Underwater multi-target orientation estimation method based on improved adaptive matching pursuit

Abstract

The invention discloses an underwater multi-target orientation estimation method based on improved adaptive matching pursuit, and belongs to the technical field of array signal processing and underwater target orientation estimation. The method comprises the following steps: acquiring a target signal through an array, and carrying out complex processing on acquired target signal data; constructing an airspace dictionary set used for improving adaptive matching pursuit based on a target signal incoming wave direction; carrying out adaptive search on the constructed airspace dictionary set, and completing atom screening according to the change condition of residual energy; and according to a complete incoming wave direction angle set used when the airspace dictionary set is constructed, in combination with an index set determined by completing atom screening, completing estimation of the number of incoming waves and a sparse signal, and converting the estimated airspace sparse signal into a target orientation. According to the method, the problem that in existing underwater multi-target azimuth estimation, the number of target incoming waves and the direction of the target incoming waves cannot be estimated in a self-adaptive and high-precision mode frequently is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of array signal processing and underwater target azimuth estimation, and in particular relates to an underwater multi-target azimuth estimation method based on improved adaptive matching pursuit. Background Art

[0002] Underwater multi-target position estimation is one of the core technologies in the fields of ocean exploration, underwater robot navigation, etc. Accurately obtaining the number and position information of underwater targets can not only improve the performance of underwater communication, navigation and detection systems, but also has important significance for practical applications. However, the existing target position estimation algorithms have the following limitations: traditional beamforming algorithms: it is difficult to achieve both high resolution and high robustness; subspace algorithms: they rely on the known number of targets and cannot effectively process coherent signals; neural network-based algorithms: they have poor adaptability in different scenarios.

[0003] In recent years, with the development of compressed sensing theory, sparse recovery methods have gradually become an important means of high-precision target position estimation. As a representative of sparse recovery algorithms, greedy algorithms have attracted widespread attention due to their advantages of low computational complexity and high reconstruction accuracy. However, traditional sparse recovery algorithms require pre-estimation of sparsity, which limits their application scope. Although the adaptive matching pursuit algorithm (SAMP) eliminates the dependence on sparsity estimation, the following challenges still exist in the target position estimation problem:

[0004] 1. Difficulty in selecting the iteration step size: Improper step size setting may introduce additional atoms and cause the algorithm to fail;

[0005] 2. The cardinality of the support set affects the estimation effect: If the cardinality of the support set is too large, it will reduce the accuracy of the least squares estimation and thus affect the sparse recovery effect;

[0006] 3. The termination condition is difficult to predict: the algorithm’s a priori dependence on noise energy limits its practicality.

[0007] To solve the above problems, the present invention proposes an underwater multi-target position estimation method based on an improved adaptive matching pursuit algorithm, which optimizes the incoming wave signal to achieve high-precision, multi-target underwater position estimation. This method aims to overcome the shortcomings of traditional algorithms and provide more reliable technical support for underwater target positioning. Summary of the invention

[0008] Technical issues to be solved:

[0009] In order to avoid the shortcomings of the prior art, the present invention provides an underwater multi-target azimuth estimation method based on improved adaptive matching pursuit, which selects atoms by the least squares similarity criterion, and makes an iterative termination judgment by referring to the energy change degree of adjacent residuals, so as to realize adaptive and accurate estimation of the number and azimuth of incoming waves. The present invention solves the problem that the number and direction of target incoming waves cannot be adaptively and accurately estimated in the existing underwater multi-target azimuth estimation.

[0010] The technical solution of the present invention is: an underwater multi-target azimuth estimation method based on improved adaptive matching pursuit, the specific steps are as follows:

[0011] The target signal is acquired through array acquisition, and the acquired target signal data is processed into complex numbers;

[0012] Based on the incoming direction of the target signal, a spatial domain dictionary set used for improving adaptive matching pursuit is constructed;

[0013] Adaptively search the constructed spatial domain dictionary set and complete atom screening according to the change of residual energy;

[0014] According to the complete set of incoming wave direction angles used when constructing the spatial dictionary set, combined with the index set determined by completing the atomic screening, the number of incoming waves and sparse signals are estimated, and the estimated spatial sparse signals are converted into target directions.

