Comprehensive multi-snapshot combined Newton orthogonal matching pursuit positioning method

Abstract

The invention relates to a comprehensive multi-snapshot combined Newton orthogonal matching pursuit positioning method, and belongs to the field of acoustics. According to the method, all possible in-network sound source coordinates are estimated through an improved comprehensive orthogonal matching pursuit method, a grid is locally refined with the coordinates as the center to reconstruct a target sound source area, and a new sensing matrix is constructed; randomly selecting a plurality of groups of sub-arrays with the same number of array elements from an original array, preliminarily estimating coordinates of a sound source in a reconstruction region based on multi-sub-array data, and performing compensation optimization on the estimated coordinates and source intensity of the sound source through a multi-snapshot combined Newton method; the method has the advantages that by improving the atomic selection process, it is guaranteed that a reconstruction area contains all sound sources to the maximum extent, the coherence problem caused by finer grid spacing is avoided, and the on-network positioning capacity is enhanced; according to the multi-subarray joint estimation method, measurement data can be fully utilized, and the sound source positioning precision is improved.

Description

Technical Field

[0001] The invention belongs to the field of acoustics and relates to a comprehensive multi-quick-beat combined Newton orthogonal matching tracking positioning method. Background Art

[0002] In order to transform the sound source estimation problem into a sparse reconstruction problem for compressed sensing, the traditional compressed beamforming method constructs a sparse reconstruction model by gridding the target sound source area. It assumes that the target sound source falls on the divided grid points. However, the fact is that no matter how finely the grid of the target area space is divided, the sound source may not be on the grid point, and there is always a mismatch between the assumed basis and the real basis. For greedy algorithms such as orthogonal matching pursuit methods, the existence of basis mismatch may even lead to positioning errors. Although a finer grid can alleviate the basis mismatch problem, it will increase the demand for computing resources and the increase in the correlation between dictionary atoms, which will increase the difficulty of estimation. At the same time, the presence of noise also increases the difficulty of correctly selecting atoms.

[0003] In addition, it is common to improve positioning performance by increasing the observation dimension of the sound source signal, but this will undoubtedly increase the difficulty of measurement. However, in existing studies, most of the measurement data have only been used once, and the effective information in the data has obviously not been fully utilized. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a comprehensive multi-snapshot combined with Newton orthogonal matching pursuit positioning method to more fully utilize the effective information in the existing measurement data and improve the positioning performance of the sound source.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method, the method comprising the following steps:

[0007] S1: Based on the measurement data of the original M array, the improved comprehensive orthogonal matching pursuit method is used to obtain the candidate set of sound source coordinates, and the target sound source area is reconstructed by locally refining the grid around it to construct a new perception matrix A. sub ;

[0008] S2: According to the principle of overall sparseness of the array, multiple groups of M sub The sub-arrays are composed of array elements, the measurement data corresponding to each sub-array is constructed, and the maximum likelihood function is constructed;

[0009] S3: Based on the reconstructed target grid area, a rough on-grid estimation result of the sound source is obtained;

[0010] S4: Correct the estimated results by Newton's method to improve positioning accuracy. This is divided into two steps: first, a single-step correction is performed on the online estimation results obtained in the current iteration, and then, based on the current latest residual, a multi-step correction is performed on all the estimated results in sequence.

[0011] S5: Set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area.

[0012] Further, the S1 comprises the following steps:

[0013] S11, divide the plane where the sound source is located into Ng grid points according to a certain spatial spacing d, and calculate the a(r ng ) is normalized to get a n (r ng ), a(r ng ) is the spatial coordinate r of the ngth grid ng To the array element space coordinate r composed of M array elements mic.k ,k=1,2,...,the guiding vector (atom) between M;

[0014] a n (r ng )=a(r ng ) / ||a(r ng )||2,ng=1,2,...,Ng

[0015] a(r ng )=[g(r ng ,r mic.k ),g(r ng ,r mic.2 ),...,g(r ng ,r mic.M )] T

[0016]

[0017] Among them, g(r ng ,r mic.m ) is the Green function, ||·||2 represents the 2-norm operation, [·] T represents transpose, f is the frequency of the sound source, and c is the speed of sound;

