A high resolution directional method and system for a vector hydrophone

Abstract

The present application belongs to the field of underwater acoustic signal processing, and particularly relates to a high-resolution directional method and system for a vector hydrophone, which is aimed at solving the problems of insufficient resolution of the existing single vector hydrophone directional method, large calculation amount and easy divergence of the adaptive filter directional method by fusing time reversal convolution processing with orthogonal matching pursuit sparse reconstruction. Firstly, the time reversal convolution sound field dictionary is constructed by traversing all target directions and search angles using the ideal received signal relationship of the vector hydrophone. Secondly, the time domain filtering, channel combination weighting and time reversal convolution transformation are performed on the measured three-channel signals, and the complex amplitude at the target line spectrum frequency is extracted to construct the time reversal convolution observation vector. Finally, the target direction is sparsely reconstructed in the dictionary by the orthogonal matching pursuit algorithm. The divergence problem of the adaptive filter is overcome, the calculation amount is reduced, and high-resolution direction estimation under low signal-to-noise ratio is realized.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic signal processing, and in particular to a high-resolution directional method and system for vector hydrophones. Background Technology

[0002] Vector hydrophones are composed of a point-to-point acoustic pressure hydrophone and a particle velocity hydrophone. They can measure sound pressure and particle velocity information in the sound field in a spatially co-located and temporally synchronized manner. A single vector hydrophone can achieve target orientation by combining sound pressure and velocity information, which has important engineering application value for underwater small platform detection.

[0003] Currently, single-vector hydrophone location estimation algorithms mainly fall into three categories. The first category is conventional direction-of-arrival methods based on acoustic energy flow. These methods have clear physical meaning and high robustness, but their resolution is limited by the main lobe width. Under strong interference, weak targets are easily masked by sidelobe energy, making it difficult to effectively distinguish multiple targets in spatial proximity. The second category consists of high-resolution spatial spectrum estimation algorithms, represented by MVDR and MUSIC, which overcome the Rayleigh limit. However, MVDR suffers from a signal-to-noise ratio threshold effect; its performance degrades sharply below the threshold, and its robustness is poor. Subspace fitting algorithms like MUSIC require multi-dimensional spectral peak searches, resulting in enormous computational costs. The third category is sparse signal processing methods based on compressed sensing. These methods utilize the sparsity of the target's spatial distribution to achieve high-resolution DOA estimation through sparse reconstruction. However, existing research primarily focuses on vector hydrophone arrays, limiting their application in single-vector hydrophone scenarios.

[0004] Temporal deconvolution processing technology has been introduced into the field of single-vector hydrophone signal processing. Combined-channel temporal deconvolution can effectively suppress irrelevant noise and improve detection capability under low signal-to-noise ratio conditions. Existing technologies have proposed a azimuth estimation method based on adaptive cancellation of residual accumulation after temporal deconvolution. This method requires constructing LMS adaptive filters point-by-point in each spatial orientation and accumulating steady-state errors, which suffers from high computational cost, filter divergence, and difficulty in meeting the requirements for rapid detection. Summary of the Invention

[0005] The purpose of this invention is to provide a high-resolution direction finding method and system for vector hydrophones, aiming to solve the comprehensive technical problems of insufficient resolution in conventional algorithms, high-resolution spectral estimation methods having signal-to-noise ratio thresholds and large computational loads, and time-deconvolution adaptive cancellation methods having filter divergence and large computational loads.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a high-resolution orientation method for a vector hydrophone, comprising: S1. For each preset target orientation, traversing all preset search angles, performing channel combination weighting using the ideal received signals from the sound pressure channel, x-direction velocity channel, and y-direction velocity channel of the vector hydrophone, and convolving the combined signal with the corresponding time-reversed combined signal to transform it to the frequency domain, extracting the complex amplitude at the target line spectrum frequency and concatenating them sequentially to form the dictionary atom corresponding to the current target orientation; traversing all preset target orientations and combining all dictionary atoms to obtain a time-reversed sound field dictionary; S2. Filtering the actually measured sound pressure channel signal, x-direction velocity channel signal, and y-direction velocity channel signal of the vector hydrophone, performing channel combination weighting on the filtered three signals for each preset search orientation, and convolving the combined signal with the time-reversed combined signal to form the dictionary atom corresponding to the current target orientation; After convolving the signal with the corresponding time-reversed combined signal, transform it to the frequency domain, extract the complex amplitude at the target line spectrum frequency, traverse all preset search azimuths and concatenate them in order to obtain the time-deconvolution observation vector; S3, assign the time-deconvolution observation vector as the initial value of the residual vector, and initialize the sparse vector to the zero vector; S4, calculate the inner product of each column of the time-deconvolution sound field dictionary with the current residual vector, select the column index corresponding to the largest absolute value of the inner product and store it in the selected index set; use the dictionary columns corresponding to the selected index set to form a submatrix, and solve for the true amplitude corresponding to the target azimuth using the least squares method; fill the solved true amplitude into the sparse vector at the position corresponding to the selected index set; update the residual vector, and repeat the above process of step S4 until the iteration stopping condition is met, and take the target azimuth corresponding to the largest amplitude in the final sparse vector as the azimuth estimation result.