[0015] A further technical solution of the present invention is: the method for complex processing the acquired target signal data is to represent the real signal received by each array element as x(t), and perform Hilbert transform on the real signal x(t), and the expression is as follows:

[0016]

[0017] In the formula, is a complex signal, which is the Hilbert transform result of x(t); τ represents the integral variable;

[0018] The result of Hilbert transforming the real-number signal received by each array element in the same snapshot is represented by a vector y.

[0019] A further technical solution of the present invention is: the method for constructing the spatial domain dictionary set is:

[0020] K real-number signals from K incoming wave directions are received through M array elements;

[0021] Based on the incoming wave direction of the real signal, a complete incoming wave direction angle set is established;

[0022] Based on the complete set of incoming wave direction angles, the first array element of the array is taken as a reference point, and the spatial domain dictionary set A is designed according to the array structure characteristics as follows:

[0023]

[0024] In the formula, j is the imaginary unit, f is the carrier frequency of the incoming signal, τ m (θ n ) is the time delay of the signal arriving at the mth array element relative to the reference point when the signal is incident at an angle θn, m=2,…,M,n=1,…,N.

[0025] A further technical solution of the present invention is that: before the atomic screening, the coherent parameters need to be initialized, that is, the residual signal r 0 =y, number of iterations t = 1, number of stages stage = 1, index set Step size s=1, number of atoms screened L=1.

[0026] A further technical solution of the present invention is: the specific process of the atom screening is:

[0027] Calculate the spatial domain dictionary set A and the residual r t-1 The inner product U of

[0028] Select the largest value from U, and record the A sequence number j corresponding to this value as the potential index set J;

[0029] Update the index set and candidate support set;

[0030] Solve the least squares solution of the updated candidate support set expression array received signal y;

[0031] Select the first L elements with the largest values ​​from the vector element values ​​of the least squares solution, and record the corresponding indexes as F;

[0032] Update residuals;

[0033] The change in residual energy is judged. When the number of iterations and the change in residual energy meet the requirements, the atom screening is completed and the iteration is stopped.

[0034] A further technical solution of the present invention is: the calculation process of the atomic screening is as follows:

[0035] The spatial domain dictionary set A and the residual r t-1 The inner product U of is expressed as: u = abs(A T r t-1 );

[0036] The index set is updated to Λ t =Λ t-1 ∪J;

[0037] The candidate support set is updated to A t =A t-1 ∪{a j}, j∈J;

[0038] The least squares solution expression is:

[0039] The residual is updated as

[0040] A further technical solution of the present invention is: the specific process of judging according to the change of residual energy is as follows:

[0041] If || r new || 2 ≥||r t-1 || 2 , then update stage = stage + 1, L = stage * s, and return to the next iteration;

[0042] If || r new || 2 <||r t-1 || 2 , then update the index set Λ t =F, and record the residual value change I t =||r t-1 || 2 -||r new || 2 ; At the same time, make a multi-target position estimation iteration termination judgment: if the number of iterations t ≥ 3, I t <0.7*I t-2 with I t-1 <0.7*I t-2 If both conditions are met, the atomic screening is completed and the iteration is stopped; otherwise, take r t =r new , t=t+1, continue iterating.

[0043] A further technical solution of the present invention is: the number of incoming waves K is estimated as: column vector The dimension of .

[0044] A further technical solution of the present invention is: the wave direction of the target signal is estimated by: the pre-constructed complete wave direction angle set θ = [θ 1 ,θ 2 ,…,θ N ]According to the index set Λ t-2 The angle obtained by indexing the element values ​​in .

[0045] An underwater multi-target position estimation system based on improved adaptive matching pursuit, comprising an array sensor, a signal complex processing module, an atom screening module and an estimation module;

[0046] Acquire a real signal of a target signal through an array sensor;

[0047] Convert the real signal into a complex signal by a signal complex processing module;

[0048] The atom screening module is used to adaptively search the spatial domain dictionary set and complete the atom screening according to the change of residual energy;

[0049] The estimation module is used to estimate the number of incoming waves and sparse signals, and convert the estimated spatial sparse signals into the target direction.