[0018] S12, in each iteration, by calculating the current residual Y res=[y(1), y(2), ..., y(L)], where L is the number of snapshots and the inner product of each column of the normalized atom in the dictionary. The atoms corresponding to the first three largest inner product values ​​are selected, and a circular area that just contains its neighboring atoms is covered with the coordinates of the atom with the largest inner product value as the center. The candidate set of the sound source coordinates is determined by judging whether the other two atoms fall within the circle. The coordinates of the atoms outside the area and the coordinates of the atoms with the largest inner product value together constitute the candidate set of coordinates R su ;

[0019]

[0020] R su =R su {r k.1 ,r k.i ⊙(r k.1 )},r k = {r k.1 ,r k.2 ,r k.3}

[0021] Where ⊙(r k.1 ) represents the coordinate r k.1 The center covers a certain range of circular space. Like the orthogonal matching pursuit method, in each iteration, the coordinates of the atom with the maximum inner product value are used as the estimation result to promote the iteration process.

[0022] S13. Based on the atomic coordinates in the candidate set, half of the original space grid spacing d is used as the new grid spacing, and a square grid space with a radius of 1.5d is radiated. These local grid spaces together constitute our new target sound source area R Ns , and construct a new perception matrix A sub , which is composed of the steering vectors corresponding to all grid points in the target area.

[0023] Further, the S2 comprises the following steps:

[0024] S21, randomly select L from the original M array of elements according to the overall sparseness principle of the array A Group by M sub The corresponding array data together constitute a new measurement signal in Represents the first A The measurement data corresponding to each sub-array;

[0025] S22. For each subarray, based on By minimizing the residual energy Obtain the maximum likelihood solution of the sound source coordinates and intensity, and construct the maximum likelihood function

[0026]

[0027] in is the perception matrix corresponding to the real sound source coordinates, Corresponding source strength, is noise, ||·|| F represents the Frobenius norm, represents a real number, (·) H stands for conjugate transpose.

[0028] Further, the S3 comprises the following steps:

[0029] S31, by estimating a sound source estimate in each iteration process, a maximum likelihood ratio test cost function GLRT(·) is established according to the least squares method;

[0030]

[0031] S32. For each subarray, in each iteration, the coordinates corresponding to the first three maximum values ​​of GLRT(·) are taken as the current possible estimated values. A circular area with a radius that just contains its neighboring grid points is radiated from the estimated value with the most occurrences among all possible estimated values ​​as the center, and the number of estimated values ​​falling in the area is counted. The central grid point of the area corresponding to the largest number is taken as the final on-grid estimated coordinate.

[0032]

[0033] count(g) represents the number of counts that meet the conditions;

[0034] S33, calculating the estimated sound source strength on the network

[0035]

[0036] Further, the S4 comprises the following steps:

[0037] S41, performing a single-step correction on the current online estimation result by using Newton's method;

[0038]

[0039] In the formula, is the Hessian matrix, is the Jacobian matrix;

[0040] S42, update residual Y res , atomic support set A sup , coordinate support set R supAnd the source strength support set Q sup ;

[0041]

[0042] S43, based on the latest current residual, all estimated results are corrected in multiple steps in sequence; first, the source strength is updated by the least squares method to complete the orthogonalization operation, and then the result is corrected by the Newton method;

[0043]

[0044] Further, the S5 comprises the following steps:

[0045] S51, set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area; update it through the least squares method Y estimated and Y res , calculate Tol1 and Tol2, determine whether the iteration stop condition is met, and the iteration terminates when (Tol1≤ε1) or (Tol2≤ε2); where Y res0 Refers to the measurement residual before each iteration;

[0046]

[0047] Tol1=Σ(|Y res0 | 2 -|Y res | 2 ) / ∑(|Y res0 | 2 )

[0048] Tol2=∑(|Y res0 | 2 -|Y res | 2 ) / ∑(|Y Estimated | 2 )

[0049] After the iteration stops, the coordinate support set R is output sup and source strong support set Q sup , that is, the estimated results of the sound source coordinates and their corresponding sound source intensity are obtained.

[0050] The beneficial effects of the present invention are:

[0051] Since the present invention ensures that all sound sources fall within the reconstruction area as much as possible according to the correlation when reconstructing the target area with a greatly reduced range, the correlation problem caused by finer grids in the future is avoided, and the on-grid positioning capability is improved; at the same time, the present invention combines the estimation results of multiple subarrays, so that the estimated on-grid coordinates in each iteration process are as close to the real sound source coordinates as possible, and the joint use of multiple groups of subarray data in the Newton method off-grid correction process makes the Newton method correction direction more stable and reliable, thereby more reliably realizing the accurate positioning of the sound source.