[0007] In step S1, forming the dictionary atoms corresponding to the current target orientation specifically includes: for the current target orientation, using the ideal sound pressure channel signal, the ideal x-direction vibration velocity channel signal, and the ideal y-direction vibration velocity channel signal, performing channel combination weighting according to the currently traversed search angle to obtain the first combination signal and the second combination signal; convolving the sum of the ideal sound pressure channel signal and the first combination signal with the time-reversed signals of the first combination signal and the second combination signal respectively to obtain two time-deconvolution time-domain signals; transforming the two time-deconvolution time-domain signals to the frequency domain, extracting the complex amplitude at the target line spectrum frequency as the atom element corresponding to the current search angle, and sequentially splicing them into dictionary atoms after traversing all search angles.

[0008] The channel combination weighting method includes: summing the product of the x-direction vibration velocity signal and the cosine of the current search angle with the product of the y-direction vibration velocity signal and the sine of the current search angle to obtain the first combined signal; and summing the negative value of the product of the x-direction vibration velocity signal and the sine of the current search angle with the product of the y-direction vibration velocity signal and the cosine of the current search angle to obtain the second combined signal.

[0009] Step S2 extracts the complex amplitude at the target line spectrum frequency, specifically including: for the current preset search azimuth, the measured sound pressure channel signal, x-direction vibration velocity channel signal, and y-direction vibration velocity channel signal are combined and weighted to obtain a first measured combined signal and a second measured combined signal; the sum of the measured sound pressure channel signal and the first measured combined signal is convolved with the time-reversed signals of the first and second measured combined signals, respectively, to obtain two measured time-deconvolutioned time-domain signals; the two measured time-deconvolutioned time-domain signals are transformed to the frequency domain, and the complex amplitude at the target line spectrum frequency is extracted as the observation vector element corresponding to the current search azimuth.

[0010] The filtering process is an in-band filtering, and the frequency band is determined by the target line spectrum frequency.

[0011] In step S4, the iteration stops when the norm of the current residual vector is less than a preset threshold, or when the number of iterations reaches a preset sparsity constraint value.

[0012] The expression for solving the true amplitude using the least squares method in step S4 is as follows:

[0013] in, The selected index set is represented by a submatrix consisting of the dictionary columns. The initial value of the residual vector is the deconvolution observation vector. To obtain the true magnitude vector, for The transpose of .

[0014] The target line spectrum frequency is obtained by line spectrum detection of the actual measured vector hydrophone sound pressure channel signal.

[0015] The preset target orientation and preset search angle both range from 0° to 360°, and the interval between adjacent angles is 1°.

[0016] Secondly, the present invention provides a high-resolution directional system for a vector hydrophone, comprising: a single-vector hydrophone, wherein the single-vector hydrophone is composed of an omnidirectional acoustic pressure hydrophone and a particle velocity hydrophone, the particle velocity hydrophone having mutually orthogonal x-direction channels and y-direction channels; the output terminal of the acoustic pressure hydrophone is used to output the acoustic pressure channel signal, and the x-direction channel output terminal and y-direction channel output terminal of the particle velocity hydrophone are respectively used to output the x-direction velocity channel signal and the y-direction velocity channel signal; and a signal acquisition module, the input terminal of the signal acquisition module being connected to the output terminal of the acoustic pressure hydrophone and the particle velocity hydrophone respectively. The x-channel and y-channel outputs of the hearing aid are connected for synchronous sampling and analog-to-digital conversion of the three output signals. A signal processing module is connected to the output of the signal acquisition module. The signal processing module internally includes a line spectrum detection unit and an orthogonal matched pursuit sparse reconstruction unit. The line spectrum detection unit performs line spectrum detection on the sound pressure channel signal to obtain the target line spectrum frequency. The orthogonal matched pursuit sparse reconstruction unit performs signal processing operations from steps S1 to S4 based on the target line spectrum frequency. The output of the signal processing module outputs the target azimuth estimation result.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention fully utilizes the advantages of multi-dimensional information fusion between the single-vector hydrophone's sound pressure channel and two orthogonal vibration velocity channels. The combined gain of the joint processing of sound pressure and vibration velocity significantly improves the line spectrum detection capability under low signal-to-noise ratio conditions. At the same time, the time-deconvolution processing utilizes the essential difference in correlation characteristics between signal and noise to spatiotemporally focus the target energy dispersed in the multipath channel, further suppressing the impact of unrelated noise and channel distortion on the orientation accuracy.

