A deep-sea volume array target azimuth estimation method based on array geometry decoupling

Abstract

This invention discloses a method for estimating the azimuth of deep-sea volumetric array targets based on array geometry decoupling, belonging to the field of array signal processing. The method includes acquiring data matrices for a vertical linear array and a horizontal circular array; constructing the elevation angle spatial spectrum of the vertical linear array using a minimum variance distortion-free response algorithm to extract elevation angle estimates for each sound source; extracting the signal subspace vectors of each sound source using a stepwise orthogonal projection to obtain the signal subspace projection matrix of the horizontal circular array; constructing orthogonal projection azimuth angle spatial spectra under the constraints of each elevation angle estimate; sequentially extracting the coarse azimuth angle estimates corresponding to each elevation angle estimate; combining this with a circular protection zone to suppress repetitive peaks; performing a local fine-grid search on the coarse azimuth angle estimates to obtain fine azimuth angle estimates, which are then paired and output as two-dimensional direction-of-arrival (DOA) estimation results for multiple sound sources. This invention improves the accuracy and stability of two-dimensional DOA estimation for multiple targets.

Description

Technical Field

[0001] This invention relates to the field of array signal processing, and specifically to a method for estimating the azimuth of a deep-sea volume array target based on array geometric decoupling. Background Technology

[0002] Underwater acoustic array signal processing is one of the core technologies in underwater communication, navigation, and other fields. By processing the signals received by spatially distributed multi-element hydrophone arrays, the location information of underwater sound sources can be extracted, thereby serving applications such as marine scientific research, environmental monitoring, and autonomous sensing of underwater operating platforms.

[0003] Direction of Arrival (DOA) estimation is a crucial step in underwater acoustic array signal processing. Its principle is to invert the direction of arrival of the sound source signal at the array by utilizing the phase difference or time delay difference of the received signals between different array elements. In typical shallow sea or homogeneous media environments, the sound propagation path is nearly straight. Traditional two-dimensional spatial spectrum DOA estimation methods are based on far-field plane wave models. By constructing a two-dimensional joint steering vector related to the azimuth and elevation angles, and performing a two-dimensional spectrum search in the spatial angular domain, the source direction can be estimated relatively accurately.

[0004] However, when the aforementioned direction-of-arrival (DOA) estimation method is applied to deep-sea low-frequency acoustic environments, its performance degrades significantly due to waveguide mismatch. In the deep-sea environment, the sea surface, seabed, and non-uniform sound velocity profiles varying with depth together constitute a complex acoustic propagation waveguide structure. When a sound source signal propagates in this waveguide, it excites multiple normal mode modes, each with different horizontal and vertical wave numbers. The sound field actually received by the array is the result of the superposition of these multiple modes, and its spatial phase distribution is determined by the physical characteristics of the waveguide, rather than simply corresponding to a single phase delay relationship under the free-field plane wave assumption. If the ideal plane waveguide vector established based on the reference sound velocity is still used for DOA estimation, a non-negligible mismatch will occur between the actual array manifold and the scanning guide vector, leading to a series of problems such as spectral peak shift in the two-dimensional spatial spectrum, spectral peak broadening, decreased resolution of weak target directions, and reduced reliability of multi-target angle estimation results.

[0005] To address the deep-sea waveguide mismatch problem, existing technologies have mainly improved the technology from two aspects: array geometry design and array signal subspace estimation.

[0006] In terms of array geometry design, existing technologies attempt to combine vertical linear arrays with horizontal circular arrays. The vertical linear array is used to estimate the target's elevation angle, which is then used to constrain the azimuth search range of the horizontal circular array. This decomposes the two-dimensional joint angle search process into a sequentially executed one-dimensional search process, reducing the computational complexity of the two-dimensional angle search. However, the core objective of these methods is to reduce computational load. Their technical solutions still rely on the premise of ideal plane wave propagation and the matching of the array steering vector with the actual spatial characteristics of the received signal. They do not provide a corresponding mechanism for addressing the array manifold mismatch problem in deep-sea waveguide environments.

[0007] In the estimation of array signal subspaces, existing techniques attempt to reduce the computational complexity of subspace estimation through orthogonal projection decomposition methods. For example, by constructing array signal subspace basis vectors through successive orthogonal projections, the traditional eigenvalue decomposition operation can be replaced. However, these methods lack branched constraints on elevation and azimuth estimation, making it difficult to effectively suppress spectral peak crosstalk and mispairing between different sound sources under multi-target conditions. As a result, the stability and accuracy of azimuth estimation in deep-sea waveguide environments remain unsatisfactory.

[0008] Therefore, how to achieve reliable pairing of pitch and azimuth results under deep-sea volumetric array conditions, and improve the accuracy and stability of two-dimensional direction-of-arrival estimation for multiple targets, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0009] To address the shortcomings of existing technologies, this invention provides a deep-sea volumetric array target azimuth estimation method based on array geometry decoupling. By decomposing the two-dimensional angle joint search into two stages—elevation branch extraction and one-dimensional azimuth search within the branch—and introducing cross-elevation branch circumferential protection and local refinement search, the method mitigates the impact of deep-sea waveguide array manifold mismatch, suppresses multi-target angle pairing errors, and improves the accuracy and stability of two-dimensional direction of arrival estimation.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] A method for estimating the azimuth of a deep-sea volumetric array target based on array geometry decoupling includes the following steps:

[0012] Step 1: Construct a volume array consisting of a vertical linear array and a horizontal circular array, acquire the received data of the volume array, extract the vertical linear array data matrix and the horizontal circular array data matrix from the received data of the volume array, calculate the sample covariance matrix of the vertical linear array based on the vertical linear array data matrix, and calculate the sample covariance matrix of the horizontal circular array based on the horizontal circular array data matrix.