[0052] The method for reconstructing the sound source target area proposed in this method improves the sound source estimation capability and can avoid erroneous estimation results to a certain extent; the multi-subarray joint estimation model introduced in the invention of this method can make more full use of the useful information in each measurement data, improve noise robustness, improve positioning accuracy, and reduce positioning deviation.

[0053] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0055] Figure 1 It is a schematic diagram of the process of the present invention;

[0056] Figure 2 It is a principle diagram. Figure 2 (a) is a schematic diagram of the principle of the improved orthogonal matching pursuit method. Figure 2 (b) is a schematic diagram of the reconstructed sound source area;

[0057] Figure 3 This is the simulation result diagram of three sound sources at 20kHz; Figure 3 (a) is the positioning result of the MNOMP method under the condition of separated sound sources; Figure 3 (b) is the positioning result of the proposed invention method COMP-MJNOMP under the separated sound sources; Figure 3 (c) is the positioning result of the MNOMP method under adjacent sound sources; Figure 3 (d) is the positioning result of the proposed invention method COMP-MJNOMP under adjacent sound sources;

[0058] Figure 4 This is the result of Monte Carlo randomized experiment; Figure 4(a) is the sound source correct recognition rate CRR, which is used to measure the probability of a single correct positioning; Figure 4 (b) is the root mean square error RMSE, which is used to measure the deviation of positioning coordinates; Figure 4 (c) is the cumulative correct recognition rate CCRR, which is used to measure the probability of correct estimation of a single sound source; Figure 4 (d) is the mean amplitude deviation MAE, which is used to measure the deviation of the estimated source strength;

[0059] Figure 5 Position the result map for the test; Figure 5 (a) is the positioning result of the multi-snapshot orthogonal matching pursuit method; Figure 5 (b) is the positioning result of the multi-shot Newton orthogonal matching pursuit method; Figure 5 (c) is the positioning result obtained by the method of the present invention. DETAILED DESCRIPTION

[0060] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0061] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0062] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0063] See also Figure 1 , Figure 2It is a principle diagram. Figure 2 (a) is a schematic diagram of the principle of the improved orthogonal matching pursuit method. Figure 2 (b) is a schematic diagram of the reconstructed sound source area. The present invention is a comprehensive multi-quick shot combined with Newton orthogonal matching tracking positioning method, comprising the following steps:

[0064] S1: Based on the measurement data of the original M array, the improved comprehensive orthogonal matching pursuit method is used to obtain the candidate set of sound source coordinates, and the target sound source area is reconstructed by locally refining the grid around it to construct a new perception matrix A. sub :

[0065] S11. Normalize the atoms in the original sound source plane to obtain a n (r ng );

[0066] a n (r ng )=a(r ng ) / ||a(r ng )||2,ng=1,2,...,Ng

[0067] S12, in each iteration, by calculating the current residual Y res The inner product of each column of normalized atoms in the dictionary is used to select the atoms corresponding to the first three largest inner product values. A circular area that can just contain its neighboring atoms is covered with the coordinates of the atom with the largest inner product value as the center of the circle. The candidate set of sound source coordinates is determined by judging whether the other two atoms fall within the circle. The coordinates of the atoms outside the area and the coordinates of the atoms with the largest inner product value together constitute the candidate set of coordinates R. su .

[0068]

[0069] R su =R su {r k.1 ,r k.i ⊙(r k.1 )},r k = {r k.1 ,r k.2 ,r k.3}

[0070] Where ⊙(r k.1 ) represents r k.1 The circular space covers a certain range as the center. Like the orthogonal matching pursuit method, in each iteration, the coordinates corresponding to the atom with the maximum inner product value are used as the estimation result to promote the iteration process.

[0071] S13. Based on the atomic coordinates in the candidate set, half of the original space grid spacing d is used as the new grid spacing, and a square grid space with a radius of 1.5d is radiated. These local grid spaces together constitute our new target sound source area R Ns .

