix of columns ; Step A33, placing each sub-matrix in the same window sliding order as in step A12 , the elements at the same position are superimposed and averaged, and finally a primary enhancement matrix of the original time-frequency spectrum matrix is ​​obtained .

[0049] Obviously, the above steps are Figure 5The reverse operation of the matrix expansion operation shown in the figure, wherein, since during the matrix expansion operation, when elements are intercepted in the window sliding order, elements at some positions may appear in multiple sub-matrices, correspondingly, during the inverse matrix expansion operation, multiple elements may be superimposed at a certain position. Therefore, when multiple elements are superimposed at the same position, it is necessary to perform an averaging operation on the superimposed results.

[0050] Figure 8 Shows the Figure 2 The original time-frequency spectrum matrix in After performing the matrix enhancement operations from step A1 to step A3, the primary enhancement matrix .pass Figure 8 It can be seen that after the interference structure enhancement operation based on local similarity, the time-frequency spectrum of the single Ping reverberation signal has shown obvious interference fringe structure characteristics, which proves that the method provided in the present application can effectively extract the reverberation interference fringe structure from the single Ping reverberation signal without the support of multi-Ping data, providing a more reliable method for reverberation suppression and target detection in complex marine environments, and has significant advantages in practical engineering applications.

[0051] Fig. 9 Schematic diagram of the original time-frequency spectrum matrix of another single Ping shallow sea reverberation signal. Figure 2 From the same dataset, its number is 9. Fig.10 A schematic diagram of a primary enhancement matrix of the original time-frequency spectrum matrix obtained after performing an interference structure enhancement operation based on local similarity is shown. Fig.10 It can be seen that the Ping shallow sea reverberation signal has only undergone one enhancement operation, and its interference structure characteristics have not been significantly improved. It can be seen that due to the complexity and variability of the marine environment, the various types of noise superimposed on the signals emitted at different times in the same experiment have different degrees of influence on the effective information. Even after one enhancement operation, some Ping signals may still not show a significant interference structure. Therefore, it is necessary to further enhance the structural characteristics of some of the one-time enhancement matrices. Obviously, a different mechanism from the first enhancement operation should be used at this time to have a more significant impact on the signal that has not been significantly improved after the first enhancement operation.

[0052] observe Figure 4 It can be seen that the ideal time-frequency spectrum of the shallow sea reverberation signal (in terms of (represented by) there are regular interference fringes on the time-frequency plane, and these interference fringes satisfy the following waveguide invariant relationship: , in, is a waveguide invariant, which characterizes the slope of the interference fringes, which is also determined by the inherent characteristics of the shallow sea environment. Therefore, this characteristic can be used to perform an interference structure enhancement operation based on the waveguide invariant on the primary enhancement matrix. Based on the above analysis, in some preferred embodiments of the present application, Fig.11 As shown, the method also includes operation three: performing an interference structure enhancement operation based on a waveguide invariant on the primary enhancement matrix of the original time-frequency spectrum matrix.

[0053] Specifically, the primary enhancement matrix of the original time-frequency spectrum matrix is Performing an interference structure enhancement operation based on a waveguide invariant further includes the following steps: Step B1, a primary enhancement matrix of the original time spectrum Each grid point Slope Path interpolation to obtain the values ​​at several interpolation points ,in, , , , are the frequency resolution and time resolution, respectively. is the preset waveguide invariant, is the interpolation point; Step B2, The window is drawn around : , in, is the window size parameter; Step B3, determining the secondary enhancement matrix of the original time-frequency spectrum matrix based on the following formula: , in, For the secondary reinforcement matrix The value at is a counting function.

[0054] refer to Fig.12 ,right The interference structure enhancement operation based on the waveguide invariant is to interpolate each point along the direction specified by the waveguide invariant on the time-frequency plane, and replace the original value of each point with the average of several interpolation points. Obviously, through this operation, the parts of the slope that show the same trend due to the interference fringes can be superimposed and enhanced, while components such as random noise offset each other through accumulation, thereby achieving the enhancement operation of the interference structure.

[0055] The preset waveguide invariant can be set according to the measurement results of the shallow sea environment. In addition, in some preferred embodiments, multiple candidate values ​​can be set for the waveguide invariant. Then, after using each candidate value to perform the interference structure enhancement operation based on the waveguide invariant, the operation with the most obvious enhancement effect is selected from each operation result (that is, the alternating interference fringe characteristics after the operation are the most significant), and the result is used as the secondary enhancement matrix of the original time-frequency spectrum matrix. For each grid point in the waveguide, only the interpolation and superposition in the direction defined by its waveguide invariant can be significantly enhanced. Therefore, the value of the waveguide invariant can be obtained more accurately through the above steps.

[0056] Fig.13 Shows the Fig. 9 The schematic diagram of the secondary enhanced matrix obtained after the original time-frequency spectrum matrix is ​​enhanced twice is shown in Figure 1. Fig.10 and Fig.13 It can be seen that by further strengthening the single Ping signal using the waveguide invariant characteristics, the problem of poor effect of single strengthening operation on some Ping signals can be effectively solved, and the accuracy and robustness of the method can be effectively improved.

[0057] The above is a detailed introduction to the specific implementation methods of the present application. For those skilled in the art, several improvements and modifications may be made to the present application without departing from the principles of the present application. These improvements and modifications also fall within the scope of protection of the claims of the present application.