[0050] Beneficial Effects

[0051] The present invention has the following beneficial effects: The algorithm proposed in the present invention converts the target azimuth estimation problem into a sparse signal reconstruction problem by establishing a spatial sparse direction of arrival estimation model, thus breaking through the limitation of array aperture and achieving higher resolution azimuth estimation. The specific advantages are as follows:

[0052] Improved iterative step size selection: By analyzing the specific characteristics of the BOA estimation problem, a safer iterative step size selection scheme is proposed, which effectively solves the problem of difficult step size selection in the classic adaptive matching pursuit algorithm and improves the robustness and applicability of the algorithm.

[0053] Optimize the cardinality of the potential index set: By reasonably reducing the cardinality of the potential index set, the problem of too many outliers caused by a large support set, which in turn affects the effect of the least squares estimation, is solved, thereby improving the accuracy and robustness of the BOA estimation.

[0054] Improved iterative termination condition: Based on observing the ratio of residual changes between two adjacent iterations, an iterative termination condition based on residual energy changes is proposed, which overcomes the dependence of the classical algorithm on noise energy estimation and enhances the applicability of the algorithm in complex noise environments.

[0055] In addition, the improved algorithm proposed in the present invention breaks through the limitation of the number of snapshots, and can adaptively achieve high-precision estimation of the number and orientation of targets even in the case of a limited number of snapshots or even a single snapshot. This advantage makes the algorithm more widely applicable and more reliable in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flow chart of the present invention;

[0057] Figure 2This is a simulation test result diagram of an example of the present invention when the signal-to-noise ratio is 20dB;

[0058] Figure 3 This is a simulation test result diagram of an example of the present invention when the signal-to-noise ratio is 10dB;

[0059] Figure 4 This is a simulation test result diagram of an example of the present invention when the signal-to-noise ratio is 0dB. DETAILED DESCRIPTION

[0060] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0061] The existing greedy algorithm is a mature algorithm in compressed sensing theory. Its basic idea is to use the sparse representation of the signal in the redundant dictionary, and gradually select the atom that best matches the current signal residual from the dictionary in an iterative manner, and update the residual until the accuracy requirement is met. Since the target wave direction is sparse in the spatial domain, here we will construct a spatial sparse dictionary and use the matching pursuit algorithm to adaptively estimate the number and direction of the waves.

[0062] Based on the existing underwater multi-target position estimation, there is often a problem that the number of target waves and the direction of target waves cannot be estimated adaptively and accurately. The present invention provides an underwater multi-target position estimation method based on improved adaptive matching pursuit. Figure 1 As shown, the specific steps are as follows:

[0063] Step 1: complexify the target signal data collected by the array;

[0064] In this embodiment, it is assumed that the water depth is 100m in the experimental conditions, the depths of the transmitting transducer are 20m, 30m and 40m respectively, the 10-element uniform linear array carried by the underwater vehicle is 30m below the sea level, the horizontal distance between the transmitting transducer and the receiving array is 1km, the incoming wave directions are: -70°, -15° and 34° respectively, and the simulation signal is a carrier frequency of f 0 =3kHz single frequency narrowband signal, the receiving array element spacing d is half wavelength, and the speed of sound is c = 1500m / s. The real signal received by each array element is represented as x(t), and the real signal x(t) is converted into a complex signal by the following formula:

[0065]

[0066] The observation vector obtained by converting the real signal received by each array element at the same snapshot into a complex signal is denoted as y;

[0067] Step 2: Construct a spatial domain dictionary set used to improve the adaptive matching pursuit algorithm;

[0068] When K narrowband signals with the same frequency come from the direction When the signals are incident on the M-element array at the same time, considering that the possible directions of the signals are limited, an exhaustive method can be used to obtain a complete set of angles of the directions of the signals Θ = [θ 1 ,θ 2 ,…,θ N ], so that the signal incident direction constitutes a smaller subset of that set.