[0013] Step 2: Construct the vertical linear array pitch angle spatial spectrum based on the vertical linear array sample covariance matrix, and extract the pitch angle estimates of multiple sound sources from the vertical linear array pitch angle spatial spectrum in combination with the pitch angle protection zone.

[0014] Step 3: Based on the covariance matrix of the horizontal circular array samples, the signal subspace vector of each sound source is extracted using the stepwise orthogonal projection method, and the horizontal circular array signal subspace projection matrix is ​​constructed; according to the horizontal circular array signal subspace projection matrix, the orthogonal projection azimuth space spectrum is constructed under the constraints of each pitch angle estimate.

[0015] Step 4: In the spatial spectrum of azimuth angles of each orthogonal projection, suppress the repeated peaks under different pitch angle estimates by using the circular protection zone, and extract the coarse azimuth angle estimates corresponding to each pitch angle estimate.

[0016] Step 5: Perform a local fine-grid search on the coarse azimuth estimate to obtain a fine azimuth estimate. Pair each elevation angle estimate with the corresponding fine azimuth estimate to output the two-dimensional direction of arrival estimation results for multiple sound sources.

[0017] Furthermore, step 2 specifically includes:

[0018] The covariance matrix of the vertical linear array sample is subjected to conjugate symmetry processing to obtain the conjugate symmetric vertical linear array sample covariance matrix.

[0019] The covariance matrix of the conjugate symmetric vertical linear array sample is diagonally loaded to obtain the diagonally loaded vertical linear array sample covariance matrix.

[0020] For the elevation angle to be scanned, a vertical linear array guide vector is constructed. Each element of the vertical linear array guide vector is determined by the reference wavenumber, the relative depth of each element of the vertical linear array, and the cosine value of the elevation angle to be scanned.

[0021] The vertical linear array steering vector is normalized to obtain the normalized vertical linear array steering vector.

[0022] Based on the normalized vertical linear array steering vector and the sample covariance matrix of the diagonally loaded vertical linear array, the pitch angle spatial spectrum of the vertical linear array is constructed using the minimum variance distortionless response algorithm.

[0023] The maximum spectral peak is extracted sequentially from the vertical linear array pitch angle spatial spectrum. After each extraction, a pitch angle protection zone is set with the currently extracted spectral peak as the center, and the spectral values ​​within the pitch angle protection zone are set to negative infinity to update the current vertical linear array pitch angle spatial spectrum. The next extraction of the maximum spectral peak is performed in the updated pitch angle spatial spectrum until the pitch angle estimates of multiple sound sources are obtained.

[0024] Further, in step 3, based on the covariance matrix of the horizontal circular array samples, a successive orthogonal projection method is used to extract the signal subspace vector of each sound source, and a horizontal circular array signal subspace projection matrix is ​​constructed, specifically including:

[0025] The horizontal circular array sample covariance matrix is ​​subjected to conjugate symmetry processing to obtain the conjugate symmetric horizontal circular array sample covariance matrix.

[0026] Initialize the orthogonal complement projection matrix;

[0027] For the Secondary signal subspace vector extraction, where , Given the number of sound sources, calculate the projection covariance matrix in the current orthogonal complement space. The projection covariance matrix is ​​calculated based on the current orthogonal complement projection matrix and the sample covariance matrix of the conjugate symmetric horizontal circular array.

[0028] The projection covariance matrix is ​​subjected to conjugate symmetry processing to obtain the conjugate symmetric projection covariance matrix;

[0029] Find the unit eigenvector corresponding to the largest eigenvalue of the conjugate symmetric projective covariance matrix;

[0030] Project the unit feature vector onto the current orthogonal complement space to obtain the projection vector;

[0031] Normalize the projection vector to obtain the first... The signal subspace vector of each sound source;

[0032] Update the orthogonal complement projection matrix by subtracting the product of the currently extracted signal subspace vector and its conjugate transpose from the current orthogonal complement projection matrix;

[0033] The updated orthogonal complement projection matrix is ​​used for the next extraction of signal subspace vectors;

[0034] Repeat execution The next signal subspace vector extraction process yields... A horizontal circular array of signal subspaces composed of signal subspace vectors;

[0035] The horizontal circular array signal subspace is orthogonalized using QR to obtain a signal subspace basis matrix with mutually orthogonal column vectors. The horizontal circular array signal subspace projection matrix is ​​then constructed based on the signal subspace basis matrix.

[0036] Furthermore, ,in This represents the number of elements in the vertical linear array. This represents the number of elements in a horizontal circular array.

[0037] Furthermore, in step 3, the process of constructing the orthogonal projection azimuth spatial spectrum under the constraints of each pitch angle estimate includes:

[0038] For the One pitch angle estimate, , To determine the number of sound sources, the direction of the sound sources is calculated based on the current estimated elevation angle and the azimuth angle to be scanned. Direction cosines along the axis Direction cosine along the axial direction;

[0039] Construct the horizontal circular array steering vector based on the reference wavenumber, the circumference angle of the horizontal circular array element, the radius of the horizontal circular array, the current elevation angle estimate, and the azimuth angle to be scanned.