[0018] 2. This invention replaces the conventional array manifold dictionary with a sound field dictionary in the time-deconvolution transform domain, and replaces the existing method of constructing LMS adaptive filters point by point with a greedy iterative search of orthogonal matched pursuit. This fundamentally avoids the problem of adaptive filters easily diverging under low signal-to-noise ratio and channel mismatch conditions, while significantly reducing the computational load in the spatial orientation scanning process and improving the stability and engineering practicality of the orientation method.

[0019] 3. This invention organically combines the spatiotemporal focusing processing of temporal deconvolution with the sparse reconstruction of orthogonal matching pursuit. By sparsely decomposing the temporal deconvolution observation vector in the temporal deconvolution sound field dictionary, it breaks through the Rayleigh resolution limit of the vector hydrophone aperture. Even under low signal-to-noise ratio conditions, it can still obtain a sharp azimuth spectrum and a small angle estimation error, and can effectively distinguish multiple line spectrum targets with small spatial angular intervals.

[0020] 4. This invention adopts an architecture that constructs the deconvolutional sound field dictionary offline, constructs the observation vector online, and performs orthogonal matching pursuit. The signal processing flow is clear and controllable, the system structure is simple, and there is no need for multidimensional peak search and filter iteration convergence. It is suitable for real-time high-resolution orientation requirements in computing-restricted occasions such as underwater unmanned platforms. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the operation of a vector hydrophone provided in an embodiment of this application; Figure 2 This is a flowchart of a high-resolution orientation method for a vector hydrophone provided in an embodiment of this application; Figure 3 This is a comparison diagram of orientation results of different vector hydrophone orientation estimation algorithms provided in the embodiments of this application; Figure 4 This is a performance curve of the root mean square error of orientation estimation using different algorithms as a function of signal-to-noise ratio, provided in the embodiments of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0023] See Figure 1 , Figure 1 This is a schematic diagram of a vector hydrophone, showing the spatial coordinate system model of a single vector hydrophone: the origin O is the position of the vector hydrophone, xy is the horizontal plane, and z is the vertical axis; This is the horizontal azimuth angle of the underwater target relative to the vector hydrophone, used to define the target's incident direction. It forms the geometric basis of the signal model in this invention and is used to derive the ideal signal expressions for sound pressure and x / y direction vibration velocity. The vector hydrophone is used for underwater target detection and azimuth estimation. The vector hydrophone is composed of a point-to-point composite of an omnidirectional sound pressure hydrophone and a particle velocity hydrophone. The particle velocity hydrophone has mutually orthogonal x-direction and y-direction channels, enabling spatially co-located and temporally synchronized measurement of sound pressure and particle velocity information in the sound field. The vector hydrophone sound pressure channel signal satisfies the ideal far-field conditions. x-direction vibration velocity channel signal y-direction vibration velocity channel signal The following relationship exists:

[0024]

[0025]

[0026] in, The angular frequency of the target signal. Let be the true horizontal projection azimuth angle of the target relative to the vector hydrophone. The above ideal signal model is used as the theoretical atom for constructing the deconvolutional sound field dictionary, while the actual received signal is superimposed with environmental noise and multipath channel distortion.

[0027] For example, refer to Figure 2 This application provides a high-resolution orientation method for a vector hydrophone, comprising: S1. For each preset target orientation, traverse all preset search angles, and use the ideal received signals from the vector hydrophone sound pressure channel, x-direction vibration velocity channel, and y-direction vibration velocity channel to perform channel combination weighting. Then, convolve the combined signal with the corresponding time-reversed combined signal and transform it to the frequency domain. Extract the complex amplitude at the target line spectrum frequency and splice them in order to form the dictionary atom corresponding to the current target orientation. Traverse all preset target orientations and combine all dictionary atoms to obtain the time-reversed convolution sound field dictionary.

[0028] The target line spectrum frequency was obtained by line spectrum detection of the actual measured sound pressure channel signal from the vector hydrophone. Line spectrum detection is a common technique in underwater acoustic signal processing, aiming to extract stable line spectrum components from the target radiated noise from a broadband noise background. Line spectra are characterized by frequency stability and concentrated energy, making them important features for underwater target detection and parameter estimation.

