Deconvolution focusing positioning method based on spatial matrix pre-filtering

Abstract

The present invention discloses a deconvolution focusing localization method based on spatial domain matrix pre-filtering, which includes using a hydrophone array to receive noise source signals from the near field; obtaining the cross-spectrum matrix corresponding to each sub-band; obtaining the conventional focusing beamforming for the corresponding sub-band; obtaining the broadband conventional focusing beamforming; obtaining the position coordinates of the strong noise source; designing the spatial domain matrix filters corresponding to each sub-band near the strong noise source; obtaining the covariance matrix filtered by each sub-band; obtaining the conventional focusing output pre-filtered by the corresponding spatial domain matrix filter at each sub-band; obtaining the array directivity function dictionary corresponding to each sub-band; performing deconvolution iteration on the array directivity function dictionary corresponding to each sub-band and the conventional focusing output corresponding to each sub-band to obtain the noise source distribution corresponding to each sub-band; summing the noise source distributions corresponding to each sub-band to obtain the final weak noise source distribution.

Description

Technical Field

[0001] The present invention relates to the field of array signal processing, and particularly to a deconvolution focusing localization method based on spatial domain matrix pre-filtering. Background Art

[0002] Conventional focusing beamforming is a common sound source localization method, which only requires arranging sensors in the measurement area. Compared with the acoustic holography algorithm, acoustic focusing has simpler equipment, and at the same time has high positioning accuracy, high gain, and high robustness, and is widely used in the measurement of the position of noise sources. The article "Research on the Principle of Focusing Beamforming Sound Map Measurement" (published in Acta Acustica, 2007, 32(4): 356-361) proposed to use the conventional focusing beamforming algorithm to localize noise sources, but when using this algorithm to localize low-frequency noise sources, there are often phenomena such as wide beams and high sidelobes.

[0003] When there are multiple noise sources in the scanning area at the same time, for a single noise source among them, other noise sources can be regarded as background interference. When the strength difference between noise sources is obvious, the weak noise source is very likely to be submerged in the sidelobes. In order to improve the localization performance of weak noise sources, suppressing other noise sources is one way, and an idea for suppressing noise sources is to form nulls using a spatial domain matrix filter. The article "Research on the Robustness and Algorithm of Sonar Beamforming" (Harbin Engineering University, 2011: 65-85) proposed a near-field sound source localization method based on a matrix spatial domain filter, which can realize the localization of weak noise sources, but this method still has problems such as low positioning accuracy for the localization of low-frequency weak noise sources. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0005] The purpose of the present invention is to provide a deconvolution focusing localization method based on spatial domain matrix pre-filtering. When strong and weak noise sources exist simultaneously, using this method can avoid problems such as the low-frequency weak noise source being submerged by the strong noise source and the large scale of the focusing peak during the localization process, further reducing the influence of the strong noise source on the localization of the low-frequency weak noise source, and thus realizing high-precision localization of the low-frequency weak noise source.

[0006] According to an embodiment of the present application, a deconvolution focusing localization method based on spatial domain matrix pre-filtering is provided, including the following steps:

[0007] Step 1: Use a hydrophone array to receive the noise source signal from the near field. The noise source signal received by the entire hydrophone array is x(t), and x(t) is expressed by the formula:

[0008] x(t) = [x1(t), x2(t),..., x m(t),…,x M (t)]

[0009] where m and M are both integers greater than or equal to 1; M represents the number of array elements in the pressure hydrophone array; x m (t) represents the noise source signal received by the m-th array element;

[0010] Step 2: Transform the noise source signal x(t) received by the entire pressure hydrophone array into the frequency domain for processing; and split the receiving frequency band of the noise source signal x(t) received by the entire pressure hydrophone array into multiple sub-bands f j , and obtain the cross-spectral matrix R(f j ) corresponding to each sub-band f j ), where j is an integer greater than or equal to 1;

[0011] Step 3: Convert the two-dimensional scanning area into a one-dimensional scanning area; and rearrange the scanning points in the two-dimensional scanning area in a column splicing manner so that each two-dimensional scanning point corresponds to a one-dimensional scanning point;