[0040] The horizontal circular array guide vector is normalized to obtain the normalized horizontal circular array guide vector.

[0041] Calculate the projection energy of the normalized horizontal circular array steering vector in the horizontal circular array signal subspace;

[0042] The projected energy is subjected to amplitude limiting processing;

[0043] Construct the orthogonal projection azimuth spatial spectrum corresponding to the current pitch angle estimate based on the projection energy after amplitude limiting.

[0044] Furthermore, step 4 specifically includes:

[0045] Set a coarse search interval and perform a coarse grid search within the azimuth angle range of 0° to 360°;

[0046] For the first pitch angle estimate, the azimuth angle corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum is extracted as the coarse azimuth angle estimate;

[0047] Define any two azimuth angles and circumferential distance between for:

[0048]

[0049] Set a circular protection zone half the width;

[0050] For the One pitch angle estimate, , To determine the number of sound sources, a circular protection zone is established centered on the previously extracted coarse azimuth estimates. Only outside the scope of the circular protection zone, the azimuth corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum is extracted as the coarse azimuth estimate of the current elevation angle estimate.

[0051] Furthermore, step 5 specifically includes:

[0052] For each pitch angle estimate, a corresponding local azimuth fine search set is established centered on the coarse azimuth angle estimate. The local azimuth fine search set is determined by the local search half-width and the fine search interval, and the fine search interval is smaller than the coarse search interval.

[0053] For the first pitch angle estimate, the corresponding local azimuth angle fine search set is used as the allowed fine search set;

[0054] For the One pitch angle estimate, Circular protection zones are set up centered on each previously obtained fine azimuth angle estimate. The circular protection zones are removed from the corresponding local azimuth angle fine search set to obtain the allowable fine search set.

[0055] In the set of allowed fine search, the corresponding pitch angle estimate is kept unchanged, and a fine grid search is performed at a fine search interval to extract the azimuth angle corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum as the fine azimuth angle estimate.

[0056] After extracting fine azimuth estimates for each elevation angle estimate, the elevation angle estimates are paired with the corresponding fine azimuth estimates to output the two-dimensional direction of arrival (DOA) estimates for multiple sound sources.

[0057] Furthermore, the upper limit for the projection energy during amplitude limiting is: .

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] (1) This invention combines the division of labor between vertical linear arrays and horizontal circular arrays in array geometry with the physical constraints between wavenumber projection components. First, it utilizes the sensitivity of the vertical linear array to the vertical wavenumber component to extract the elevation angle estimate, and then uses the elevation angle estimate as a constraint to construct the azimuth scanning steering vector of the horizontal circular array. Since there is an inherent sum of squares relationship between the total wavenumber, horizontal wavenumber and vertical wavenumber of the same mode in the real sound field, the corresponding horizontal wavenumber projection factor can be determined from the elevation angle estimate, making the horizontal wavenumber component used by the steering vector of the horizontal circular array closer to the horizontal wavenumber of the real propagation mode. This reduces the degree of mismatch between the real array manifold and the conventional plane wave scanning steering vector caused by multimode superposition in deep-sea waveguides, and reduces the spectral peak shift and spectral peak broadening caused by array manifold mismatch in the two-dimensional spatial spectrum.

[0060] (2) This invention extracts the signal subspace vector corresponding to each sound source sequentially through a stepwise orthogonal projection method. After each extraction, the extracted signal components are removed from the orthogonal complement space, ensuring that the subsequently extracted signal subspace vectors remain orthogonal. Since the signal subspace vectors of different sound sources are orthogonal to each other, during the subsequent azimuth space spectrum construction process, the spectral peaks of each sound source will not leak or crosstalk due to the non-orthogonality of the signal subspace direction. This maintains the independence and distinguishability of the azimuth spectral peaks of each sound source in the deep-sea multi-target environment, improving the detection capability and azimuth estimation stability of weak targets under strong target sidelobe interference.

[0061] (3) When extracting the coarse azimuth estimates corresponding to each elevation angle estimate in sequence, the present invention establishes a circular protection zone centered on the extracted azimuth angle, so that the subsequent azimuth angle search under the elevation angle estimate is only carried out outside the protection zone; and in the local fine search stage, the circular protection zone is maintained centered on the obtained fine azimuth angle estimate. Since the same sound source will produce response peaks with similar positions in the azimuth spatial spectrum of adjacent elevation angle estimates, and the interval between different sound sources in azimuth angle is usually large, the circular protection zone can exclude the spectral peaks within a certain range around the extracted azimuth angle, which can effectively suppress the main lobe, side lobe or leakage peak of the same strong sound source from being repeatedly extracted as false targets under different elevation angle estimates, while not affecting the azimuth angle extraction of real sound sources under different elevation angle estimates, thereby reducing the angle mismatch rate and false target detection rate in the two-dimensional direction of arrival estimation of multiple targets.

[0062] (4) This invention performs a local fine-grid search on the coarse azimuth estimates corresponding to each pitch angle estimate. Centered on the coarse azimuth estimate, the maximum spectral peak of the orthogonal projection azimuth spatial spectrum is extracted within a local range with a smaller step size. Since the azimuth scanning interval used in the coarse search stage will cause discretization and quantization errors in the estimation results, the local fine search reduces the scanning step size from the coarse search interval to the fine search interval, which effectively improves the positioning accuracy of the spectral peak. At the same time, the search range is limited to a local interval near the coarse azimuth estimate, and the computational load is much smaller than that of the global fine search. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the volumetric array configuration in this invention;

[0064] Figure 2 This is a diagram showing the distribution of propagation loss with radial distance at different depths in the deep-sea environment used in the simulation of this invention embodiment;

[0065] Figure 3 The image shows the two-dimensional direction-of-arrival estimation results obtained using the method of this invention under single sound source conditions.