[0069] In this embodiment, the entire airspace from -90° to 90° is divided into N grids at equal intervals, and the airspace can be sampled at an angle interval of 1° to construct Θ C = a complete set of incoming wave direction angles of [-90°, -89°, ..., 90°]. And according to the center frequency of the incoming wave signal and the array structure, a spatial domain dictionary set A as shown in the following formula is established;

[0070]

[0071] Step 3: Initialize algorithm-related parameters including residual signal, number of iterations, step size, number of stages, and index set;

[0072] Initialize the algorithm related parameters: residual signal r 0 =y, number of iterations t = 1, number of stages stage = 1, index set Step size s = 1, number of atoms screened L = 1;

[0073] Step 4: Adaptively search the dictionary set using the invented algorithm and complete atom screening based on the change of residual energy;

[0074] Step 4.1: Calculate the column vectors and residual r in the spatial domain dictionary set A t-1 The correlation coefficient is as follows:

[0075] u=abs(A T r t-1 ) (3)

[0076] Step 4.2: Take the largest value in u(i), and record the A sequence number j corresponding to this value as the potential index set J;

[0077] Step 4.3: Update the index set and candidate support set respectively. The formula is as follows:

[0078] Λ t =Λ t-1 ∪J (4)

[0079] A t =At-1 ∪{a j},j∈J (5)

[0080] Among them, a j represents the j-th column of matrix A.

[0081] Step 4.4: Solve for the matrix A t The least squares solution of the expression array receiving signal y is as follows:

[0082]

[0083] Step 4.5: From column vector The values ​​of each element Select the element with the largest value in the first L items, and record the corresponding index as F; and update the residual by the following formula

[0084]

[0085] Step 4.6: Observe the change of residual energy and make the following judgment:

[0086] If at this time ‖r new ‖ 2 ≥‖r t-1 ‖ 2 , then update stage = stage + 1, L = stage * s, and return to the next iteration;

[0087] If at this time || r new || 2 <||r t-1 || 2 , then update the index set Λ t =F, and record the residual value change I t =||r t-1 || 2 -||r new || 2 At the same time, the multi-target position estimation iteration termination judgment is made. If the number of iterations t ≥ 3, I t <0.7*I t-2 with I t-1 <0.7*I t-2 If both are satisfied, the atomic screening is completed, the iteration is stopped, and the fifth step is entered. Otherwise, take r t =r new , t=t+1, continue iterating;

[0088] Step 5: The number of incoming waves K is estimated as: column vector The wave direction of each target is estimated as follows: the pre-constructed complete wave direction angle set Θ = [θ 1 ,θ2 ,…,θ N ]According to the index set Λ t-2 The angle obtained by indexing the values ​​of each element in , for example: the estimated direction of the incoming wave of the i-th target (i=1,2,…,K) is where Λ t-2 (i) represents the index set Λ t-2 The element value of the i-th element in .

[0089] Reference Figure 2-Figure 4 As shown in the figure, the analysis results of the experimental data are respectively shown when the signal-to-noise ratio is 20dB, 10dB and 0dB, and the performance of the underwater multi-target azimuth estimation method based on improved adaptive matching pursuit proposed by the present invention is compared with the classic MUSIC subspace azimuth estimation method. It can be seen from the experimental result figure that when the MUSIC algorithm estimates the target azimuth of the coherent signal source, multiple pseudo peaks appear, and the spectrum peak is not sharp enough, and it is almost impossible to distinguish the direction of arrival of the source. The method of the present invention can accurately estimate the direction of arrival of the three sources under different signal-to-noise ratios, and successfully realizes the estimation of the direction of arrival of the coherent signal source. Its advantage is that the direction of arrival estimation method based on compressed sensing does not require the characteristic decomposition of the data covariance matrix, so it has a natural decoherence ability, which overcomes the deficiency that the subspace algorithm cannot estimate the coherent source. In addition, in this experiment, the algorithm proposed by the present invention realizes the accurate estimation of the direction of arrival through only a single snapshot data, which proves that the algorithm has a low demand for the amount of sampling information and can reduce the data storage and calculation costs.

[0090] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and purpose of the present invention.

