Noise elimination method and system based on matrix reconstruction, terminal and storage medium

Abstract

The invention discloses a noise elimination method and system based on matrix reconstruction, a terminal and a storage medium, and relates to the technical field of underwater acoustic array signal processing. The noise elimination method comprises the following steps: receiving a sound source signal; establishing a noise-containing data model of the sound source signal; obtaining a data snapshot sample of the sound source signal based on the noise-containing data model, and calculating a sample covariance matrix; performing characteristic decomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace; obtaining a distance relationship between two projection matrixes formed by each column vector in the signal subspace and the noise subspace based on a translation invariant subspace theory, and establishing a diagonal matrix based on the distance relationship between the two projection matrixes; and reconstructing a sample covariance matrix by using the diagonal matrix to realize noise elimination. The method has lower sidelobe, better robustness and overall performance, can adaptively process data, does not need to adjust the diagonal load shedding coefficient, and can improve the angular resolution of a target signal to a greater extent.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater acoustic array signal processing, and in particular to a noise elimination method, system, terminal and storage medium based on matrix reconstruction. Background Art

[0002] In the complex and ever-changing underwater environment, there are noises such as ocean self-noise, ship noise, and biological sounds. These noises cause the performance of traditional Direction of Arrival (DOA) estimation methods to be seriously degraded, bringing certain difficulties and challenges to sonar array signal processing. Underwater DOA estimation is an important research direction in the field of array signal processing. In practical applications, the underwater noise field is very complex, and the power of the noise received by each array element is not equal, which will seriously affect the angular resolution of underwater targets.

[0003] Conventional beamforming methods achieve the "guiding" effect of the beam by adjusting the weighting coefficients, so that the beamformer can respond to signals in a certain direction in space and suppress signals in other directions. However, in the face of complex ocean noise environments, the resolution of conventional beamforming methods is difficult to meet the requirements of practical applications. At present, the noise elimination method based on diagonal unloading is widely used. The core idea of ​​this method is to subtract as much noise components on the diagonal of the covariance matrix as possible and reconstruct the sample covariance matrix to achieve the purpose of noise elimination, so that the output signal-to-noise ratio is maximized. However, this method requires continuous adjustment of the diagonal unloading coefficient to maintain a compromise between robustness and array gain, and is not suitable for complex underwater noise environments. Summary of the invention

[0004] The purpose of the present invention is to provide a noise elimination method, system, terminal and storage medium based on matrix reconstruction, which utilizes the difference between signal subspace and noise subspace to construct a diagonal matrix, reconstruct the sample covariance matrix and improve the performance of underwater noise elimination.

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

[0006] In a first aspect, the present invention provides a noise elimination method based on matrix reconstruction, comprising the following steps:

[0007] S1. Receive sound source signal;

[0008] S2. Establishing a noisy data model of the sound source signal;

[0009] S3. Obtain data snapshot samples of the sound source signal based on the noisy data model and calculate the sample covariance matrix;

[0010] S4. Perform eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace;

[0011] S5. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, and a diagonal matrix is ​​established based on the distance relationship between the two projection matrices;

[0012] S6. Use the diagonal matrix to reconstruct the sample covariance matrix to eliminate noise.

[0013] As a possible implementation method, a noisy data model of the sound source signal is established, specifically:

[0014] X=AS+N

[0015] Where X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array.

[0016] As a possible implementation method, data snapshot samples of the sound source signal are obtained based on the noisy data model, and the sample covariance matrix is ​​calculated, specifically:

[0017]

[0018] Where R is the sample covariance matrix, L is the number of snapshots, H represents the conjugate transpose operation of the matrix, and X l represents the noisy data received by the array at the lth snapshot.

