Coherent signal decoherence method based on noise suppression

By constructing an interpolated virtual uniform array model on a coprime array and performing covariance matrix pre-whitening and sparse recovery, the problem of DOA estimation accuracy under the influence of non-uniform noise and coherent signals is solved, and high-precision DOA estimation is achieved.

CN118962580APending Publication Date: 2024-11-15Chinese People's Liberation Army Cyberspace Force Information Engineering University
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411071884.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Existing technologies have insufficient accuracy in direction-of-arrival (DOA) estimation under non-uniform noise and coherent signal conditions. Furthermore, conventional methods involve large computational loads and are only applicable to uniform arrays, resulting in array aperture loss.

Method used

By using the array interpolation concept, virtual sensors are filled with physical holes in coprime arrays to construct an interpolated virtual uniform array model. Signal initialization and covariance matrix pre-whitening are performed, and sparse recovery is achieved by utilizing the low-rank property of the Toeplitz matrix. DOA estimation is then performed in conjunction with the MUSIC algorithm.

Benefits of technology

High-precision coherent signal DOA estimation was achieved in non-uniform noise environments, avoiding array aperture loss and improving estimation performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118962580A_ABST
    Figure CN118962580A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of signal processing, in particular to a coherent signal decoherence method based on noise suppression, and the method comprises the steps: building a receiving signal model; physical hole filling virtual sensors in the co-prime array are equivalent to an interpolation virtual uniform array, and then an interpolation virtual uniform array receiving signal model is constructed; initializing the interpolation virtual uniform array receiving signal model, and obtaining a sampling covariance matrix of the initialized signal model; decomposing the sampling covariance matrix and replacing diagonal elements to obtain a pre-whitened covariance matrix, and suppressing non-uniform noise; the method comprises the following steps: pre-whitening a covariance matrix, selecting a first row of the pre-whitened covariance matrix to reconstruct a Toeplitz matrix, performing sparse recovery on the Toeplitz matrix by using a low-rank characteristic of the reconstructed Toeplitz matrix to obtain T-(m), performing characteristic decomposition on the matrix T-(m), and estimating a target DOA by using a MUSIC algorithm. According to the invention, DOA estimation of coherent signals in a non-uniform noise environment is realized, and the DOA estimation precision is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and in particular to a coherent signal decorrelation method based on noise suppression. Background Art

[0002] Direction of Arrival (DOA) estimation is an important branch of array signal processing and is widely used in radar, wireless communication and other fields. Due to the influence of multipath propagation, coherent signals that cause covariance matrix rank deficiency are widely present in practical applications. At the same time, due to the imperfect hardware of the receiving channel or the uncalibrated antenna array, non-uniform noise is prevalent in the array. The presence of non-uniform noise and coherent signals will cause the performance of conventional subspace methods to degrade or even fail to accurately estimate. Therefore, a DOA estimation method is considered to suppress multipath interference while suppressing non-uniform noise.

[0003] Under the conditions of coherent signals and non-uniform noise, the current method is usually applied to uniform linear arrays, suppressing the influence of non-uniform noise and multipath by combining spatial smoothing technology and iterative technology. However, this method has too much computational complexity when estimating the noise covariance matrix. At the same time, it is only applied to uniform arrays, has the problem of array aperture loss, and requires the number of signal sources to be less than the number of array elements. Summary of the invention

[0004] The purpose of the present invention is to solve the problem in the prior art that non-uniform noise and coherent signals affect the accuracy of DOA estimation, and propose a coherent signal decorrelation method based on noise suppression to realize DOA estimation of coherent signals in a non-uniform noise environment, thereby effectively improving the DOA estimation accuracy.

[0005] To achieve the above purpose, the technical solution adopted is:

[0006] A coherent signal decorrelation method based on noise suppression, comprising:

[0007] Establish a received signal model;

[0008] Based on the array interpolation idea, the virtual sensors filling the physical holes in the coprime array are equivalent to the interpolated virtual uniform array, and then the interpolated virtual uniform array receiving signal model is constructed.

[0009] Initializing the interpolated virtual uniform array receiving signal model and obtaining the sampling covariance matrix of the initialized signal model;

[0010] Decomposing the above sampling covariance matrix and replacing the diagonal elements to obtain the pre-whitened covariance matrix to suppress non-uniform noise;

[0011] The first row of the pre-whitened covariance matrix is ​​selected to reconstruct the Toeplitz matrix. The low-rank property of the reconstructed Toeplitz matrix is ​​used to perform sparse recovery on the Toeplitz matrix to obtain Pair Matrix Perform eigendecomposition and use the MUSIC algorithm to estimate the target DOA.