[0012] Step 4: Based on the cross-spectral matrix R(f j ) of each sub-band f j ), perform phase compensation on each sub-band f j respectively according to the order of the one-dimensional scanning point u to obtain the conventional focusing beamforming P(u, f j ) at the corresponding sub-band f j ); and sum the conventional focusing beamforming P(u, f j ) at each sub-band f j ) to obtain the broadband conventional focusing beamforming P(u);

[0013] Step 5: Obtain the position coordinates of the strong noise source based on the broadband conventional focusing beamforming P(u); and design the corresponding spatial matrix filter G(f j ) for each sub-band f j near the strong noise source based on the MS criterion and the second-order cone programming theory;

[0014] Step 6: Based on the spatial matrix filter G(f j ) corresponding to each sub-band f j ) and the cross-spectral matrix R(f j ) corresponding to each sub-band f j ), obtain the filtered covariance matrix R j (f G ) for each sub-band f j ;

[0015] Step 7: Based on the covariance matrix RG (f j ) to obtain the conventional focusing output B pre-filtered by the corresponding spatial domain matrix filter G(f j ) at each sub-band f j ); G (u, f j );

[0016] Step 8: Obtain the corresponding array directivity function dictionary p(u|v, f j ) at each sub-band f j , where v represents a one-dimensional noise source point;

[0017] Step 9: Based on the extended RL algorithm, perform deconvolution iteration on the corresponding array directivity function dictionary p(u|v, f j ) at each sub-band f j and the corresponding conventional focusing output B j at each sub-band f G (u, f j ) to obtain the noise source distribution corresponding to each sub-band f j ;

[0018] Step 10: Sum the noise source distributions corresponding to each sub-band f j to obtain the final weak noise source distribution.

[0019] In the above method, in Step 1, the noise source signal x m (t) received by the m-th array element is expressed by the formula:

[0020]

[0021] where m and k are both integers greater than or equal to 1; A m,k represents the signal strength of the k-th noise source received by the m-th array element; s k (t) represents the radiation signal of the k-th noise source; r m,k represents the distance from the k-th noise source to the m-th array element, and c represents the speed of sound.

[0022] In the above method, in Step 2, the cross-spectral matrix R(f j ) corresponding to each sub-band is expressed by the formula:

[0023] R(f j ) = E{X(f j )X H (f j )}

[0024] where X(f j ) represents the hydrophone array of the sound pressure in the sub-band fj is the Fourier transform of the noise source signal x(t) received at ; X H (f j ) represents the conjugate transpose of the Fourier transform of the noise source signal x(t) received by the hydrophone array in the sub-band f j ; E{·} represents taking the average value.

[0025] In the above method, in step 4, in the sub-band f j , the conventional focusing beamforming P(u, f j ) is expressed by the formula:

[0026] P(u, f j ) = a H (u, f j ) R(f j ) a(u, f j )

[0027] where represents the weight vector at the corresponding sub-band f j ; r m ' represents the distance from the one-dimensional scanning point u to the m-th array element; a H (u, f j ) represents the conjugate transpose of the weight vector for phase compensation at the one-dimensional scanning point u in the sub-band f j .

[0028] In the above method, in step 4, the broadband conventional focusing beamforming P(u) is expressed by the formula:

[0029]

[0030] In the above method, in step 5, in the broadband conventional focusing beamforming p(u), the peak point p(u m ) is searched for, and the peak point p(u m ) corresponds to the one-dimensional scanning point u m , and the position coordinates of the strong noise source are the two-dimensional coordinate points corresponding to the one-dimensional scanning point u m .

[0031] In the above method, in step 5, the set of scanning points near the position of the strong noise source is the stopband region, and the set of the remaining scanning points is the passband region.