[0066] Figure 4The elevation angle spatial spectrum of a vertical linear array obtained using the method of this invention under single sound source conditions;

[0067] Figure 5 This is an orthogonal projection azimuth spatial spectrum obtained using the method of this invention under single sound source conditions;

[0068] Figure 6 The image shows the two-dimensional direction of arrival estimation results obtained by the direct two-dimensional orthogonal projection method under dual sound source conditions.

[0069] Figure 7 The elevation angle spatial spectrum of the vertical linear array obtained by the method of this invention under dual sound source conditions;

[0070] Figure 8 The orthogonal projection azimuth spatial spectrum obtained by the method of this invention under dual sound source conditions;

[0071] Figure 9 The image shows the two-dimensional direction of arrival estimation results obtained using the method of this invention under dual sound source conditions. Detailed Implementation

[0072] 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Example

[0074] This embodiment provides a method for estimating the azimuth of a deep-sea volumetric array target based on array geometry decoupling, which is performed according to the following process:

[0075] Step 1: Construct a volume array consisting of a vertical linear array and a horizontal circular array, refer to... Figure 1 The horizontal circular array is located at the deepest position of the vertical linear array, and the center of the horizontal circular array coincides with the position of the lowest element of the vertical linear array.

[0076] With the center of the horizontal circular array as the origin Establish a three-dimensional rectangular coordinate system, where, The plane is a horizontal plane. The positive axis points vertically upwards. The pitch angle is defined using spherical coordinates. Indicates the direction of the sound source and The angle between the positive directions of the axis, Indicates that the sound source is located at Positive axis direction Indicates that the sound source is located in the horizontal plane; azimuth angle The unit direction vector corresponding to the direction of the sound source is in Projection in the plane and The angle between the positive directions of the axis has a range of values. .

[0077] Unit direction vector corresponding to the direction of the sound source for:

[0078]

[0079] Vertical linear array includes There are 1 array element, and the distance between adjacent array elements is 1. Let the center of the horizontal circular array be... The axis coordinates are Then the first ( ) vertical linear array elements Axis coordinates A horizontal circular array contains There are [number] array elements, and the radius of the horizontal circular array is [radius]. .by Starting from the positive direction of the axis, the first ( The circular angles corresponding to the elements of the horizontal circular array .

[0080] Let the center frequency of the signal received by the volume array be... The reference speed of sound is Then refer to wavenumber In this embodiment Take 1500 m / s.

[0081] The narrowband underwater acoustic signals received by each element of the volumetric array are acquired, and the received narrowband underwater acoustic signals are sampled to obtain the volumetric array observation data matrix. Assume that a total of [number missing] data are acquired during the observation period. Each snapshot is a complex sampled value of all array elements at the same time. Therefore, the volumetric array observation data matrix... ,in For the first ( ( ) Snapshots 3D column vector, This represents the total number of elements in the volumetric array.

[0082] In this embodiment, the volume array observation data matrix The former The rows correspond to the vertical linear array Each array element, later The row corresponds to the horizontal circular array Each array element.

[0083] Extract the previous data from the volumetric array observation data matrix. Rows of data form a vertical linear array data matrix. After extraction The rows of data form a horizontal circular matrix. .

[0084] Based on the vertical linear array data matrix Calculate the sample covariance matrix of the vertical linear array :

[0085]

[0086] In the formula, This is the conjugate transpose operation.

[0087] Based on the horizontal circular array data matrix Calculate the sample covariance matrix of the horizontal circular array :

[0088]

[0089] Step 2: Construct the vertical linear array elevation angle spatial spectrum based on the covariance matrix of the vertical linear array samples, and extract the elevation angle estimates of multiple sound sources. The specific process is as follows:

[0090] For the sample covariance matrix of the vertical linear array Perform conjugate symmetry processing to obtain the conjugate symmetric vertical linear array sample covariance matrix. :

[0091]

[0092] To improve the stability of covariance matrix inversion, the sample covariance matrix of the conjugate symmetric vertical linear array is calculated. diagonal loading Utilizing diagonal loading For the sample covariance matrix of the conjugate symmetric vertical linear array Diagonal loading is performed to obtain the sample covariance matrix of the diagonally loaded vertical linear array. :

[0093]

[0094]

[0095] In the formula, For the trace operation of a matrix, for An identity matrix of order 1.

[0096] For the elevation angle to be scanned Its vertical cosine is Vertical linear array guide vector for:

[0097]

[0098] In the formula, The imaginary unit, For the vertical linear array Individual elements Relative coordinates of the axes .

[0099] For vertical linear array steering vector Normalization is performed to obtain the normalized vertical linear array steering vector. :

[0100]

[0101] In the formula, It is a 2-norm. To prevent the regularization constant from having a denominator of zero, in this embodiment, .

[0102] Constructing the elevation spatial spectrum of a vertical linear array using the minimum variance distortionless response algorithm :

[0103]

[0104] In the formula, To perform the operation of taking the real part, for The inverse matrix.