[0012] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, assuming that K far-field narrowband coherent signal sources are emitted from different directions θ k (k=1,2,...,K) is incident on a coprime array, and the array element position is represented by d0. Then the received signal model expression on the coprime array is:

[0013]

[0014] Where s1(t) is the original reference signal, ρ is a non-zero complex-valued constant coefficient representing the coherence coefficient, is the array flow matrix, is the non-uniform noise matrix on the coprime array.

[0015] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, the physical holes in the coprime array are filled with virtual sensors, which is equivalent to an interpolated virtual uniform array. and Interpolate the uth virtual uniform array i The received signal of each array element is expressed as:

[0016]

[0017] Where λ is the incident wavelength; the interpolation virtual uniform array receiving signal model is expressed as:

[0018]

[0019] Among them, the array flow matrix is the non-uniform noise matrix on the interpolated virtual uniform array.

[0020] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, the interpolation virtual uniform array receiving signal model Initialize it as:

[0021]

[0022] in, Indicates that the lth array element is in the set But not in the collection If it is not in the set The signal received by the array element in is considered to be 0.

[0023] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, obtaining the sampling covariance matrix of the initialized signal model includes:

[0024] The sampling covariance matrix of the interpolated virtual uniform array received signal model is expressed as follows:

[0025]

[0026] Where T represents the number of snapshots, (·) H represents the conjugate transpose of a matrix;

[0027] Then, the signal model after initialization The sampling covariance matrix of is expressed as:

[0028]

[0029] Where (p,q) represents the matrix The element at row p and column q in .

[0030] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, the sampling covariance matrix The breakdown is as follows:

[0031]

[0032] in

[0033]

[0034] in, is a matrix with all 1s except the diagonal which is 0.

[0035] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, replacing the diagonal elements to obtain the pre-whitened covariance matrix includes:

[0036] Replace all diagonal elements in R2 with the smallest diagonal element<diag(R2)> min , the replaced R2 is expressed as:

[0037]

[0038] The covariance matrix after pre-whitening is expressed as:

[0039]

[0040] According to the coherent signal decorrelation method based on noise suppression of the present invention, further, by utilizing the low rank property of the reconstructed Toeplitz matrix, the Toeplitz matrix is ​​sparsely restored to obtain Include:

[0041] Construct the optimization problem:

[0042]

[0043] Where C is a binary matrix with the same dimension as the Toeplitz matrix, whose value is 1 at the non-zero position of the Toeplitz matrix and 0 at the other positions. is the matrix of the Hermitian-Toeplitz structure, rank(·) is the rank of the matrix, represents the Hadamard product;

[0044] Since the rank minimization problem is an NP-hard problem, the above equation is convexly relaxed to the following form:

[0045]

[0046] The final recovery is an estimate of the Toeplitz matrix, which implements pre-whitening of non-uniform noise and decorrelation of coherent signals.

[0047] The beneficial effects achieved by adopting the above technical solution are:

[0048] The coherent signal decorrelation method based on noise suppression of the present invention suppresses the influence of non-uniform noise on a coprime array, thereby accurately estimating the DOA of the coherent signal. Compared with the existing method, the present invention does not cause array aperture loss, suppresses non-uniform noise by decomposing the sampling covariance matrix and replacing the diagonal elements, and sparsely restores the Toeplitz matrix by utilizing the low-rank characteristics of the reconstructed Toeplitz matrix, thereby realizing decorrelation of the coherent signal and achieving better estimation performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings of the embodiments of the present invention, wherein the drawings are only used to illustrate some embodiments of the present invention, but not to limit all embodiments of the present invention thereto.

[0050] Figure 1 is a flow chart of a coherent signal decorrelation method based on noise suppression according to an embodiment of the present invention;

[0051] Figure 2 is the influence of the number of snapshots in the fully coherent signal of the embodiment of the present invention on the estimation accuracy of each method;

[0052] Figure 3 It is the RMSE variation trend of each method in the embodiments of the present invention as the signal-to-noise ratio changes. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings of specific embodiments of the present invention to clearly and completely describe the exemplary scheme of the embodiment of the present invention. Unless otherwise defined, the technical terms or scientific terms used in the present invention should be the common meanings understood by people with ordinary skills in the field.