[0032] In the above method, in step 5, the vectorized form g of the spatial domain matrix filter G(f j ) corresponding to each sub-band f j is expressed by the formula:

[0033]

[0034] ||g|| ≤ η

[0035] Wherein, represents the transpose of the steering vectors at all scan points in the passband region;

[0036] U P represents the set of scan points in the passband region, and N P represents the number of scan points in the passband region;

[0037] represents the vectorization of the set of steering vectors after spatial domain filtering at all scan points in the passband region; represents the sub - frequency band f j the steering vector after filtering of the scan points in the passband region at this point;

[0038] a T (u s , f j ) represents the conjugate transpose of the steering vector after filtering of the scan points in the stopband region at the sub - frequency band f j at this point, U S represents the set of scan points in the stopband region, and N S represents the number of scan points in the passband region;

[0039] represents a d - dimensional unit array; δ represents the maximum value of the stopband attenuation; η represents the limit value of the Frobenius norm of the matrix filter;

[0040] represents the objective linear function;

[0041] u s ∈ U S represents the second - order cone constraint;

[0042] And, use the cvx toolbox to obtain the vectorized form g, and rearrange the vectorized form g to obtain the spatial domain matrix filter G(f j ) at each sub - frequency band f j .

[0043] In the above method, in step 6, for each sub - frequency band f j the covariance matrix R G (f j ), is expressed by the formula:

[0044] R G (f j ) = GH (f j )R(f j )G(f j )。

[0045] In the above method, in step 7, the pre-filtered conventional focusing output B G (u, f j ), which is expressed by the formula as:

[0046]

[0047] In the formula, is the steering vector after filtering of the scanning point located in the passband region at the sub-band f j .

[0048] According to the technical solution provided by the present application, it has at least the following beneficial effects:

[0049] a) By filtering the noise source signal generated by the strong noise source through the spatial domain matrix filter, the signal-to-interference ratio of the weak noise source is improved, which is more conducive to the positioning of the weak noise source.

[0050] b) The position estimation of the strong noise source has fault tolerance. When the position estimation of the strong noise source is incorrect, as long as its position is still within the stopband range of the designed spatial domain matrix filter, the noise source signal generated by the strong noise source can still be suppressed.

[0051] c) Compared with the conventional focusing beamforming based on spatial domain matrix pre-filtering, a smaller focusing peak scale and lower sidelobes can be obtained by using the extended RL algorithm after filtering.

[0052] Other features and advantages of the present application will be described in the subsequent specification, and part of them will become obvious from the specification or be understood by implementing the present application. The objectives and other advantages of the present application can be achieved and obtained through the structures specifically pointed out in the specification, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, where:

[0054] Figure 1 is the flowchart of the positioning method provided by the embodiment of the present application;

[0055] Figure 2 is the horizontal array sound source positioning model provided by the embodiment of the present application;

[0056] Figure 3 is the schematic diagram of converting two-dimensional scanning points into one-dimensional scanning points provided by the embodiment of the present application;

[0057] Figure 4 The effect diagram of conventional focusing beamforming provided by the embodiment of the present application;

[0058] Figure 5 The performance effect diagram of the spatial domain matrix filtering pre-processor provided by the embodiment of the present application;

[0059] Figure 6 The effect diagram of conventional beamforming based on spatial domain matrix pre-filtering provided by the embodiment of the present application;

[0060] Figure 7 The effect diagram of conventional beamforming based on spatial domain matrix pre-filtering after estimating the wrong position provided by the embodiment of the present application;

[0061] Figure 8 The effect diagram of the extended RL algorithm based on spatial domain matrix pre-filtering provided by the embodiment of the present application. Detailed implementation manners

[0062] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

[0063] It should be noted that although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from the module division in the device or the flowchart in the flowchart. The terms "first", "second", etc. in the description and claims of the present application and the above drawings are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0064] To make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0065] As Figure 1 shown, the embodiment of the present application provides a deconvolution focusing positioning method based on spatial domain matrix pre-filtering, and the method includes the following steps:

[0066] Step 1: Use a hydrophone array to receive the noise source signal from the near field. The noise source signal received by the entire hydrophone array is x(t), and x(t) is expressed by the formula:

[0067] x(t) = [x1(t), x2(t), …, x m (t), …, x M (t)]