[0019] As a possible implementation method, the sample covariance matrix is ​​eigen-decomposed to obtain the signal subspace and the noise subspace, specifically:

[0020] R=UΣU H

[0021] Among them, U=[u1,u2,…,u N ], is the eigenvector matrix, u i is the i-th eigenvector, 1≤i≤N; Σ=diag[λ1,λ2,…,λ N ], is the diagonal matrix composed of eigenvalues, λ i is the i-th eigenvalue, 1≤i≤N; and the eigenvalue λ i Arrange in order from largest to smallest;

[0022] The eigenvectors corresponding to the first Q eigenvalues ​​form the signal subspace: U Q =[u1,u2,…,u Q ];

[0023] The eigenvectors corresponding to the last NQ eigenvalues ​​form the noise subspace: U N =[u Q+1 ,uQ+2 ,…,u N ].

[0024] As a possible implementation method, S5 specifically includes the following sub-steps:

[0025] S50. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, specifically:

[0026] When the number of snapshots L→∞, there exists a square non-singular matrix F such that A≈U Q F, where A is the steering vector matrix of the incident signal and Q is the number of signal sources; according to the translation invariant subspace theory, we can get:

[0027]

[0028] in, i∈[1,Q], is the column vector u i The matrix obtained by deleting the first row of elements is is the column vector u i The matrix obtained by deleting the last row of elements is for The projection matrix, for The projection matrix;

[0029] S51. According to the difference between each pair of projection matrices in the signal subspace and the noise subspace, the distance relationship expression between the two projection matrices is configured:

[0030]

[0031] in, is the column vector u p The matrix obtained by deleting the first row is is the column vector u p The matrix obtained by deleting the last row, 1≤p≤N, ||·|| F represents the Frobenius norm, tr(·) represents the trace operator, and N is the number of array elements; for The projection matrix, for The projection matrix;

[0032] S52. Establish a diagonal matrix based on the distance relationship expression:

[0033]

[0034] Among them, diag means taking diagonal elements.

[0035] As a possible implementation manner, S6 specifically includes the following sub-steps:

[0036] S60. Subtract the matrix obtained by conjugate multiplication of the diagonal matrices from the identity matrix to obtain a new diagonal matrix:

[0037] V=I-HH H

[0038] S61. Reconstruct the sample covariance matrix using a new diagonal matrix to eliminate the noise in the sample covariance matrix R, specifically:

[0039] R opt =VR

[0040] P(θ h )=a(θ h )R opt a Η (θ h )

[0041] Among them, θ h ∈[0°,180°], R is the sample covariance matrix, P(θ h ) is the angle θ h The power spectrum of the direction, a(θ h ) is the angle θ h Direction of the steering vector, R opt Represents the reconstructed sample covariance matrix.

[0042] In a second aspect, the present invention provides a noise elimination system based on matrix reconstruction, comprising:

[0043] A noisy data model building unit is configured with a uniform linear array, and uses the uniform linear array to receive a sound source signal, and builds a noisy data model of the sound source signal: X=AS+N, wherein X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array;

[0044] A sample covariance matrix solving unit, which obtains data snapshot samples of the sound source signal based on the noisy data model and calculates the sample covariance matrix;

[0045] A diagonal matrix solving unit performs eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace, obtains a distance relationship between two projection matrices formed by each column vector in the signal subspace and the noise subspace based on the translation invariant subspace theory, configures a distance relationship expression between the two projection matrices according to the difference between each pair of projection matrices in the signal subspace and the noise subspace, and establishes a diagonal matrix based on the distance relationship expression;

[0046] The noise elimination unit uses a unit matrix to subtract a matrix obtained by conjugate multiplication of the diagonal matrix to obtain a new diagonal matrix, and uses the new diagonal matrix to reconstruct the sample covariance matrix to eliminate noise in the sample covariance matrix.

[0047] In a third aspect, the present invention provides a terminal, including a processor and a communication interface coupled to the processor, wherein the processor is used to run a computer program or instruction to implement the noise elimination method based on matrix reconstruction provided in the first aspect.

[0048] In a fourth aspect, the present invention provides a storage medium storing a program, wherein the program is used to be executed by a processor to execute the noise elimination method based on matrix reconstruction provided in the first aspect.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] 1. The noise elimination method based on matrix reconstruction provided by the present invention is suitable for complex underwater noise environments and has higher resolution than traditional conventional beamforming methods.