[0072] S2: According to the principle of overall sparseness of the array, multiple groups of M sub The subarrays are composed of array elements, the measurement data corresponding to each subarray is constructed, and the maximum likelihood function is constructed:

[0073] S21, randomly select L from the original M array of elements according to the overall sparseness principle of the array A Group by M sub The corresponding array data together constitute a new measurement signal in

[0074] S22. For each subarray, based on By minimizing the residual energy Obtain the maximum likelihood solution of the sound source coordinates and intensity, and construct the maximum likelihood function in is the perception matrix corresponding to the real sound source coordinates, Corresponding source strength, is noise, ||·|| F Represents the Frobenius norm.

[0075]

[0076] S3: Based on the reconstructed target grid area, obtain a rough on-grid estimation result of the sound source:

[0077] S31, by estimating a sound source estimate in each iteration process, a maximum likelihood ratio test cost function GLRT(·) is established according to the least squares method;

[0078]

[0079] S32. For each subarray, in each iteration, the coordinates corresponding to the first three maximum values ​​of GLRT(·) are taken as the current possible estimated values. A circular area with a radius that just contains its neighboring grid points is radiated from the estimated value with the most occurrences among all possible estimated values ​​as the center, and the number of estimated values ​​falling in the area is counted. The central grid point of the area corresponding to the largest number is taken as the final on-grid estimated coordinate.

[0080]

[0081] S33, calculating the estimated sound source strength on the network

[0082]

[0083] S4: Correct the estimated results by Newton's method to improve positioning accuracy. This is divided into two steps. First, a single-step correction is performed on the online estimation results obtained in the current iteration. Then, based on the current latest residual, a multi-step correction is performed on all the estimated results in turn:

[0084] S41, performing a single-step correction on the current online estimation result by using Newton's method;

[0085]

[0086]

[0087] In the formula, is the Hessian matrix, is the Jacobian matrix.

[0088] S42, update residual Y res , atomic support set A sup , coordinate support set R sup And the source strength support set Q sup ;

[0089]

[0090] S43, based on the latest current residual, all estimated results are corrected in multiple steps in sequence; first, the source strength is updated by the least squares method to complete the orthogonalization operation, and then the result is corrected by the Newton method;

[0091]

[0092] S5: Set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area:

[0093] S51, set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area. Update by least squares method Y estimated and Y res , calculate Tol1 and Tol2, determine whether the iteration stop condition is met, and the iteration terminates when (Tol1≤ε1) or (Tol2≤ε2).

[0094]

[0095] Tol1=∑(|Y res0 | 2 -|Y res |2 ) / ∑(|Y res0 | 2 )

[0096] Tol2=∑(|Y res0 | 2 -|Y res | 2 ) / ∑(|Y Estimated | 2 )

[0097] After the iteration stops, the coordinate support set R is output sup and source strong support set Q sup , that is, the estimated results of the sound source coordinates and their corresponding sound source intensity are obtained.

[0098] In this example, the beneficial effects of the present invention are demonstrated through simulation test verification and field test verification:

[0099] Simulation test: A microphone array with a diameter of 0.32m and an array element number of M = 128 was used for the simulation test. The target sound source was located 2m in front of the array and the area was 16m 2 In the target area, the target area is divided with a spacing of d = 0.1m, and a total of 1681 grid points are obtained. Assume that there are three sound sources with a frequency of f = 20kHz and a sound pressure level of 100dB in the target area, and sample at a sampling rate of 192k. In order to fully reflect the positioning performance of the three algorithms, we set up two types of positioning situations: separated sound sources and adjacent sound sources according to the distance between the sound sources. Without loss of generality, we assume that the coordinates of the sound sources in the separated situation are (-1.1, -1.3, 2), (0.16, 0.93, 2), (1.07, -0.73, 2), and the coordinates of the sound sources in the adjacent situation are (-1.1, -1.3, 2), (0.16, 0.93, 2), (0.16, 0.83, 2). The positioning test was carried out in a simulated environment with a signal-to-noise ratio (SNR) of 0dB and a signal length of 100 snapshots.

[0100] The above sound source is identified by using the method and Newton orthogonal matching pursuit algorithm respectively. The results are as follows: Figure 3 As shown in the figure, "*" and "Ο" represent the estimated sound source position and the real sound source position respectively. Figure 3 (a) and Figure 3 (b) It can be seen that for the separated sound sources, both the method of the present invention and the Newton orthogonal matching pursuit algorithm can correctly locate the real sound source. Figure 3 (c) and Figure 3 (d) It can be seen that the off-grid compensation effect of the Newton orthogonal matching pursuit algorithm is affected when the sound source is adjacent, while the positioning effect of the present invention is still very good. Therefore, it can be seen that the present invention has a more stable and accurate positioning effect.