[0068] In this step, the hydrophone array has multiple array elements. If the hydrophone array is a horizontal linear array, a coordinate system as shown in Figure 2 is established with the line where the multiple array elements are located as the x-axis. Both m and M are integers greater than or equal to 1, and M represents the number of array elements in the hydrophone array. x m (t) represents the noise source signal received by the m-th array element, and the coordinate of the m-th array element is (x m , 0, 0); the noise source moves on the plane of z = z0 and generates corresponding noise source signals. There are a total of K noise sources, and the coordinate of the k-th noise source is (x k , y k , z0). It should be noted that in this application, the noise sources include strong noise sources and weak noise sources, and each generates its corresponding noise source signal.

[0069] Specifically, the multiple array elements are evenly distributed on the x-axis to receive the corresponding noise source signals x m (t). x m (t) is expressed by the formula:

[0070]

[0071] In the formula, both m and k are integers greater than or equal to 1; A m,k represents the signal strength of the k-th noise source received by the m-th array element; s k (t) represents the radiation signal of the k-th noise source; r m,k represents the distance from the k-th noise source to the m-th array element, and c represents the speed of sound.

[0072] Step 2: Transform the noise source signal x(t) received by the entire hydrophone array into the frequency domain for processing; and, split the receiving frequency band of the noise source signal x(t) received by the entire hydrophone array into multiple sub-bands f j , and obtain the cross-spectral matrix R(f j ) corresponding to each sub-band f j . j is an integer greater than or equal to 1.

[0073] In this step, the noise source signal x(t) received by the entire hydrophone array is Fourier-transformed to be processed in the frequency domain.

[0074] In this step, if the lower limit of the receiving frequency band of the noise source signal x(t) received by the entire hydrophone array is f l , and the upper limit is f h , then the frequency band from f l to f h is evenly split into j sub-bands on average, and each sub-band f jThe cross-spectral matrix \(R(f j ) is expressed by the formula as follows:

[0075] R(f j ) = E\{X(f j )X H (f j )\}

[0076] Wherein, \(X(f j )\) represents the Fourier transform of the noise source signal \(x(t)\) received by the hydrophone array at the sub-band \(f j \); \(X H (f j )\) represents the conjugate transpose of the Fourier transform of the noise source signal \(x(t)\) received by the hydrophone array at the sub-band \(f j \); \(E\{\cdot\}\) represents taking the average.

[0077] Step 3: Convert the two-dimensional scanning area into a one-dimensional scanning area; and rearrange the scanning points in the two-dimensional scanning area in a column splicing manner so that each two-dimensional scanning point can correspond to a one-dimensional scanning point \(u\).

[0078] In this step, if the number of scanning points in the \(x\) direction in the two-dimensional scanning area is \(N X \), and the number of scanning points in the \(y\) direction is \(N Y \), then the total number of scanning points is \(\Omega = N X \times N Y .

[0079] In this step, the scanning points in the two-dimensional scanning area are rearranged in a column splicing manner, as shown in Figure 3 , and each two-dimensional scanning point \((x, y)\) corresponds to a one-dimensional scanning point \(u\).

[0080] Step 4: Based on the cross-spectral matrix \(R(f j )\) of each sub-band, perform phase compensation on each sub-band \(f j \) in the order of the one-dimensional scanning point \(u\) to obtain the conventional focusing beamforming \(P(u, f j )\) at the corresponding sub-band \(f j \); and sum the conventional focusing beamforming \(P(u, f j )\) at each sub-band \(f j \) to obtain the broadband conventional focusing beamforming \(P(u)\).

[0081] In this step, the conventional focusing beamforming \(P(u, f j )\) is expressed by the formula as follows:

[0082] P(u, f j ) = a H(u, f j )R(f j )a(u, f j )

[0083] wherein, represents the weight vector at the corresponding sub - frequency band f j ; r m ' represents the distance from the one - dimensional scanning point u to the m - th array element; a H (u, f j ) represents the conjugate transpose of the weight vector for phase compensation at the one - dimensional scanning point u at the sub - frequency band f j .