[0105] The maximum spectral peak is extracted successively from the elevation angle spatial spectrum of the vertical linear array. The specific process is as follows:

[0106] For the Secondary peak extraction, among which , The number of sound sources is preset. Extract the pitch angle estimate corresponding to the largest spectral peak from the current pitch angle spatial spectrum. :

[0107]

[0108] In the formula, The range of the elevation angle scan. For the first The current pitch angle spatial spectrum at the time of the next extraction. for The elevation angle scanning interval is 0.1°;

[0109] After extraction, with A pitch angle protection zone is set up around the center, with a half-width of the protection zone being [missing information]. In this embodiment Setting the spectral values ​​within the protected area to negative infinity yields the first... The elevation angle spatial spectrum used in the second extraction :

[0110]

[0111] Repeat the above peak extraction process until the peaks are obtained. Each pitch angle estimate will be obtained Arrange the pitch angle estimates in ascending order to obtain the pitch angle estimate set. :

[0112]

[0113] In the formula, for The Middle One pitch angle estimate, ,and .

[0114] In this step, sorting the pitch angle estimates is used to determine the processing order of multiple pitch angle estimates.

[0115] Step 3: Based on the covariance matrix of the horizontal circular array samples, extract the signal subspace vector of each sound source using a stepwise orthogonal projection method, and construct the horizontal circular array signal subspace projection matrix. The specific process is as follows:

[0116] Covariance matrix of horizontal circular array samples Perform conjugate symmetry processing to obtain the sample covariance matrix of the conjugate symmetric horizontal circular matrix. :

[0117]

[0118] Initialize the orthogonal complement projection matrix:

[0119]

[0120] In the formula, for An identity matrix of order 1. This is the initial orthogonal complement projection matrix when no signal subspace vectors have been extracted. At this point, the orthogonal complement space is... Dimensional space.

[0121] For the Secondary signal subspace vector extraction, where Perform the following sub-steps:

[0122] Calculate the projection covariance matrix in the current orthogonal complement space. :

[0123]

[0124] In the formula, For the first The orthogonal complementary projection matrix after the second update.

[0125] For the projection covariance matrix Perform conjugate symmetry transformation to obtain the conjugate symmetric projective covariance matrix. :

[0126]

[0127] Seeking The unit eigenvector corresponding to the largest eigenvalue :

[0128]

[0129] In the formula, These are the candidate unit eigenvectors.

[0130] unit eigenvector Projecting onto the current orthogonal complement space yields the projection vector. :

[0131]

[0132] For projection vector Normalization is performed to obtain the first... The signal subspace vector of each sound source :

[0133]

[0134] Subtract the product of the currently extracted signal subspace vector and its conjugate transpose from the current orthogonal complement space to remove the components along the direction of the signal subspace vector, and update the orthogonal complement projection matrix:

[0135]

[0136] Repeat execution The above signal subspace vector extraction process yields the horizontal circular array signal subspace. :

[0137]

[0138] For the horizontal circular array signal subspace Perform QR orthogonalization to obtain the signal subspace basis matrix with mutually orthogonal column vectors. :

[0139]

[0140] In the formula, It is an upper triangular matrix.

[0141] According to the signal subspace basis matrix Constructing the horizontal circular array signal subspace projection matrix :

[0142]

[0143] For each estimated elevation angle, a one-dimensional orthogonal projection azimuth spatial spectrum is constructed. The specific process is as follows:

[0144] For pitch angle estimation set The Middle One pitch angle estimate The azimuth angle to be scanned is The direction of the sound source is in Direction cosine along the axis and Direction cosine along the axis They are respectively:

[0145]

[0146]

[0147] Since all elements of the horizontal circular array are at the same depth, the vertical phase term is a common phase term and does not affect the position of the azimuth spectral peak. Therefore, the elevation angle estimate is... Corresponding horizontal circular array guiding vector for:

[0148]

[0149] In the formula, For the horizontal circular array Individual elements The coordinates of the axis, For the horizontal circular array Individual elements The coordinates of the axis. ; .

[0150] Horizontal circular array guiding vector Normalization is performed to obtain the normalized horizontal circular matrix steering vector. :

[0151]

[0152] Calculate the normalized horizontal circular array steering vector Projected energy in the horizontal circular array signal subspace :

[0153]

[0154] For projected energy Amplitude limiting is applied to ensure numerical stability:

[0155]

[0156] Construction pitch angle estimation The corresponding orthogonal projection azimuth space spectrum :

[0157]

[0158] In the formula, In this embodiment, the peak sharpening index is taken as... To enhance the identifiability of azimuth peaks without altering their positions.

[0159] Step 4: In the spatial spectrum of orthogonal projection azimuth corresponding to each pitch angle estimate, suppress the recurring peaks under different pitch angle estimates through the circular protection zone, and extract the coarse azimuth estimate corresponding to each pitch angle estimate, specifically including:

[0160] A coarse-grid search was performed within the azimuth angle range of 0 to 359.75°, resulting in the azimuth coarse-grid search set. for:

[0161]

[0162] in, As a coarse search interval, this embodiment takes... .

[0163] For the first pitch angle estimate Extract the spatial spectrum of orthogonal projection azimuth angle Coarse azimuth estimate corresponding to the maximum spectral peak :

[0164]

[0165] To avoid the extracted sound source direction being repeatedly selected in subsequent pitch angle estimates, we define any two azimuth angles. and circumferential distance between for:

[0166]

[0167] Let the half-width of the circular protected area be In this embodiment, .