[0101] Monte Carlo random simulation test: simulates random tests of different algorithms under various conditions. For each algorithm, the SNR value is taken at intervals of 5dB in the range of -10dB to 30dB, and positioning tests are carried out under various signal-to-noise ratio conditions with the number of snapshots L = 100. 500 random tests are performed for each case. To evaluate the performance of the algorithm, we set four evaluation parameters: correct recognition rate CRR, root mean square error RMSE, cumulative recognition correct rate CCRR, and average amplitude error MAE.

[0102] from Figure 4 (a) It can be seen that the CRR of the proposed method is close to 95% at 0dB. Compared with the existing orthogonal matching pursuit algorithm and Newton orthogonal matching pursuit algorithm, the correct recognition rate of the proposed method is significantly increased under low signal-to-noise ratio, and the anti-noise robustness is significantly better. In addition, Figure 4 The size of CCRR in (c) also illustrates the superiority of the present invention. Figure 4 (b) Under medium and high signal-to-noise ratio, the positioning deviation of the orthogonal matching pursuit algorithm is around 4.3 cm, the positioning deviation of the Newton orthogonal matching pursuit algorithm is around 3.3 cm, and the positioning deviation of the present invention is around 2 cm, indicating that the proposed algorithm has good positioning accuracy. Figure 4 (d) The source strength estimation results of the proposed algorithm in the entire signal-noise space are closer to the actual sound source strength.

[0103] Field test verification: A 128-element array with a sampling rate of 192k and a diameter of about 0.16m arranged in a spiral structure was used for positioning experiments. Three sound sources (speakers) with consistent intensity and random positions were placed in a plane 1.5m away from the array plate. The sound sources were emitted by Philips S1009. The positioning of the three algorithms at a frequency of 20k was compared. A single positioning was performed with 100 sampling points, i.e. 100 snapshots. The positioning results are as follows: Figure 5 As shown, Figure 5 (a) is the positioning result of the multi-snapshot orthogonal matching pursuit method; Figure 5 (b) is the positioning result of the multi-shot Newton orthogonal matching pursuit method; Figure 5 (c) is the positioning result of the method of the present invention. It can be seen from the results that in this experimental scenario, the orthogonal matching pursuit algorithm has an obvious grid mismatch problem. Although the Newton orthogonal matching pursuit algorithm corrects the positioning deviation on the grid, the correction effect is limited. In contrast, the present method can basically locate the center of the real sound source.

[0104] The results of simulation tests and field tests show that the method has better sound source localization capability than the prior art.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. A comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method, characterized by: The method comprises the following steps: S1: Based on the measurement data of the original M array, the improved comprehensive orthogonal matching pursuit method is used to obtain the candidate set of sound source coordinates, and the target sound source area is reconstructed by locally refining the grid around it to construct a new perception matrix A. sub ; S2: According to the principle of overall sparseness of the array, multiple groups of M sub The sub-arrays are composed of array elements, the measurement data corresponding to each sub-array is constructed, and the maximum likelihood function is constructed; S3: Based on the reconstructed target grid area, a rough on-grid estimation result of the sound source is obtained; S4: Correct the estimated results by Newton's method to improve positioning accuracy. This is divided into two steps: first, a single-step correction is performed on the online estimation results obtained in the current iteration, and then, based on the current latest residual, a multi-step correction is performed on all the estimated results in sequence. S5: Set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area.