[0084] It should be noted that when this formula is actually running, u represents the output of the conventional focusing beam at the one - dimensional scanning point u.

[0085] In this step, the wide - band conventional focusing beamforming P(u) is expressed by the formula as:

[0086]

[0087] Step 5: Obtain the position coordinates of the strong noise source based on the wide - band conventional focusing beamforming P(u); and, design the spatial domain matrix filter G(f j ) corresponding to each sub - frequency band f j based on the MS criterion and the second - order cone programming theory respectively.

[0088] In this step, find the peak point p(u m ) in the wide - band conventional focusing beamforming p(u). This peak point p(u m ) corresponds to the one - dimensional scanning point u m . The two - dimensional coordinate point corresponding to the one - dimensional scanning point u m is the position coordinate of the strong noise source; the set of scanning points near the position of the strong noise source is the stop - band region, and the set of the remaining scanning points is the pass - band region.

[0089] Specifically, design the vectorized form g of the spatial domain matrix filter G(f j ) based on the MS criterion and the second - order cone programming theory. In this step, it is necessary to design for each sub - frequency band f j respectively, which is expressed by the formula as:

[0090]

[0091] ||g||≤η

[0092] wherein, represents the transpose of the steering vector at all scanning points located in the pass - band region;

[0093] U P represents the set of scanning points within the passband region, N P represents the number of scanning points within the passband region;

[0094] represents the vectorization of the set of steering vectors after spatial domain filtering at all scanning points within the passband region; represents the sub - frequency band f j the steering vector of the scanning points within the passband region at f after filtering;

[0095] a T (u s , f j ) represents the conjugate transpose of the steering vector of the scanning points within the stopband region at f after filtering, j where the scanning points are within the stopband region; U S represents the set of scanning points within the stopband region, N S represents the number of scanning points within the passband region;

[0096] represents a - dimensional unit array; δ represents the maximum value of the stopband attenuation; η represents the limit value for the Frobenius norm of the matrix filter, which can limit the attenuation of noise;

[0097] represents the objective linear function;

[0098] u s ∈U S represents the second - order cone constraint.

[0099] Furthermore, using the cvx toolbox to obtain the vectorized form g (g is a one - dimensional vector), and re - arranging the vectorized form g can obtain the spatial domain matrix filter G(f j ) at each sub - frequency band f. j )

[0100] In this application, design the spatial domain matrix filter G(f j ) corresponding to each sub - frequency band f near the strong noise source j to achieve the effect of filtering the noise source signal generated by the strong noise source, so that only the noise source signal of the weak noise source is retained in the noise source signal received by the hydrophone array, which is beneficial to improving the signal - to - interference ratio of the weak noise source and preparing for the precise positioning of the weak noise source in the subsequent process.

[0101] Step 6, based on the spatial domain matrix filter G(f j ) corresponding to each sub - frequency band f j) and the cross-spectrum matrix R(f corresponding to each sub-band j ), obtain each sub-band f j of the filtered covariance matrix R G (f j ).

[0102] In this step, each sub-band f j of the filtered covariance matrix R G (f j ) is expressed by the formula:

[0103] R G (f j ) = G H (f j )R(f j )G(f j ).

[0104] Step 7: Based on the covariance matrix R G (f j ), obtain the conventional focusing output B j pre-filtered by the corresponding spatial domain matrix filter G(f j ) at each sub-band f G (u, f j ).

[0105] In this step, the pre-filtered conventional focusing output B G (u, f j ) is expressed by the formula:

[0106]

[0107] In the formula, represents the filtered steering vector of the scanning points within the passband region at sub-band f j .

[0108] Step 8: Obtain the array directivity function dictionary p(u|v, f j ) corresponding to each sub-band f j .

[0109] In this step, the array directivity function dictionary p(u|v, f j ) is expressed by the formula:

[0110]

[0111] In the formula, r m ' represents the distance from the one-dimensional scanning point u to the m-th array element; r m represents the distance from the noise source at the one-dimensional noise source point v to the m-th array element.