[0051] 2. The noise elimination method based on matrix reconstruction provided by the present invention is based on the theory of translation invariant subspace. It utilizes the distance relationship between each pair of projection matrices in the signal subspace and the noise subspace to construct them into a diagonal matrix, thereby reconstructing the sample covariance matrix to eliminate noise. It can process data adaptively without adjusting the diagonal unloading coefficient.

[0052] 3. Experimental data show that the noise elimination method based on matrix reconstruction provided by the present invention has lower side lobes, better robustness and overall performance. Compared with conventional beamforming methods and noise elimination methods based on diagonal load shedding, it can improve the angular resolution of the target signal to a greater extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0054] Figure 1 is a flow chart of a noise elimination method based on matrix reconstruction in an embodiment of the present invention;

[0055] Figure 2 1 is a diagram showing the result of processing the sound source signal data of the conventional beamforming method in an embodiment of the present invention;

[0056] Figure 31. It is a diagram showing the result of processing the sound source signal data of the noise elimination method based on diagonal load shedding in an embodiment of the present invention;

[0057] Figure 4 1 is a diagram showing the result of processing the sound source signal data of the noise elimination method based on matrix reconstruction in an embodiment of the present invention;

[0058] Figure 5 It is a diagram of spatial spectrum estimation results of sound source signal data processing at a certain moment by a conventional beamforming method, a noise elimination method based on diagonal load shedding, and a noise elimination method based on matrix reconstruction in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, words such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and effects. For example, the first threshold and the second threshold are only used to distinguish different thresholds, and their order is not limited. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different.

[0060] It should be noted that, in the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0061] In the present invention, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " generally indicates that the associated objects before and after are in an "or" relationship. The following at least one item (items) or similar expressions thereof refer to any combination of these items, including any combination of single items (items) or plural items (items). For example, at least one item (items) of a, b or c can mean: a, b, c, the combination of a and b, the combination of a and c, the combination of b and c, or the combination of a, b and c, where a, b, c can be single or multiple.

[0062] In a complex and changeable underwater environment, the underwater noise field is very complex, and the power of the noise received by each array element is not equal, which will seriously affect the angular resolution of underwater targets and bring certain difficulties and challenges to sonar array signal processing. To solve the above problems, the present invention provides a noise elimination method, system, terminal, and readable storage medium based on matrix reconstruction, which utilizes the difference between the signal subspace and the noise subspace to construct a diagonal matrix, reconstruct the sample covariance matrix, and improve the performance of underwater noise elimination.

[0063] In a first aspect, the present invention provides a noise elimination method based on matrix reconstruction, see Figure 1 , including the following steps:

[0064] S1. Receive sound source signal;

[0065] As an example, a uniform linear array composed of hydrophone array elements is placed underwater to receive sound source signals.

[0066] S2. Establishing a noisy data model of the sound source signal;

[0067] As a possible implementation method, a noisy data model of the sound source signal is established, specifically:

[0068] X=AS+N

[0069] Where X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array.

[0070] S3. Obtain data snapshot samples of the sound source signal based on the noisy data model and calculate the sample covariance matrix;

[0071] As a possible implementation method, data snapshot samples of the sound source signal are obtained based on the noisy data model, and the sample covariance matrix is ​​calculated, specifically:

[0072]

[0073] Where R is the sample covariance matrix, L is the number of snapshots, H represents the conjugate transpose operation of the matrix, and X l represents the noisy data received by the array at the lth snapshot.