Claims

1. A method for processing a single Ping shallow sea reverberation signal, characterized in that: The following operations are included: The original time-frequency spectrum matrix of a single Ping shallow sea reverberation signal is obtained, and an interference structure enhancement operation based on local similarity is performed on the original time-frequency spectrum matrix, wherein the interference structure enhancement operation based on local similarity includes steps A1 to A3: Step A1, based on the local similarity of the interference structure of shallow sea reverberation, performing a matrix expansion operation on the original time-frequency spectrum matrix to obtain an expanded matrix of the original time-frequency spectrum matrix; Step A2, performing a robust principal component analysis operation on the extended matrix to obtain a low-rank approximate matrix of the extended matrix; Step A3: performing a matrix inverse expansion operation based on sliding window averaging on the low-rank approximate matrix to obtain a primary enhancement matrix of the original time-frequency spectrum matrix.

2. The method for processing a single Ping shallow sea reverberation signal according to claim 1, characterized in that: The local similarity of the interference structure is specifically: In the time-frequency spectrum of the shallow sea reverberation signal, local areas that are similar to each other are located on both sides of the diagonal of the time-frequency spectrum and can be mapped to each other through translation.

3. The method for processing a single Ping shallow sea reverberation signal according to claim 2, characterized in that: The local similarity of the interference structure is analyzed and determined by the following steps: Establish a shallow sea reverberation model; Determine the ideal time-frequency spectrum matrix of a single-ping shallow sea reverberation signal based on the shallow sea reverberation model; Local regions with similar interference structures are extracted from the ideal time-frequency spectrum matrix, and their local similarity of interference structures is determined.

4. The method for processing a single Ping shallow sea reverberation signal according to claim 1, characterized in that: The expansion matrix is ​​obtained by following the steps below: Step A11, based on the original time-frequency spectrum matrix of the single Ping shallow sea reverberation signal The size is determined by A sliding window, where , are the minimum and maximum frequency sampling points respectively, , are the minimum and maximum time sampling points respectively, , are the number of rows and columns of the sliding window respectively; Step A12, making the sliding window take 1 as the step length, Slide in the order of rows first and columns later, and intercept the elements in the window one by one, and get the following formula Sub-matrices : ， in, for OK A two-dimensional matrix of columns, , They are the serial numbers of the frequency sampling points and time sampling points in the original time-frequency spectrum matrix respectively; Step A13: According to the window sliding order, each sub-matrix Reshape it into a length of One-dimensional column vector of : ， in, () means extracting all elements of a two-dimensional matrix in the order of rows first and columns and arranging them into column vectors; Step A14: slide each window in the order of the window sliding. Arrange them row by row and get the expanded matrix as shown below : ， in, for OK A two-dimensional matrix of columns.

5. The method for processing a single Ping shallow sea reverberation signal according to claim 4, characterized in that: Step A2 further comprises the following steps: Step A21, initialize the low rank matrix , sparse matrix And set , ,in, , The size and Consistent, for The expected value of the rank, for The number of non-zero entries in ; Step A22: Iteratively execute steps A221 to A222 until the number of iterations reaches a preset upper limit, or the result of the robust principal component analysis reaches a preset convergence target: Step A221, and difference Rank The low-rank approximation is processed and the low-rank matrix is ​​updated according to the low-rank approximation result. ; Step A222, from the residual matrix The one with the largest absolute value is retained elements, and the remaining elements are set to zero, thus obtaining the updated sparse matrix ; Step A23, the low-rank matrix at the end of the iteration As an expansion matrix The low-rank approximation matrix of .

6. The method for processing a single Ping shallow sea reverberation signal according to claim 5, characterized in that: Step A3 further comprises the following steps: Step A31, split each column of the low-rank approximation matrix and number them in the same order as in step A14, to obtain the following formula: Column vectors: ； Step A32, for each column vector , by dividing each element into The way to convert the elements into a row vector is to Convert to OK Submatrix of columns ; Step A33, placing each sub-matrix in the same window sliding order as in step A12 , the elements at the same position are superimposed and averaged, and finally a primary enhancement matrix of the original time-frequency spectrum matrix is ​​obtained .

7. The method for processing a single Ping shallow sea reverberation signal according to any one of claims 4 to 6, characterized in that: when When is an odd number, for The rounded value of When is an even number, for ; when When is an odd number, for The rounded value of When is an even number, for .

8. The method for processing a single Ping shallow sea reverberation signal according to claim 1, characterized in that: Also includes: An interference structure enhancement operation based on a waveguide invariant is performed on the primary enhancement matrix.

9. The method for processing a single Ping shallow sea reverberation signal according to claim 8, characterized in that: Performing an interference structure enhancement operation based on a waveguide invariant on the primary enhancement matrix further comprises the following steps: Step B1, a primary enhancement matrix of the original time spectrum Each grid point Slope Path interpolation to obtain the values ​​at several interpolation points ,in, , , , are the frequency resolution and time resolution, respectively. is the preset waveguide invariant, is the interpolation point; Step B2, The window is drawn around : ， in, is the window size parameter; Step B3, determining the secondary enhancement matrix of the original time-frequency spectrum matrix based on the following formula: ， in, For the secondary reinforcement matrix The value at is a counting function.

10. The method for processing a single Ping shallow sea reverberation signal according to claim 9, characterized in that: The preset waveguide invariant has multiple candidate values; After using various candidate values ​​to perform interference structure enhancement operations based on waveguide invariants, the operation with the most obvious enhancement effect is selected from various operation results, and its result is used as the secondary enhancement matrix of the original time-frequency spectrum matrix.

Images