[0054] like Figure 1 As shown, this embodiment discloses a coherent signal decorrelation method based on noise suppression, which includes the following contents:

[0055] Step S101: Establish a received signal model.

[0056] Assume that K far-field narrowband coherent sources come from different directions θ k (k=1,2,...,K) is incident on a coprime array, and the array element position is represented by d0. Then the received signal model expression on the coprime array is:

[0057]

[0058] Where s1(t) is the original reference signal, ρ is a non-zero complex-valued constant coefficient representing the coherence coefficient, is the array flow matrix, is a non-uniform noise matrix on a coprime array, and the noises are independent of each other and the noises are independent of the signal source.

[0059] Since the K information sources are coherent, the covariance matrix of the received signal can be obtained as:

[0060]

[0061] in, is the power of the original reference signal; the noise is non-uniform white noise, so the covariance matrix of the noise vector is expressed as:

[0062]

[0063] in, For the first i The noise power on each array element.

[0064] Step S102: Based on the array interpolation concept, the virtual sensors filling the physical holes in the coprime array are equivalent to an interpolated virtual uniform array, and then an interpolated virtual uniform array receiving signal model is constructed.

[0065] Fill the physical holes in the coprime array with virtual sensors, which is equivalent to an interpolated virtual uniform array and Interpolate the uth virtual uniform array i The received signal of each array element is expressed as:

[0066]

[0067] Where λ is the incident wavelength; the interpolation virtual uniform array receiving signal model is expressed as:

[0068]

[0069] Among them, the array flow matrix is the non-uniform noise matrix on the interpolated virtual uniform array.

[0070] Step S103: Initialize the interpolated virtual uniform array received signal model and obtain a sampling covariance matrix of the initialized signal model.

[0071] In theory, the covariance matrix of the interpolated virtual uniform array receiving signal model is expressed as:

[0072]

[0073] in, For u i The noise on the interpolated virtual uniform array elements.

[0074] In practical applications, the number of snapshots is limited. It can be equivalent to the sampling covariance matrix, that is

[0075]

[0076] Where T represents the number of snapshots, (·) H Represents the conjugate transpose of a matrix.

[0077] Since virtual sensors are inserted at the holes in the coprime array, the virtual sensors at the holes have virtually no received signals. The received signal model of the interpolated virtual uniform array can be initialized as:

[0078]

[0079] in, Indicates that the lth array element is in the set But not in the collection If it is not in the set The signal received by the array element in can be regarded as 0.

[0080] At this point, the initialized signal model The sampling covariance matrix of can be expressed as:

[0081]

[0082] Where (p,q) represents the matrix The element at row p and column q in .

[0083] Step S104: Sampling the covariance matrix The covariance matrix after pre-whitening is obtained by decomposing and replacing the diagonal elements to suppress non-uniform noise.

[0084] The sampling covariance matrix The breakdown is as follows:

[0085]

[0086] in

[0087]

[0088] in, is a matrix with all 1s except the diagonal which is 0.

[0089] Since the non-uniform noise power is only reflected in the diagonal elements of the array covariance matrix, in order to eliminate The non-uniformity caused by the noise in the Perform pre-whitening. Select the smallest diagonal element in R2 as the replacement element <r2> min , replace all diagonal elements of R2 to reduce the impact of non-uniform noise. At the same time, it should be noted that The diagonal elements of contain 0 elements, that is, the received signal power at the virtual sensor position is 0. The 0 elements in the diagonal should be excluded when selecting the minimum diagonal element.

[0090] Depend on From the calculation formula of R2, we can see that each diagonal element of R2 is the uth i The sum of the power of K signals and the noise power at each array element, so theoretically the difference in diagonal elements is caused by the unevenness of the noise power. Therefore, replace all diagonal elements in R2 with the minimum diagonal element<diag(R2)> min , the replaced R2 is expressed as:

[0091]

[0092] At this point, the pre-whitening of the non-uniform noise is completed, and the covariance matrix after pre-whitening is expressed as:

[0093]

[0094] Step S105: Select the first row of the pre-whitened covariance matrix to reconstruct the Toeplitz matrix, and use the low-rank property of the reconstructed Toeplitz matrix to perform sparse recovery on the Toeplitz matrix to obtain Pair Matrix Perform eigendecomposition and use the MUSIC algorithm to estimate the target DOA.