2. The comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method according to claim 1, characterized in that: The S1 comprises the following steps: S11, divide the plane where the sound source is located into Ng grid points according to a certain spatial spacing d, and calculate the a(r ng ) is normalized to get a n (r ng ), a(r ng ) is the spatial coordinate r of the ngth grid ng To the array element space coordinate r composed of M array elements mic.k ,k=1,2,...,the guiding vector (atom) between M; a n (r ng )=a(r ng ) / ||a(r ng )||2,ng=1,2,...,Ng a(r ng )=[g(r ng ,r mic.k ),g(r ng ,r mic.2 ),...,g(r ng ,r mic.M )] T Among them, g(r ng ,r mic.m ) is the Green function, ||·||2 represents the 2-norm operation, [·] T represents transpose, f is the frequency of the sound source, and c is the speed of sound; S12, in each iteration, by calculating the current residual Y res =[y(1), y(2), ..., y(L)], where L is the number of snapshots and the inner product of each column of the normalized atom in the dictionary. The atoms corresponding to the first three largest inner product values ​​are selected, and a circular area that just contains its neighboring atoms is covered with the coordinates of the atom with the largest inner product value as the center. The candidate set of the sound source coordinates is determined by judging whether the other two atoms fall within the circle. The coordinates of the atoms outside the area and the coordinates of the atoms with the largest inner product value together constitute the candidate set of coordinates R su ; R su =R su {r k.1 ,r k.i ⊙(r k.1 )},r k ={r k.1 ,r k.2 ,r k.3 } Where ⊙(r k.1 ) represents the coordinate r k.1 The center covers a certain range of circular space. Like the orthogonal matching pursuit method, in each iteration, the coordinates of the atom with the maximum inner product value are used as the estimation result to promote the iteration process. S13. Based on the atomic coordinates in the candidate set, half of the original space grid spacing d is used as the new grid spacing, and a square grid space with a radius of 1.5d is radiated. These local grid spaces together constitute our new target sound source area R Ns , and construct a new perception matrix A sub , which is composed of the steering vectors corresponding to all grid points in the target area.

3. The comprehensive multi-snapshot combined Newton orthogonal matching tracking positioning method according to claim 1, characterized in that: The S2 comprises the following steps: S21, randomly select L from the original M array of elements according to the overall sparseness principle of the array A Group by M sub The corresponding array data together constitute a new measurement signal in Represents the first A The measurement data corresponding to each sub-array; S22. For each subarray, based on By minimizing the residual energy Obtain the maximum likelihood solution of the sound source coordinates and intensity, and construct the maximum likelihood function in is the perception matrix corresponding to the real sound source coordinates, Corresponding source strength, is noise, ||·|| F represents the Frobenius norm, represents a real number, (·) H stands for conjugate transpose.

4. The comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method according to claim 1, characterized in that: The S3 comprises the following steps: S31, by estimating a sound source estimate in each iteration process, a maximum likelihood ratio test cost function GLRT(·) is established according to the least squares method; S32. For each subarray, in each iteration, the coordinates corresponding to the first three maximum values ​​of GLRT(·) are taken as the current possible estimated values. A circular area with a radius that just contains its neighboring grid points is radiated from the estimated value with the most occurrences among all possible estimated values ​​as the center, and the number of estimated values ​​falling in the area is counted. The central grid point of the area corresponding to the largest number is taken as the final on-grid estimated coordinate. count(g) represents the number of counts that meet the conditions; S33, calculating the estimated sound source strength on the network 5. The comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method according to claim 1, characterized in that: The S4 comprises the following steps: S41, performing a single-step correction on the current online estimation result by using Newton's method; In the formula, is the Hessian matrix, is the Jacobian matrix; S42, update residual Y res , atomic support set A sup , coordinate support set R sup And the source strength support set Q sup ; S43, based on the latest current residual, all estimated results are corrected in multiple steps in sequence; first, the source strength is updated by the least squares method to complete the orthogonalization operation, and then the result is corrected by the Newton method; 6. The comprehensive multi-snapshot combined with Newton orthogonal matching tracking positioning method according to claim 1, characterized in that: The S5 comprises the following steps: S51, set up a global feedback link to provide an opportunity to make up for the missed sound source in the process of S1 reconstructing the sound source area; update it through the least squares method Y estimated and Y res , calculate Tol1 and Tol2, determine whether the iteration stop condition is met, and the iteration terminates when (Tol1≤ε1) or (Tol2≤ε2); where Y res0 Refers to the measurement residual before each iteration; Tol1=Σ(|Y res0 | 2 -|And res | 2 ) / ∑(|Y res0 | 2 ) Tol2=∑(|Y res0 | 2 -|And res | 2 ) / Σ(|Y Estimated | 2 ) After the iteration stops, the coordinate support set R is output sup and source strong support set Q sup , that is, the estimated results of the sound source coordinates and their corresponding sound source intensity are obtained.

Images