[0112] It should be noted that the array directivity function dictionary p(u|v,f j ) represents the contribution of the noise source at the one-dimensional noise source point v to the scanning point u at the sub-band f j . The scanning point u and the one-dimensional noise source point v are located in the same scanning area. When this formula is actually run, u represents the output of the conventional focusing beam at the one-dimensional scanning point u, and v represents the sound power at the one-dimensional noise source point. The set of all p(u|v,f j ) is called the array directivity function dictionary (PSF) at the sub-band f j .

[0113] Step 9: Perform deconvolution iteration on the array directivity function dictionary p(u|v,f j ) corresponding to each sub-band f j and the conventional focusing output B j (u,f G ) corresponding to each sub-band f j to obtain the noise source distribution corresponding to each sub-band f j .

[0114] In this step, when performing iteration based on the extended RL algorithm, it is necessary to select an appropriate number of iterations. The difference between the results of the previous and current iterations can be used as the iteration termination condition, and the initial iteration value is the conventional focusing beamforming after pre-filtering. The iteration formula of the extended RL algorithm is as follows:

[0115]

[0116] In the formula, u represents the one-dimensional scanning point; v represents the one-dimensional noise source point; B (r) (u,f j ) represents the beam result after the r-th iteration at each sub-band f j ; q (r) (v,f j ) represents the noise source distribution after the r-th iteration at each sub-band f j .

[0117] It should be noted that the number of iterations can be set according to actual needs and is not specifically limited in this application.

[0118] In this application, applying the extended RL algorithm to near-field sound source localization can improve the localization accuracy of the sound source.

[0119] Step 10: Sum the noise source distributions corresponding to each sub-band f j to obtain the final weak noise source distribution.

[0120] In this step, the final distribution of the weak noise sources is expressed by the formula:

[0121]

[0122] Next, the extended RL algorithm based on spatial domain matrix pre-filtering of this application will be compared and verified from the perspective of computer simulation.

[0123] The specific parameters of the simulation environment are as follows:

[0124] The sea depth is 20 m;

[0125] The number of array elements of the horizontal linear array is 20;

[0126] The element spacing is 5 m;

[0127] The deployment depth is 18 m;

[0128] Assume the sound source movement depth is 5 m;

[0129] The sound speed c is taken as 1450 m / s;

[0130] The band-limited noise frequency band is 300 Hz - 500 Hz;

[0131] The sound source signal is a broadband signal, the sampling frequency is 100 kHz, the frequency band is 300 Hz - 500 Hz, the bandwidth of each sub-band is 10 Hz, the signal-to-noise ratio of the strong noise source is 10 dB, and the signal-to-noise ratio of the weak noise source is 0 dB;

[0132] The sound sources are located at (-8 m, 25 m, 13 m) and (12 m, 20 m, 13 m) respectively.

[0133] For the extended RL algorithm based on spatial domain matrix filtering, the number of iterations is taken as 100 to locate the weak noise sources. When the position of the strong noise source is estimated incorrectly, assume the estimated position of the strong noise source is (-9 m, 26 m, 13 m).

[0134] Combined with Figure 4 It can be seen that when there are two sound sources with a large difference in sound source intensity in the scanning area, the strong noise source will cover up the weak noise source. For the weak noise source, the existence of the strong noise source is equivalent to increasing the background noise, reducing the signal-to-noise ratio relatively, and it is impossible to locate the weak noise source. Only the strong noise source appears in the positioning result.

[0135] Design a spatial domain matrix filter in the strong noise source area x ∈ [-12 m, -4 m], y ∈ [23 m, 28 m]. From Figure 5 It can be seen that there is stopband attenuation in this area, and the steering vector of the scanning points in this area will be approximately set to zero, and the signals in this area are filtered out.

[0136] Combined with Figure 6As can be seen, if a stopband matrix filter is designed in the area where the strong noise source is located and then conventional focused beamforming is performed, the strong noise source is suppressed at this time, and the position of the weak noise source can be obtained. The normalized background level of the positioning result is about 0.2.