[0168] For the One pitch angle estimate Using the previously extracted coarse azimuth estimates respectively Establish a circumferential protection zone around the center, and remove azimuth angles falling within the circumferential protection zone from the coarse azimuth search set, resulting in... Corresponding allowed search set :

[0169]

[0170] In the formula, This is the index for the extracted coarse azimuth estimates.

[0171] In the allowed search set Periodic local peak detection is performed in the process, assuming... To allow searching the set of the first There are azimuth angles to be scanned, and their left and right adjacent azimuth angles to be scanned are arranged according to... If the beginning and end of the period are connected, then the first... The set of local peaks corresponding to each pitch angle estimate for:

[0172]

[0173] When local peaks aggregate When not empty, the first One pitch angle estimate Corresponding coarse azimuth estimate for:

[0174]

[0175] When local peaks aggregate When empty, the first One pitch angle estimate Corresponding coarse azimuth estimate for:

[0176]

[0177] After extracting coarse azimuth estimates sequentially for each pitch angle estimate, a set of coarse azimuth estimates is obtained. .

[0178] Step 5: Perform a local fine-grid search on the coarse azimuth estimates corresponding to each elevation angle estimate to obtain the fine azimuth estimates:

[0179] For the One pitch angle estimate The initial estimate of the coarse azimuth angle Establish a local azimuth fine-search set centered on the target:

[0180]

[0181] In the formula, For local search half-width, For finer search intervals, In this embodiment, the index of the azimuth angle to be scanned in the local fine search set is used. , , , This is a modulo operation with 360° as the modulus, used to map angle values ​​to a periodic range of 0° to 360°.

[0182] when hour, The corresponding set of allowed fine search .

[0183] when At that time, the previously obtained fine azimuth estimates were used respectively. A circular protection zone is set up in the local azimuth fine search set centered on the azimuth angle, and azimuth angles falling into the protection zone are removed to obtain the following result. The corresponding set of allowed fine search :

[0184]

[0185] In the formula, This is an index for the extracted fine azimuth estimates.

[0186] In sets that allow fine-grained search In the middle, maintain the corresponding pitch angle estimate. Unchanged, with finer search intervals A fine-grid search is performed to extract the azimuth angle corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum as the elevation angle estimate. fine azimuth estimate :

[0187]

[0188] After extracting fine azimuth estimates sequentially for each elevation angle estimate, a set of fine azimuth estimates is obtained. .

[0189] Each pitch angle estimate is paired with its corresponding fine azimuth angle estimate to output a set of two-dimensional direction-of-arrival (DOA) estimates for multiple sound sources. :

[0190]

[0191] To better illustrate the beneficial effects of this invention, a typical deep-sea waveguide environment was used, and the performance of the deep-sea volume array target azimuth estimation method based on array geometry decoupling (the method of this invention) and the direct two-dimensional orthogonal projection method provided in this embodiment were compared and verified under the same simulation conditions. The direct two-dimensional orthogonal projection method directly performs a joint spectrum search in the azimuth-elevation two-dimensional space, without using a vertical linear array to pre-extract elevation angle estimates, nor using the elevation angle estimates as constraints for searching the azimuth angle. The simulation environment and parameter settings are as follows:

[0192] The Munk sound velocity profile was used to describe the deep-sea waveguide environment at a depth of 5000m, a signal center frequency of 190Hz, and a reference sound velocity of 1500m / s. In the volumetric array, the vertical linear array contains 30 elements with a spacing of 2.5m. The lowest element of the vertical linear array is located at a depth of 4990m. A horizontal circular array is deployed at the same depth (4990m), with its center coinciding with the position of the lowest element of the vertical linear array. The horizontal circular array contains 30 elements with a radius of 3m. The propagation response from the sound source to each array element was calculated using a normal mode model. Two sound sources were set up: source 1 had a depth of 150m and a horizontal distance of 7000m from the center of the horizontal circular array, with a true azimuth angle of 32° and a true elevation angle of 55.34°; source 2 had a depth of 320m and a horizontal distance of 12500m from the center of the horizontal circular array, with a true azimuth angle of 150° and a true elevation angle of 69.51°. The sampling frequency was 1000Hz, the observation time was 2s, the number of snapshots was 2000, and additive noise with a signal-to-noise ratio of 5dB was added to the array received signal.

[0193] Figure 2 This demonstrates the distribution of propagation loss with radial distance at different depths in a simulated environment, with sound source 1 as the reference point. Figure 2 It can be seen that the acoustic propagation loss in the deep-sea waveguide environment exhibits obvious spatial fluctuations with distance, reflecting the multimodal interference effect.

[0194] Figure 3 The diagram shows the two-dimensional direction-of-arrival estimation results obtained using the method of this invention under single-source (source 1) conditions. Figure 3 It can be seen that the orthogonal projection azimuth spatial spectrum forms a distinct main peak near the actual angle of the target, and the estimated position is close to the actual position, indicating that the present invention can effectively extract the direction information of the sound source under the mixed sound field conditions in the deep sea.

[0195] Figure 4 The pitch angle spatial spectrum of a vertical linear array obtained using the method of this invention under single sound source conditions is shown. Figure 5 Shown in Figure 3 The orthogonal projection azimuth spatial spectrum is obtained by intercepting the image at an elevation angle of 55° along the azimuth direction. Figure 4 and Figure 5 It can be seen that both the vertical linear array elevation angle spatial spectrum and the orthogonal projection azimuth angle spatial spectrum can form clear spectral peaks.