[0029] The actual measured sound pressure channel signal of the vector hydrophone Spectral analysis can be performed using classic power spectrum estimation methods such as the periodogram method and the Welch method to obtain the power spectral density of the signal. The amplitude peaks are searched within the power spectrum; if the amplitude at a certain frequency is significantly higher than that of adjacent frequencies (exceeding a preset threshold), then that frequency is determined to be a line spectrum frequency. .

[0030] In one specific implementation, the sound pressure channel signal Perform a Fast Fourier Transform to obtain its spectrum. Calculate the power spectrum Search for local maxima within the frequency band of interest. If a certain frequency... If the power spectral value at a given location exceeds a preset multiple (e.g., 3 times) of the mean of its neighborhood, then it is determined that... The target line spectrum frequency.

[0031] Step S1 aims to construct a time-deconvolution sound field dictionary. Its purpose is to pre-establish an overcomplete atomic library covering all possible target azimuths, utilizing the theoretical relationship between sound pressure and vibration velocity of a vector hydrophone under ideal far-field conditions. Each column in the time-deconvolution sound field dictionary represents an ideal time-deconvolution sound field response template corresponding to a preset target azimuth. Subsequently, the actual measured time-deconvolution observation vector is sparsely decomposed in this dictionary using an orthogonal matching pursuit algorithm, thus achieving high-resolution estimation of the target azimuth. Pre-constructing the time-deconvolution sound field dictionary offline significantly reduces the computational load of online processing, meeting real-time processing requirements.

[0032] More specifically, the implementation of step S1 includes the following sub-steps: First, construct the atomic vector elements of the deconvolutioned sound field. Then, set the target orientation index. and search angle index ,make , For the first Preset target location and the Preset search angles The first combined signal is obtained by combining and weighting the channels using the theoretical expressions of the sound pressure channel signal, x-direction vibration velocity channel signal, and y-direction vibration velocity channel signal of the vector hydrophone under ideal far-field conditions. Second combined signal .

[0033] ideal sound pressure channel signal With the first combined signal The sum of these is respectively compared with the time-reversed signal of the first combined signal. The time-reversed signal of the second combined signal Perform convolution operations to obtain two time-deconvolutioned time-domain signals. and .Will and Perform discrete Fourier transforms to the frequency domain respectively to obtain and Extract target line spectrum frequencies The complex amplitude at the current search angle is used as the current search angle. The corresponding atomic vector elements.

[0034] Change search angle index ,make Repeat the above process until all preset search angles have been traversed. (Up to 360), to obtain the atomic vector elements corresponding to all search angles.

[0035] Next, the deconvolution sound field dictionary atomic vectors for the vector hydrophone are constructed. All atomic vector elements obtained by traversing all search angles are rearranged into a column vector according to the order of the search angles. This column vector is the first one. Preset target location The corresponding dictionary atoms.

[0036] Finally, construct the complete temporal deconvolution sound field dictionary. Change the target orientation index. ,make Repeat the above two sub-steps until all preset target locations have been traversed. (Up to 360). All dictionary atoms corresponding to the target orientation. Arranged column-wise, they form a complete temporal deconvolution sound field dictionary matrix. .

[0037] In one specific implementation, the preset target orientation and preset search angle both range from 0° to 360°, with an interval of 1° between adjacent angles, meaning there are a total of 360 preset target orientations and 360 preset search angles.

[0038] Assuming the current target location index ,Right now Search Angle Index ,Right now Under ideal far-field conditions, the three signals of the vector hydrophone are:

[0039] By search angle Channel combination weighting is performed to obtain the first combined signal. Second combined signal :

[0040]

[0041] in, This is a combined vibration velocity signal along the search direction. This is a combined vibration velocity signal perpendicular to the search direction.

[0042] Constructing the deconvolution time-domain signal:

[0043]

[0044] Among them, symbols This represents the convolution operation. and Performing discrete Fourier transforms on each, we obtain the frequency domain expression:

[0045]

[0046] in, , , They are respectively , , Discrete Fourier Transform, superscript This indicates the conjugate operation. For frequency variables.

[0047] Extracting target line spectrum frequencies The complex amplitude value at the location yields the current search angle. Corresponding atomic vector elements .in, and They are respectively and At discrete frequency Complex values ​​at that location, Sampling frequency, is the number of points in the Discrete Fourier Transform.