[0074] S4. Perform eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace;

[0075] As a possible implementation method, the sample covariance matrix is ​​eigen-decomposed to obtain the signal subspace and the noise subspace, specifically:

[0076] R=UΣU H

[0077] Among them, U=[u1,u2,…,u N ], is the eigenvector matrix, u i is the i-th eigenvector, 1≤i≤N; Σ=diag[λ1,λ2,…,λ N ], is the diagonal matrix composed of eigenvalues, λ i is the i-th eigenvalue, 1≤i≤N; and the eigenvalue λ i Arrange in order from largest to smallest;

[0078] The eigenvectors corresponding to the first Q eigenvalues ​​form the signal subspace: U Q =[u1,u2,…,u Q ];

[0079] The eigenvectors corresponding to the last NQ eigenvalues ​​form the noise subspace: U N =[u Q+1 ,u Q+2 ,…,u N ].

[0080] S5. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, and a diagonal matrix is ​​established based on the distance relationship between the two projection matrices;

[0081] As a possible implementation method, S5 specifically includes the following sub-steps:

[0082] S50. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, specifically:

[0083] When the number of snapshots L→∞, there exists a square non-singular matrix F such that A≈U Q F, where A is the steering vector matrix of the incident signal and Q is the number of signal sources; according to the translation invariant subspace theory, we can get:

[0084]

[0085] in, i∈[1,Q], is the column vector u i The matrix obtained by deleting the first row of elements is is the column vector u i The matrix obtained by deleting the last row of elements is for The projection matrix, for The projection matrix;

[0086] S51. According to the difference between each pair of projection matrices in the signal subspace and the noise subspace, the distance relationship expression between the two projection matrices is configured:

[0087]

[0088] in, is the column vector u p The matrix obtained by deleting the first row is is the column vector u p The matrix obtained by deleting the last row, 1≤p≤N, ‖·‖ F represents the Frobenius norm, tr(·) represents the trace operator, and N is the number of array elements; for The projection matrix, for The projection matrix;

[0089] S52. Establish a diagonal matrix based on the distance relationship expression:

[0090]

[0091] Among them, diag means taking diagonal elements.

[0092] S6. Use the diagonal matrix to reconstruct the sample covariance matrix to eliminate noise.

[0093] As a possible implementation manner, S6 specifically includes the following sub-steps:

[0094] S60. Subtract the matrix obtained by conjugate multiplication of the diagonal matrices from the identity matrix to obtain a new diagonal matrix:

[0095] V=I-HH H

[0096] S61. Reconstruct the sample covariance matrix using a new diagonal matrix to eliminate the noise in the sample covariance matrix R, specifically:

[0097] R opt =VR

[0098] P(θ h )=a(θ h )R opt a Η (θ h )

[0099] Among them, θ h ∈[0°,180°], R is the sample covariance matrix, P(θ h ) is the angle θ h The power spectrum of the direction, a(θh ) is the angle θ h Direction of the steering vector, R opt represents the reconstructed sample covariance matrix, and H represents the conjugate transpose operation of the matrix.

[0100] In a possible implementation, a conventional beamforming method, a noise elimination method based on diagonal load shedding, and a noise elimination method based on matrix reconstruction provided by the present invention are respectively selected to compare noise elimination performance. The embodiments are as follows:

[0101] A uniform linear array consisting of 40 sensors is used. The sound source is set 90m below the water surface. The frequency range of the sound source signal is 180-280Hz, with a total of 101 frequency points. The sampling frequency is 8kHz, and the processed signal data length is 60 seconds. Frequency domain sampling is performed every 6 seconds. The data overlap rate is 50%, with a total of 10 snapshots. The water sound speed at the receiving array depth is 1518m / s, and the angle search interval is 0.1°.

[0102] See also Figure 2 , is the result of the sound source signal data processing of the conventional beamforming method. The two targets move relative to each other from about 30° to 110° and from 105° to 90° respectively. It can be seen from the figure that the existence of ocean background noise affects the angular resolution of the target to a certain extent.

[0103] See also Figure 3 , is the sound source signal data processing result of the noise elimination method based on diagonal load reduction. It can be seen from the figure that compared with Figure 2 The processing results of the conventional beamforming method and the noise elimination method based on diagonal load shedding reduce the side lobes near the target by more than 6dB, which greatly improves the angular resolution of the target signal.