[0196] Figure 6 The results of two-dimensional direction-of-arrival estimation obtained using the direct two-dimensional orthogonal projection method under dual-source conditions (source 1 and source 2) are presented. Figure 6 It can be seen that the direct two-dimensional orthogonal projection method only forms a clear spectral peak near the true location of one sound source. The spectral peak corresponding to the other sound source is obscured by the side lobes or leakage peaks of the strong sound source, and fails to form an identifiable independent peak in the two-dimensional spatial spectrum, making it difficult to effectively detect and locate the direction of the weak sound source.

[0197] Figure 7 The spatial spectrum of the vertical linear array elevation angle obtained using the method of this invention under dual-source conditions is shown. From Figure 7 It can be seen that the pitch angle spatial spectrum of the vertical linear array forms obvious peaks near the true pitch angle values ​​of both sound sources. The angles corresponding to these peaks are the estimated pitch angle values, indicating that the vertical linear array can effectively identify the vertical component of the sound source direction by utilizing the geometric features arranged along the depth direction. First, the pitch angle estimates are extracted from the vertical linear array. Then, the two pitch angle estimates are sorted by their angle magnitude, providing clear constraints for the subsequent azimuth angle estimation of the horizontal circular array.

[0198] Figure 8 The orthogonal projection azimuth spatial spectrum corresponding to each elevation angle estimate obtained by the method of this invention under dual sound source conditions is shown. Figure 8 It can be seen that, under the constraint of the pitch angle estimation, the spatial spectrum of the orthogonal projection azimuth angle forms a clear spectral peak near the true azimuth angle of the sound source, and the estimated azimuth has good consistency with the true azimuth.

[0199] Figure 9 The results of two-dimensional direction-of-arrival estimation obtained using the method of this invention under dual-source conditions are presented. Figure 9 It can be seen that the spectral peaks corresponding to the two sound sources fall near their respective true directions, and the peak shapes are clear with no obvious crosstalk between the two peaks, indicating that the method of this invention can achieve effective pairing of the pitch and azimuth angles of the two sound sources. (Comparison) Figure 6 and Figure 9 As can be seen, the method of the present invention effectively suppresses the masking of weak sound source spectral peaks by strong sound source sidelobes or leakage peaks through pitch angle estimation and circumferential protection. The direction of weak sound sources can be reliably identified, and the accuracy and stability of two-dimensional direction of arrival estimation are significantly improved.

[0200] The specific embodiments of the present invention enable those skilled in the art to understand or implement the invention. Various modifications to the above embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention.

[0201] It should be understood that the present invention is not limited to the content already described above, and various modifications and changes can be made without departing from its scope. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for estimating the azimuth of a deep-sea volumetric array target based on array geometric decoupling, characterized in that, Includes the following steps: Step 1: Construct a volume array consisting of a vertical linear array and a horizontal circular array, acquire the received data of the volume array, extract the vertical linear array data matrix and the horizontal circular array data matrix from the received data of the volume array, calculate the sample covariance matrix of the vertical linear array based on the vertical linear array data matrix, and calculate the sample covariance matrix of the horizontal circular array based on the horizontal circular array data matrix. Step 2: Construct the vertical linear array pitch angle spatial spectrum based on the vertical linear array sample covariance matrix, and extract the pitch angle estimates of multiple sound sources from the vertical linear array pitch angle spatial spectrum in combination with the pitch angle protection zone. Step 3: Based on the covariance matrix of the horizontal circular array samples, the signal subspace vector of each sound source is extracted using the stepwise orthogonal projection method, and the horizontal circular array signal subspace projection matrix is ​​constructed; according to the horizontal circular array signal subspace projection matrix, the orthogonal projection azimuth space spectrum is constructed under the constraints of each pitch angle estimate. Step 4: In the spatial spectrum of azimuth angles of each orthogonal projection, suppress the repeated peaks under different pitch angle estimates by using the circular protection zone, and extract the coarse azimuth angle estimates corresponding to each pitch angle estimate. Step 5: Perform a local fine-grid search on the coarse azimuth estimate to obtain a fine azimuth estimate. Pair each elevation angle estimate with the corresponding fine azimuth estimate to output the two-dimensional direction of arrival estimation results for multiple sound sources.

2. The method for estimating the azimuth of a deep-sea volumetric array target based on array geometric decoupling according to claim 1, characterized in that, Step 2 specifically includes: The covariance matrix of the vertical linear array sample is subjected to conjugate symmetry processing to obtain the conjugate symmetric vertical linear array sample covariance matrix. The covariance matrix of the conjugate symmetric vertical linear array sample is diagonally loaded to obtain the diagonally loaded vertical linear array sample covariance matrix. For the elevation angle to be scanned, a vertical linear array guide vector is constructed. Each element of the vertical linear array guide vector is determined by the reference wavenumber, the relative depth of each element of the vertical linear array, and the cosine value of the elevation angle to be scanned. The vertical linear array steering vector is normalized to obtain the normalized vertical linear array steering vector. Based on the normalized vertical linear array steering vector and the sample covariance matrix of the diagonally loaded vertical linear array, the pitch angle spatial spectrum of the vertical linear array is constructed using the minimum variance distortionless response algorithm. The maximum spectral peak is extracted sequentially from the vertical linear array pitch angle spatial spectrum. After each extraction, a pitch angle protection zone is set with the currently extracted spectral peak as the center, and the spectral values ​​within the pitch angle protection zone are set to negative infinity to update the current vertical linear array pitch angle spatial spectrum. The next extraction of the maximum spectral peak is performed in the updated pitch angle spatial spectrum until the pitch angle estimates of multiple sound sources are obtained.