[0048] Change the search perspective, make Repeat the above calculations to obtain the atomic vector elements corresponding to all 360 search angles. Rearrange all elements in order into column vectors. :

[0049] The dimension of this column vector is (360 search angles, each angle corresponds to 2 complex amplitudes).

[0050] Change the target's location, so that Repeat the above process to obtain dictionary atoms corresponding to all 360 target orientations. Arrange all dictionary atoms in columns to obtain the complete temporal deconvolution sound field dictionary:

[0051] The dimension of the time-deconvolution sound field dictionary matrix is Each column of this dictionary represents an ideal time-deconvolutional sound field response template corresponding to a preset target orientation. The time-deconvolutional sound field dictionary constructed in this way introduces time-deconvolution processing into the construction process of dictionary atoms, giving the dictionary atoms a natural channel-matching compression characteristic. Compared to conventional overcomplete atom libraries based on array manifolds, the dictionary atoms of this invention can more accurately match the physical model of the actual received signal in the time-deconvolutional transform domain, providing a superior atomic basis for subsequent orthogonal matching pursuit sparse reconstruction.

[0052] S2. The actual measured vector hydrophone sound pressure channel signal, x-direction vibration velocity channel signal, and y-direction vibration velocity channel signal are filtered. For each preset search orientation, the filtered three signals are combined and weighted. The combined signal is then convolved with the corresponding time-reversed combined signal and transformed to the frequency domain. The complex amplitude at the target line spectrum frequency is extracted. After traversing all preset search orientations, the signals are sequentially spliced ​​to obtain the time-deconvolution observation vector.

[0053] The construction of the time-deconvolution observation vector is a bridge connecting the actual measurement data and the sound field dictionary. It is used to process the three actual received time-domain signals according to the same construction logic as the dictionary atoms (i.e. the same channel combination weighting method and time-deconvolution transformation method) to obtain the observation vector in the same transformation domain as the dictionary atoms.

[0054] More specifically, the implementation of step S2 includes the following sub-steps: First, the three measured signals are filtered. This involves filtering the sound pressure channel signal actually measured by the vector hydrophone. x-direction vibration velocity channel signal and y-direction vibration velocity channel signal Perform in-band filtering. The passband range of the filter is determined by the target line spectrum frequency. Determined, usually with The center frequency and bandwidth are set according to the signal characteristics and noise environment to preserve the target line spectrum components and filter out out-of-band noise.

[0055] Secondly, during construction, deconvolution is used to measure the elements of the sound field vector. The search orientation index is then set. ,make For the first Preset search directions The filtered three signals are then combined and weighted according to the search direction to obtain the first measured combined signal. Second measured combined signal .

[0056] The measured sound pressure channel signal Combined with the first measured signal The sum of these two values ​​is respectively compared with the time-reversed signal of the first measured combined signal. Time-reversed signal of the second measured combined signal Perform convolution operations to obtain two measured time-domain deconvolutioned signals. and .Will and Perform discrete Fourier transforms to convert to the frequency domain and extract the target line spectrum frequencies. The complex amplitude at the location is used as the current search orientation. The corresponding observation vector elements.

[0057] Change search location index ,make Repeat the above process until all preset search locations have been traversed. Up to 360).

[0058] Finally, construct the complete temporal deconvolution observation vector. Arrange all observation vector elements obtained by traversing all search directions in the order of the search directions, and rearrange them into a column vector. This column vector is the time-deconvolution observation vector.

[0059] In one specific implementation, the preset search orientation has an angle range of 0° to 360°, with an interval of 1° between adjacent angles, meaning there are a total of 360 preset search orientations.

[0060] Set the current search location index ,Right now The three time-domain signals actually measured are: , and First, the three signals are analyzed using the target line spectrum frequency. It is a bandpass filter centered on the signal, filtering out out-of-band noise and interference.

[0061] Search by current location Perform channel combination weighting:

[0062]

[0063] in, The measured combined vibration velocity signal along the current search direction. The measured combined vibration velocity signal is perpendicular to the current search direction.

[0064] Constructing the measured deconvolution time-domain signal:

[0065]

[0066] Will and Performing discrete Fourier transforms on each, we obtain the frequency domain expression:

[0067]

[0068] in, , , They are respectively , , Discrete Fourier Transform, superscript This indicates the conjugate operation.

[0069] Extracting target line spectrum frequencies The complex amplitude value at the location gives the current search orientation. Corresponding observation vector elements .in, and They are respectively and At discrete frequency Complex values ​​at the location.