[0104] See also Figure 4 , which is the sound source signal data processing result of the noise elimination method based on matrix reconstruction provided by the present invention. It can be seen from the figure that the present invention reduces the side lobe near the target by more than 5dB at most, better improves the angular resolution of the target signal, and has better robustness.

[0105] See also Figure 5 , which is the spatial spectrum estimation result of the actual sound source signal data processing of the above three methods at a certain moment. It can be seen from the normalized power spectra of the three methods at a single time point that, compared with the conventional beamforming method and the noise elimination method based on diagonal load reduction, the sidelobe level of the method of the present invention is reduced by more than 10dB, the overall performance is better, and the angular resolution of the target signal is greatly improved.

[0106] In a second aspect, the present invention provides a noise elimination system based on matrix reconstruction, comprising:

[0107] A noisy data model building unit is configured with a uniform linear array, and uses the uniform linear array to receive a sound source signal, and builds a noisy data model of the sound source signal: X=AS+N, wherein X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array;

[0108] A sample covariance matrix solving unit, which obtains data snapshot samples of the sound source signal based on the noisy data model and calculates the sample covariance matrix;

[0109] A diagonal matrix solving unit performs eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace, obtains a distance relationship between two projection matrices formed by each column vector in the signal subspace and the noise subspace based on the translation invariant subspace theory, configures a distance relationship expression between the two projection matrices according to the difference between each pair of projection matrices in the signal subspace and the noise subspace, and establishes a diagonal matrix based on the distance relationship expression;

[0110] The noise elimination unit uses a unit matrix to subtract a matrix obtained by conjugate multiplication of the diagonal matrix to obtain a new diagonal matrix, and uses the new diagonal matrix to reconstruct the sample covariance matrix to eliminate noise in the sample covariance matrix.

[0111] In a third aspect, the present invention provides a terminal, including a processor and a communication interface coupled to the processor, wherein the processor is used to run a computer program or instruction to implement the noise elimination method based on matrix reconstruction provided in the first aspect.

[0112] In a fourth aspect, the present invention provides a storage medium storing a program, wherein the program is used to be executed by a processor to execute the noise elimination method based on matrix reconstruction provided in the first aspect.

[0113] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the drawings, etc. In the specification, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the specification. Certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0114] Although the present invention has been described in conjunction with specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present invention. Accordingly, this specification and the accompanying drawings are merely exemplary illustrations of the present invention and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present invention. Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, the present invention is intended to include such modifications and variations if they fall within the scope of the present invention and its equivalents.

Claims

1. A noise elimination method based on matrix reconstruction, characterized in that: The steps include: S1. Receive sound source signal; S2. Establishing a noisy data model of the sound source signal; S3. Acquire data snapshot samples of the sound source signal based on the noisy data model and calculate the sample covariance matrix; S4. Performing eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace; S5. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, and a diagonal matrix is ​​established based on the distance relationship between the two projection matrices; S6. Reconstruct the sample covariance matrix using the diagonal matrix to eliminate noise.

2. The noise elimination method based on matrix reconstruction according to claim 1, characterized in that: The step of establishing the noise-containing data model of the sound source signal is specifically as follows: X=AS+N Where X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array.

3. The noise elimination method based on matrix reconstruction according to claim 1, characterized in that: The data snapshot samples of the sound source signal are obtained based on the noisy data model, and the sample covariance matrix is ​​calculated, specifically: Where R is the sample covariance matrix, L is the number of snapshots, H represents the conjugate transpose operation of the matrix, and X l represents the noisy data received by the array at the lth snapshot.