3. The deep-sea volumetric array target orientation estimation method based on array geometry decoupling according to claim 1, characterized in that, In step 3, based on the covariance matrix of the horizontal circular array samples, the signal subspace vector of each sound source is extracted using a stepwise orthogonal projection method, and the horizontal circular array signal subspace projection matrix is ​​constructed, specifically including: The horizontal circular array sample covariance matrix is ​​subjected to conjugate symmetry processing to obtain the conjugate symmetric horizontal circular array sample covariance matrix. Initialize the orthogonal complement projection matrix; For the Secondary signal subspace vector extraction, where , Given the number of sound sources, calculate the projection covariance matrix in the current orthogonal complement space. The projection covariance matrix is ​​calculated based on the current orthogonal complement projection matrix and the sample covariance matrix of the conjugate symmetric horizontal circular array. The projection covariance matrix is ​​subjected to conjugate symmetry processing to obtain the conjugate symmetric projection covariance matrix; Find the unit eigenvector corresponding to the largest eigenvalue of the conjugate symmetric projective covariance matrix; Project the unit feature vector onto the current orthogonal complement space to obtain the projection vector; Normalize the projection vector to obtain the first... The signal subspace vector of each sound source; Update the orthogonal complement projection matrix by subtracting the product of the currently extracted signal subspace vector and its conjugate transpose from the current orthogonal complement projection matrix; The updated orthogonal complement projection matrix is ​​used for the next extraction of signal subspace vectors; Repeat execution The next signal subspace vector extraction process yields... A horizontal circular array of signal subspaces composed of signal subspace vectors; The horizontal circular array signal subspace is orthogonalized using QR to obtain a signal subspace basis matrix with mutually orthogonal column vectors. The horizontal circular array signal subspace projection matrix is ​​then constructed based on the signal subspace basis matrix.

4. The deep-sea volume array target orientation estimation method based on array geometry decoupling according to claim 3, characterized in that, ,in This represents the number of elements in the vertical linear array. This represents the number of elements in a horizontal circular array.

5. The deep-sea volume array target orientation estimation method based on array geometry decoupling according to claim 1, characterized in that, Step 3, the process of constructing the spatial spectrum of orthogonal projection azimuth angles under the constraints of each pitch angle estimate, includes: For the One pitch angle estimate, , To determine the number of sound sources, the direction of the sound sources is calculated based on the current estimated elevation angle and the azimuth angle to be scanned. Direction cosines along the axis Direction cosine along the axial direction; Construct the horizontal circular array steering vector based on the reference wavenumber, the circumference angle of the horizontal circular array element, the radius of the horizontal circular array, the current elevation angle estimate, and the azimuth angle to be scanned. The horizontal circular array guide vector is normalized to obtain the normalized horizontal circular array guide vector. Calculate the projection energy of the normalized horizontal circular array steering vector in the horizontal circular array signal subspace; The projected energy is subjected to amplitude limiting processing; Construct the orthogonal projection azimuth spatial spectrum corresponding to the current pitch angle estimate based on the projection energy after amplitude limiting.

6. The method for estimating the azimuth of a deep-sea volumetric array target based on array geometric decoupling according to claim 1, characterized in that, Step 4 specifically includes: Set a coarse search interval and perform a coarse grid search within the azimuth angle range of 0° to 360°; For the first pitch angle estimate, the azimuth angle corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum is extracted as the coarse azimuth angle estimate; Define any two azimuth angles and circumferential distance between for: Set a circular protection zone half the width; For the One pitch angle estimate, , To determine the number of sound sources, a circular protection zone is established centered on the previously extracted coarse azimuth estimates. Only outside the scope of the circular protection zone, the azimuth corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum is extracted as the coarse azimuth estimate of the current elevation angle estimate.

7. The deep-sea volumetric array target orientation estimation method based on array geometry decoupling according to claim 6, characterized in that, Step 5 specifically includes: For each pitch angle estimate, a corresponding local azimuth fine search set is established centered on the coarse azimuth angle estimate. The local azimuth fine search set is determined by the local search half-width and the fine search interval, and the fine search interval is smaller than the coarse search interval. For the first pitch angle estimate, the corresponding local azimuth angle fine search set is used as the allowed fine search set; For the One pitch angle estimate, Circular protection zones are set up centered on each previously obtained fine azimuth angle estimate. The circular protection zones are removed from the corresponding local azimuth angle fine search set to obtain the allowable fine search set. In the set of allowed fine search, the corresponding pitch angle estimate is kept unchanged, and a fine grid search is performed at a fine search interval to extract the azimuth angle corresponding to the maximum spectral peak of the orthogonal projection azimuth spatial spectrum as the fine azimuth angle estimate. After extracting fine azimuth estimates for each elevation angle estimate, the elevation angle estimates are paired with the corresponding fine azimuth estimates to output the two-dimensional direction of arrival (DOA) estimates for multiple sound sources.

8. The deep-sea volumetric array target orientation estimation method based on array geometry decoupling according to claim 5, characterized in that, The upper limit for the projection energy when it is subjected to amplitude limiting is .

Images