Claims

1. A method for underwater multi-target position estimation based on improved adaptive matching pursuit, characterized in that The specific steps are as follows: The target signal is acquired through array acquisition, and the acquired target signal data is processed into complex numbers; Based on the incoming direction of the target signal, a spatial domain dictionary set used for improving adaptive matching pursuit is constructed; Adaptively search the constructed spatial domain dictionary set and complete atom screening according to the change of residual energy; According to the complete set of incoming wave direction angles used when constructing the spatial dictionary set, combined with the index set determined by completing the atomic screening, the number of incoming waves and sparse signals are estimated, and the estimated spatial sparse signals are converted into target directions.

2. According to claim 1, a method for underwater multi-target position estimation based on improved adaptive matching pursuit is characterized by: The method for complex processing the acquired target signal data is to represent the real signal received by each array element as x(t), and perform Hilbert transform on the real signal x(t), and the expression is as follows: In the formula, is a complex signal, which is the Hilbert transform result of x(t); τ represents the integral variable; The result of Hilbert transforming the real-number signal received by each array element in the same snapshot is represented by a vector y.

3. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 2 is characterized by: The method for constructing the airspace dictionary set is: K real-number signals from K incoming wave directions are received through M array elements; Based on the incoming wave direction of the real signal, a complete incoming wave direction angle set is established; Based on the complete set of incoming wave direction angles, the first array element of the array is taken as a reference point, and the spatial domain dictionary set A is designed according to the array structure characteristics as follows: In the formula, j is the imaginary unit, f is the carrier frequency of the incoming signal, τ m (θ n ) is the signal from the angle θ n When incident, the time delay reaching the mth array element relative to the reference point is, m=2,…,M, n=1,…,N.

4. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 3 is characterized by: Before the atomic screening, the coherent parameters need to be initialized, that is, the residual signal r0=y, the number of iterations t=1, the number of stages stage=1, and the index set Step size s=1, number of atoms screened L=1.

5. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 4, characterized in that: The specific process of the atomic screening is: Calculate the spatial domain dictionary set A and the residual r t-1 The inner product U of Select the largest value from U, and record the A sequence number j corresponding to this value as the potential index set J; Update the index set and candidate support set; Solve the least squares solution of the updated candidate support set expression array received signal y; Select the first L elements with the largest values ​​from the vector element values ​​of the least squares solution, and record the corresponding indexes as F; Update residuals; The change in residual energy is judged. When the number of iterations and the change in residual energy meet the requirements, the atom screening is completed and the iteration is stopped.

6. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 5, characterized in that: The calculation process of the atomic screening is as follows: The spatial domain dictionary set A and the residual r t-1 The inner product U of is expressed as: u = abs(A T r t-1 ); The index set is updated to Λ t =Λ t-1 ∪J; The candidate support set is updated to A t =A t-1 ∪{a j }, j∈J; The least squares solution expression is: The residual is updated as 7. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 6, characterized in that: The specific process of judging according to the change of residual energy is as follows: If || r new ||2≥||r t-1 ||2, then update stage = stage + 1, L = stage * s, and return to the next iteration; If || r new ||2<||r t-1 ||2, then update the index set Λ t =F, and record the residual value change I t =||r t-1 ‖2-||r new ||2; At the same time, make a multi-target position estimation iteration termination judgment: if the number of iterations t≥3, I t <0.7*I t-2 with I t-1 <0.7*I t-2 If both conditions are met, the atomic screening is completed and the iteration is stopped; otherwise, take r t =r new , t=t+1, continue iterating.

8. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 7, characterized in that: The number of incoming waves K is estimated as: column vector The dimension of .

9. The underwater multi-target position estimation method based on improved adaptive matching pursuit according to claim 8, characterized in that: The target signal’s direction of arrival is estimated by: N ]According to the index set Λ t-2 The angle obtained by indexing the element values ​​in .

10. An underwater multi-target position estimation system based on improved adaptive matching pursuit, used to implement the underwater multi-target position estimation method based on improved adaptive matching pursuit according to any one of claims 1 to 9; characterized in that: It includes an array sensor, a signal complex processing module, an atom screening module and an estimation module; Acquire a real signal of a target signal through an array sensor; Convert the real signal into a complex signal by a signal complex processing module; The atom screening module is used to adaptively search the spatial domain dictionary set and complete the atom screening according to the change of residual energy; The estimation module is used to estimate the number of incoming waves and sparse signals, and convert the estimated spatial sparse signals into the target direction.

Images