4. The noise elimination method based on matrix reconstruction according to claim 3, characterized in that: The sample covariance matrix is ​​subjected to eigendecomposition to obtain a signal subspace and a noise subspace, specifically: R=UΣU H Among them, U=[u1,u2,…,u N ], is the eigenvector matrix, u i is the i-th eigenvector, 1≤i≤N; Σ=diag[λ1,λ2,…,λ N ], is the diagonal matrix composed of eigenvalues, λ i is the i-th eigenvalue, 1≤i≤N; and the eigenvalue λ i Arrange in order from largest to smallest; The eigenvectors corresponding to the first Q eigenvalues ​​form the signal subspace: U Q =[u1,u2,…,u Q ]; The eigenvectors corresponding to the last NQ eigenvalues ​​form the noise subspace: U N =[u Q+1 ,u Q+2 ,…,u N ].

5. The noise elimination method based on matrix reconstruction according to claim 1, characterized in that: The S5 specifically includes the following sub-steps: S50. Based on the translation invariant subspace theory, the distance relationship between the two projection matrices formed by each column vector in the signal subspace and the noise subspace is obtained, specifically: When the number of snapshots L→∞, there exists a square non-singular matrix F such that A≈U Q F, where A is the steering vector matrix of the incident signal and Q is the number of signal sources; according to the translation invariant subspace theory, we can get: in, is the column vector u i The matrix obtained by deleting the first row of elements is is the column vector u i The matrix obtained by deleting the last row of elements is for The projection matrix, for The projection matrix; S51. According to the difference between each pair of projection matrices in the signal subspace and the noise subspace, the distance relationship expression between the two projection matrices is configured: in, is the column vector u p The matrix obtained by deleting the first row is is the column vector u p The matrix obtained by deleting the last row, 1≤p≤N, ||·|| F represents the Frobenius norm, tr(·) represents the trace operator, and N is the number of array elements; for The projection matrix, for The projection matrix; S52. Establish a diagonal matrix based on the distance relationship expression: Among them, diag means taking diagonal elements.

6. The noise elimination method based on matrix reconstruction according to claim 5, characterized in that: The S6 specifically includes the following sub-steps: S60. Subtract the matrix obtained by conjugate multiplication of the diagonal matrices from the identity matrix to obtain a new diagonal matrix: V=I-HH H S61. Reconstruct the sample covariance matrix using a new diagonal matrix to eliminate the noise in the sample covariance matrix R, specifically: R opt =VR P(θ h )=a(θ h )R opt a Η (i h ) Among them, θ h ∈[0°,180°], R is the sample covariance matrix, P(θ h ) is the angle θ h The power spectrum of the direction, a(θ h ) is the angle θ h Direction of the steering vector, R opt Represents the reconstructed sample covariance matrix.

7. A noise elimination system based on matrix reconstruction, characterized in that: include: A noisy data model building unit is configured with a uniform linear array, and uses the uniform linear array to receive a sound source signal, and builds a noisy data model of the sound source signal: X=AS+N, wherein X is the noisy data received by the array, A is the steering vector matrix of the incident signal, S is the incident signal matrix, and N is the noise matrix received by the array; A sample covariance matrix solving unit, which obtains data snapshot samples of the sound source signal based on the noisy data model and calculates the sample covariance matrix; A diagonal matrix solving unit performs eigendecomposition on the sample covariance matrix to obtain a signal subspace and a noise subspace, obtains a distance relationship between two projection matrices formed by each column vector in the signal subspace and the noise subspace based on the translation invariant subspace theory, configures a distance relationship expression between the two projection matrices according to the difference between each pair of projection matrices in the signal subspace and the noise subspace, and establishes a diagonal matrix based on the distance relationship expression; The noise elimination unit uses a unit matrix to subtract a matrix obtained by conjugate multiplication of the diagonal matrix to obtain a new diagonal matrix, and uses the new diagonal matrix to reconstruct the sample covariance matrix to eliminate noise in the sample covariance matrix.

8. A terminal comprising a processor and a communication interface coupled to the processor, wherein the processor is used to run a computer program or instruction to implement the noise elimination method based on matrix reconstruction according to any one of claims 1 to 7.

9. A storage medium, characterized in that: The storage medium stores a program, and the program is used to be run by a processor to execute the noise elimination method based on matrix reconstruction according to any one of claims 1 to 7.

Images