[0070] Change the search direction, make Repeat the above calculations to obtain the observation vector elements corresponding to all 360 search azimuths. Rearrange all elements into column vectors in order. :

[0071] The dimension of this column vector is (360 search directions, each direction corresponds to 2 complex magnitudes).

[0072] The temporal deconvolution observation vector constructed in the above manner is in the same transform domain as the temporal deconvolution sound field dictionary constructed in step S1. They share the same physical meaning and mathematical structure, thus ensuring that the matching metric between dictionary atoms and the observation vector has a clear physical meaning when the subsequent orthogonal matching pursuit algorithm performs inner product operations. Simultaneously, temporal deconvolution processing effectively suppresses irrelevant noise by utilizing the difference in correlation characteristics between signal and noise, improving the quality of the observation vector under low signal-to-noise ratio conditions.

[0073] S3. Assign the initial value of the residual vector to the time-deconvolution observation vector, and initialize the sparse vector to the zero vector.

[0074] Orthogonal matching pursuit is a greedy iterative sparse reconstruction algorithm. In each iteration, the atom that best matches the current residual vector is selected from the overcomplete dictionary. Then, the sparse coefficients are updated using the least squares method with the selected atom, and the residual vector is updated as well. Step S3 aims to set the initial state for the entire iterative process.

[0075] Specifically, the temporal deconvolution observation vector constructed in step S2... (dimension) Assign initial values ​​to the residual vector ,Right now .

[0076] At the same time, initialize a sparse vector. Its dimensions are (Corresponding to 360 preset target orientations), all elements are initialized to zero. This sparse vector is used to record the actual amplitude corresponding to the selected target orientation in each iteration.

[0077] Initialize the selected index set (Empty set), used to store the index of the dictionary column selected in each iteration. Initialize the number of iterations. .

[0078] S4. Calculate the inner product of each column of the deconvolution sound field dictionary with the current residual vector, and select the column index corresponding to the largest absolute value of the inner product to store in the selected index set; use the dictionary columns corresponding to the selected index set to form a submatrix, and solve for the true amplitude corresponding to the target azimuth using the least squares method; fill the obtained true amplitude into the sparse vector at the position corresponding to the selected index set; update the residual vector, and repeat the above process of step S4 until the iteration stopping condition is met, and take the target azimuth corresponding to the largest amplitude in the final sparse vector as the azimuth estimation result.

[0079] More specifically, the implementation of step S4 includes the following sub-steps: First, perform time-deconvolution of the sound field dictionary. Each column and the current residual vector Perform the inner product operation to obtain the inner product vector. The larger the absolute value of the inner product, the higher the degree of matching between the dictionary column (corresponding to a certain preset target orientation) and the current residual vector.

[0080] Secondly, from the inner product vector Find the element with the largest absolute value and record its corresponding column index. Store the index in the selected index set. This ensures that the same angle is not selected again in subsequent iterations.

[0081] Next, select the dictionary columns corresponding to all historical selected angles to form a submatrix. The following optimization problem is solved using the least squares method: In the spanned space of the selected dictionary columns, find the coefficient vector that best approximates the observed vector, i.e., solve for the true magnitude. The calculated true magnitude is then filled into the sparse vector. In and Selected Index Set The corresponding position.

[0082] Update residual vector Determine whether the norm of the current residual vector is less than a preset threshold. Alternatively, check if the number of iterations has reached the preset sparsity constraint value. If either condition is met, the iteration stops; otherwise, let... Then return to the previous sub-step and continue iterating. After the iteration stops, in the sparse vector... Find the element with the largest amplitude; the target azimuth corresponding to this element is the final azimuth estimate.

[0083] In one specific implementation: inner product vector The calculation method is as follows:

[0084] in, Time-deconvolution sound field dictionary transpose, This is the current residual vector. The dimension of the inner product vector is... Each element corresponds to a correlation coefficient between a preset target orientation and the current residual vector.

[0085] The optimal index is selected as follows:

[0086] That is, take the column index corresponding to the maximum absolute value of the inner product.

[0087] Let the selected index set be... The submatrix formed by the corresponding dictionary columns is Then the expression for solving the true magnitude using least squares is:

[0088] Or equivalent to:

[0089] in, The initial value of the residual vector is the deconvolution observation vector. ; This is the solution to obtain the true magnitude vector.

[0090] For example, the method for determining the iteration stopping condition is as follows:

[0091] in, For a preset small positive number (such as This indicates that the residual energy is small enough that the currently selected atom combinations can fully explain the observation vector. Alternatively, the iteration stops when the number of iterations reaches a preset sparsity constraint value, such as a preset target number.