[0137] It can be seen from Figure 7 that when the position of the strong noise source is estimated incorrectly, the spatial domain matrix filter also has a certain degree of fault tolerance. As long as the position of the strong noise source is within the stopband, it can be suppressed, improving the positioning ability of the algorithm for weak noise sources.

[0138] Referring to Figure 8 it can be seen that when the spatial domain matrix filtering is combined with the extended RL algorithm, the focused peak scale and sidelobe of the weak noise source are smaller than those of the conventional focused beamforming, and the positioning accuracy is higher, which is more conducive to the positioning of weak noise sources.

[0139] This application combines spatial domain matrix filtering with the extended RL algorithm. First, the position of the strong noise source is estimated, and a spatial domain matrix filter near the strong noise source is designed; because during the positioning of the weak noise source, the noise source signal generated by the strong noise source will interfere with the positioning of the weak noise source, and based on the spatial domain matrix filter near the strong noise source, the noise source signal generated by the strong noise source can be filtered out, only retaining the noise source signal of the weak noise source; then, based on the extended RL algorithm, accurate positioning of the low-frequency weak noise source is achieved.

[0140] The above is a specific description of the preferred implementation of this application, but this application is not limited to the above implementation manner. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of this application, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. An anti-convolution focusing and positioning method based on spatial domain matrix pre-filtering, characterized in that Including the following steps: Step 1: Use a hydrophone array to receive the noise source signal from the near field. The noise source signal received by the entire hydrophone array is x(t), and x(t) is expressed by the formula: x(t) = [x1(t), x2(t),..., x m (t),..., x M (t)] where m and M are both integers greater than or equal to 1; M represents the number of array elements in the hydrophone array; x m (t) represents the noise source signal received by the m-th array element; Step 2: Transform the noise source signal x(t) received by the entire hydrophone array into the frequency domain for processing; and, split the receiving frequency band of the noise source signal x(t) received by the entire hydrophone array into multiple sub-bands f j , and obtain the cross-spectral matrix R(f j ) corresponding to each sub-band f j ), where j is an integer greater than or equal to 1; Step 3: Convert the two-dimensional scanning area into a one-dimensional scanning area; and rearrange the scanning points in the two-dimensional scanning area in a column-by-column splicing manner so that each two-dimensional scanning point corresponds to a one-dimensional scanning point; Step 4. Based on the cross-spectral matrix R(f j ) of each sub-band f j , perform phase compensation on each sub-band f j separately in the order of the one-dimensional scanning points u to obtain the conventional focusing beamforming P(u, f j ) at the corresponding sub-band f j ; and sum up the conventional focusing beamforming P(u, f j ) at each sub-band f j to obtain the wideband conventional focusing beamforming P(u); Step 5: Obtain the position coordinates of the strong noise source based on the broadband conventional focusing beamforming \(P(u)\); and, design the corresponding spatial domain matrix filter \(G(f\) j ) for each sub-band \(f\) near the strong noise source based on the MS criterion and the second-order cone programming theory j ). Step 6. Based on each sub-band f j corresponding spatial domain matrix filter G(f j ), and cross-spectrum matrix R(f j ), obtain the covariance matrix R j filtered for each sub-band f G (f j ); Step 7. Based on the covariance matrix R G (f j ) Obtain the conventional focusing outputs B j pre-filtered by the corresponding spatial domain matrix filter G(f j ) at each sub-band f G (u, f j ); Step 8, obtain the array directivity function dictionaries p(u|v, f j corresponding to each sub-band f j ), where v represents a one-dimensional noise source point; Step 9: Based on the extended RL algorithm, for each sub-band f j at the corresponding array directivity function dictionary p(u|v, f j ) and the conventional focusing output B G (u, f j ), perform deconvolution iteration to obtain the noise source distribution corresponding to each sub-band f j ; Step 10: Sum up the noise source distributions corresponding to each sub-band f j to obtain the final weak noise source distribution.