[0092] Finally, in sparse vectors Find the element with the largest amplitude:

[0093] The target orientation corresponding to this element is the final orientation estimation result.

[0094] This application also provides a high-resolution directional system for a vector hydrophone, including a single-vector hydrophone, a signal acquisition module, and a signal processing module.

[0095] For example, a single-vector hydrophone is composed of a point-to-point composite of an omnidirectional acoustic pressure hydrophone and a particle velocity hydrophone. The particle velocity hydrophone has mutually orthogonal x-direction and y-direction channels for measuring the two orthogonal horizontal components of the particle velocity in the sound field. The output terminal of the acoustic pressure hydrophone is used to output the acoustic pressure channel signal, and the x-direction and y-direction channel output terminals of the particle velocity hydrophone are used to output the x-direction and y-direction velocity channel signals, respectively.

[0096] The input terminals of the signal acquisition module are connected to the output terminals of the acoustic pressure hydrophone, the x-channel output terminal of the particle velocity hydrophone, and the y-channel output terminal, respectively, for synchronous sampling and analog-to-digital conversion of the three output signals. The signal acquisition module typically includes a preamplifier, an anti-aliasing filter, and an analog-to-digital converter.

[0097] The input of the signal processing module is connected to the output of the signal acquisition module. The signal processing module internally includes a line spectrum detection unit and an orthogonal matched pursuit sparse reconstruction unit. The line spectrum detection unit performs line spectrum detection on the sound pressure channel signal to obtain the target line spectrum frequency. The orthogonal matched pursuit sparse reconstruction unit performs signal processing operations from steps S1 to S4 based on the target line spectrum frequency. The output of the signal processing module outputs the target azimuth estimation result.

[0098] For example, refer to Figure 3 , Figure 3 This is a comparison chart of the orientation results of different vector azimuth estimation algorithms when the signal-to-noise ratio is 5dB and the signal length is 0.1s. The horizontal axis represents the target azimuth angle, and the vertical axis represents the normalized azimuth spectrum amplitude. The curves of different shapes correspond to various comparison algorithms. Figure 3The resolution performance of each algorithm is intuitively demonstrated by the azimuth spectrum morphology. Traditional algorithms have the problem of wide main lobe and high side lobe level, while the method of this invention can form sharp spectral peaks and has better side lobe suppression effect. Under the low signal-to-noise ratio condition, it can more accurately align with the true azimuth of the target, demonstrating better spatial resolution capability.

[0099] For example, refer to Figure 4 , Figure 4 The graph shows the performance curves of different vector azimuth estimation algorithms under a signal-to-noise ratio (SNR) condition of 0.1s signal length. The horizontal axis represents SNR (dB), and the vertical axis represents the root mean square error (RMSE) of azimuth estimation (degrees). Each curve represents the variation of the angle estimation error of different algorithms with SNR. As the SNR decreases, the azimuth estimation error of the comparative algorithms increases significantly. The curve corresponding to this invention is at a lower position overall, with a smaller RMS error. This quantitatively verifies that the method provided in this application can maintain high orientation accuracy in low SNR environments and has a more outstanding noise adaptability compared to existing algorithms.

[0100] In the description of this specification, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0101] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A high-resolution orientation method for a vector hydrophone, characterized in that, include: S1. For each preset target orientation, traverse all preset search angles, and use the ideal received signals from the vector hydrophone sound pressure channel, x-direction vibration velocity channel and y-direction vibration velocity channel to perform channel combination weighting. Then, convolve the combined signal with the corresponding time-reversed combined signal and transform it to the frequency domain. Extract the complex amplitude at the target line spectrum frequency and splice them in order to form the dictionary atom corresponding to the current target orientation. Iterate through all preset target locations and combine all dictionary atoms to obtain a time-deconvolution sound field dictionary; S2. Filter the measured vector hydrophone sound pressure channel signal, x-direction vibration velocity channel signal, and y-direction vibration velocity channel signal. For each preset search orientation, combine and weight the filtered three signals, and then convolve the combined signal with the corresponding time-reversed combined signal and transform it to the frequency domain. Extract the complex amplitude at the target line spectrum frequency. After traversing all preset search orientations, concatenate them sequentially to obtain the time-deconvolution observation vector. S3. Assign the time-deconvolution observation vector the initial value of the residual vector and initialize the sparse vector to zero. S4. Calculate the inner product of each column of the temporal deconvolution sound field dictionary with the current residual vector, and select the column index corresponding to the largest absolute value of the inner product and store it in the selected index set; use the dictionary columns corresponding to the selected index set to form a sub-matrix, and solve for the true amplitude corresponding to the target azimuth using the least squares method; fill the solved true amplitude into the sparse vector at the position corresponding to the selected index set; update the residual vector, and repeat the above process of step S4 until the iteration stopping condition is met, and take the target azimuth corresponding to the largest amplitude in the final sparse vector as the azimuth estimation result.

2. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, In step S1, forming the dictionary atoms corresponding to the current target orientation specifically includes: for the current target orientation, using the ideal sound pressure channel signal, the ideal x-direction vibration velocity channel signal, and the ideal y-direction vibration velocity channel signal, weighting the channel combination according to the currently traversed search angle to obtain a first combination signal and a second combination signal; convolving the sum of the ideal sound pressure channel signal and the first combination signal with the time-reversed signals of the first combination signal and the second combination signal respectively to obtain two time-reversed convolution time-domain signals; transforming the two time-reversed convolution time-domain signals to the frequency domain, extracting the complex amplitude at the target line spectrum frequency as the atomic element corresponding to the current search angle, and sequentially splicing them together after traversing all search angles to form the dictionary atoms.

3. The high-resolution orientation method for a vector hydrophone according to claim 2, characterized in that, The channel combination weighting method includes: summing the product of the x-direction vibration velocity signal and the cosine of the current search angle with the product of the y-direction vibration velocity signal and the sine of the current search angle to obtain the first combined signal; and summing the negative value of the product of the x-direction vibration velocity signal and the sine of the current search angle with the product of the y-direction vibration velocity signal and the cosine of the current search angle to obtain the second combined signal.

4. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, Step S2, which involves extracting the complex amplitude at the target line spectrum frequency, specifically includes: for the current preset search azimuth, combining and weighting the measured sound pressure channel signal, x-direction vibration velocity channel signal, and y-direction vibration velocity channel signal to obtain a first measured combined signal and a second measured combined signal; convolving the sum of the measured sound pressure channel signal and the first measured combined signal with the time-reversed signals of the first and second measured combined signals, respectively, to obtain two measured time-deconvolutioned time-domain signals; transforming the two measured time-deconvolutioned time-domain signals to the frequency domain, and extracting the complex amplitude at the target line spectrum frequency as the observation vector element corresponding to the current search azimuth.

5. The high-resolution orientation method for a vector hydrophone according to claim 4, characterized in that, The filtering process is an intra-band filtering, and the frequency band range is determined by the target line spectrum frequency.

6. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, The iteration stopping condition in step S4 is that the norm of the current residual vector is less than a preset threshold, or the number of iterations reaches a preset sparsity constraint value.

7. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, The expression for solving the true amplitude using the least squares method in step S4 is as follows: in, The submatrix formed by the dictionary columns corresponding to the selected index set. The initial value of the residual vector is the time-deconvolution observation vector. To obtain the true magnitude vector, for The transpose of .

8. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, The target line spectrum frequency is obtained by line spectrum detection of the actual measured vector hydrophone sound pressure channel signal.

9. The high-resolution orientation method for a vector hydrophone according to claim 1, characterized in that, The preset target orientation and the preset search angle both have an angle range of 0° to 360°, and the interval between adjacent angles is 1°.

10. A high-resolution orientation system for a vector hydrophone, used to perform the high-resolution orientation method for a vector hydrophone according to any one of claims 1 to 9, characterized in that, include: A single-vector hydrophone is constructed by combining an omnidirectional acoustic pressure hydrophone and a particle velocity hydrophone. The particle velocity hydrophone has mutually orthogonal x-direction and y-direction channels. The output terminal of the acoustic pressure hydrophone is used to output the acoustic pressure channel signal, and the x-direction channel output terminal and y-direction channel output terminal of the particle velocity hydrophone are used to output the x-direction vibration channel signal and the y-direction vibration channel signal, respectively. The signal acquisition module has its input terminals connected to the output terminals of the acoustic pressure hydrophone, the x-direction channel output terminal and the y-direction channel output terminal of the particle velocity hydrophone, respectively, and is used to synchronously sample and convert the three output signals into analog and digital signals. The signal processing module has its input connected to the output of the signal acquisition module. The signal processing module internally includes a line spectrum detection unit and an orthogonal matched pursuit sparse reconstruction unit. The line spectrum detection unit performs line spectrum detection on the sound pressure channel signal to obtain the target line spectrum frequency. The orthogonal matched pursuit sparse reconstruction unit performs the signal processing operations from steps S1 to S4 based on the target line spectrum frequency. The output of the signal processing module outputs the target azimuth estimation result.

Images