2. The positioning method according to claim 1, wherein In step 1, the noise source signal x m (t) received by the m-th array element is expressed by the formula as follows: where m and k are both integers greater than or equal to 1; A m,k represents the signal strength of the k-th noise source received by the m-th array element; s k (t) represents the radiation signal of the k-th noise source; r m,k represents the distance from the k-th noise source to the m-th array element, x k represents the position of the k-th noise source on the x-axis; x m represents the position of the m-th array element on the x-axis; y k represents the position of the k-th noise source on the y-axis; z0 represents the position of the k-th noise source on the z-axis; c represents the speed of sound.

3. The positioning method according to claim 1, characterized in that In step 2, the cross-spectrum matrix R(f j ) corresponding to each sub-band is expressed by the formula as follows: R(f j ) = E{X(f j )X H (f j )} where X(f j ) represents the Fourier transform of the noise source signal x(t) received by the hydrophone array at sub-band f j ; X H (f j ) represents the conjugate transpose of the Fourier transform of the noise source signal x(t) received by the hydrophone array at sub-band f j ; E{·} represents taking the average value.

4. The positioning method according to claim 1, characterized in that, In step 4, at sub-band f j conventional focusing beamforming P(u, f j ) is performed, which is expressed by the formula: P(u,f j ) = a H (u,f j )R(f j )a(u,f j ) In the formula, represents the weight vector at the corresponding sub - frequency band f j ; r' m represents the distance from the one - dimensional scanning point u to the m - th array element; a H (u, f j ) represents the conjugate transpose of the weight vector for phase compensation at the one - dimensional scanning point u at the sub - frequency band f j ; c represents the speed of sound.

5. The positioning method according to claim 4, characterized in that In Step 4, the broadband conventional focusing beamforming P(u) is expressed by the formula:

6. The positioning method according to claim 1, wherein In step 5, find the peak point p(u) in the broadband conventional focusing beamforming p(u m ), where the peak point p(u m ) corresponds to the one-dimensional scanning point u m , and the position coordinates of the strong noise source are the two-dimensional coordinate points corresponding to the one-dimensional scanning point u m .

7. The positioning method according to claim 1, wherein In Step 5, the set of scanning points near the position of the strong noise source is the stopband area, and the set of the remaining scanning points is the passband area.

8. The positioning method according to claim 7, characterized in that, In step 5, for each sub-band f j the corresponding spatial domain matrix filter G(f j ), the vectorized form g is expressed by the formula: In the formula, represents the transpose of the steering vectors at all the scanning points in the passband region; a(u1,f j ) represents the steering vector at the first scanning point in the passband region at sub-band f j ; a(u2,f j ) represents the steering vector at the second scanning point in the passband region at sub-band f j ; represents the steering vector at the j th scanning point in the passband region at sub-band f ; U P represents the set of scanning points in the passband region, and N P represents the number of scanning points in the passband region; Denote the vectorization of the set of steering vectors after spatial domain filtering at all scan points within the passband region; Denote the sub-band f j The steering vector after filtering of the scan points within the passband region at [location]. a T (u s ,f j ) represents the conjugate transpose of the filtered steering vector of the scanning point located in the stopband region at sub-band f j ; U S represents the set of scanning points in the stopband region, and N S represents the number of scanning points in the passband region; representation dimensional unit array; δ represents the maximum value of the stopband attenuation; η represents the limiting value of the Frobenius norm of the matrix filter; represents the target linear function; represents a second-order cone constraint; Moreover, the vectorized form g is obtained by using the cvx toolbox, and the vectorized form g is rearranged to obtain the spatial domain matrix filter G(f j at each sub-band f j ).

9. The positioning method according to claim 1, wherein In step 6, for each sub-band f j the filtered covariance matrix R G (f j ), is expressed by the formula: R G (f j ) = G H (f j )R(f j )G(f j )。 10. The positioning method according to claim 1, wherein In step 7, the pre-filtered conventional focusing output B G (u, f j ), which is expressed by the formula as: wherein, is the steering vector after filtering of the scanning point located in the passband region at the sub-band f j .

Images