[0095] Select The first line reconstructs the Toeplitz matrix as follows:

[0096]

[0097] The reconstructed Toeplitz matrix realizes the decorrelation of coherent signals, and the number of signal sources in the matrix T(m) is small relative to the number of array elements, so the matrix T(m) is a low-rank matrix. The low-rank matrix recovery theory is used to optimize the matrix elements.

[0098] Construct the optimization problem:

[0099]

[0100] Where C is a binary matrix with the same dimension as T(m), whose value is 1 at the non-zero position of T(m) and 0 at the other positions. is a matrix of Hermitian-Toeplitz structure, rank(·) is the rank of the matrix, represents the Hadamard product.

[0101] Since the rank minimization problem is an NP-hard problem, the above equation is convexly relaxed to the following form:

[0102]

[0103] The final recovery is the estimated value of T(m), which realizes the pre-whitening of non-uniform noise and the decorrelation of coherent signals. Perform eigendecomposition and use the MUSIC algorithm to search for signal spectrum peaks:

[0104]

[0105] The peak position P(θ) is the estimated value of the direction of arrival, U N is the noise subspace after eigendecomposition, and a(θ) is the direction vector of the signal subspace after eigendecomposition.

[0106] The present invention uses numerical simulation to verify the effectiveness of the proposed method. The main purpose of the proposed method is to decorrelate the coherent signal under the condition of suppressing non-uniform noise. Since the power of non-uniform noise is different, the signal-to-noise ratio is defined as:

[0107]

[0108] in, is the power of the narrowband coherent signal.

[0109] In order to compare the performance of different methods, the root mean square error (RMSE) was obtained through L = 500 Monte Carlo experiments:

[0110]

[0111] in, is the jth source estimated by the lth Monte Carlo experiment, θ j is the real j-th information source.

[0112] We choose to conduct simulation experiments on a coprime array with M=3, N=5. According to the above physical array interpolation technology, we can get the coprime array The number of array elements of the uniform array used in the comparison method is the same as the number of physical array elements of the coprime array.

[0113] First, the DOA estimation performance of this method is compared with that of the four proposed methods. Here we first give some specific details of the parameters in the method. The regularization parameter of the CAANM method is 1, the regularization parameter of the CLMR method is 0.25, the initial noise covariance vector in the ICSSM method is the diagonal element of the sampling array covariance matrix, and the iteration termination criterion is that the difference between the current solution and the previous iteration solution is less than 10 -4 , the regularization parameter in this method is 0.25. In addition, for ease of comparison, the DOA estimation methods in the above methods will use the MUSIC algorithm to perform DOA estimation on the recovered covariance matrix. Assume that the three incident signals are all coherent, the incident angles are [-42.33°, 3.22°, 48.77°], and each method is performed under non-uniform noise conditions. The non-uniform Gaussian noise power covariance matrix is ​​Q=diag(10,5,10,20,15,30,27,18,13,30). The signal-to-noise ratio is 0dB, the number of snapshots ranges from 10 to 3010, the step size is 300, and the impact of the number of snapshots in the fully coherent signal on the estimation accuracy of the method is as follows. Figure 2 As shown in the figure, under the same signal-to-noise ratio condition, with the increase of the number of snapshots, the RMSE curves of the proposed method and the comparison method tend to be stable, and the estimation error of the proposed method is the smallest, that is, the estimation performance of the proposed method is better than that of the comparison method. It can be seen that the proposed method has better estimation performance when suppressing non-uniform noise and multipath signals.

[0114] Secondly, the performance of this method is compared with that of the four methods under different SNR conditions. The number of snapshots is fixed at 1500, the SNR range is -10dB:10dB, the step size is 2dB, and the other simulation conditions remain unchanged. Figure 3 The RMSE variation trend of each method with the change of signal-to-noise ratio is given. Under the same snapshot number condition, with the increase of signal-to-noise ratio, the downward trend of RMSE curves of all methods gradually slows down. Among them, the RMSE curve of the proposed method is significantly better than other methods, which shows that the proposed method has good noise suppression effect and decorrelation ability for coherent signals.

[0115] Those skilled in the art will appreciate that all or part of the steps in the above method can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk or an optical disk. Optionally, all or part of the steps in the above embodiment can also be implemented using one or more integrated circuits, and accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or in the form of software function modules. The present invention is not limited to any specific form of combination of hardware and software.

[0116] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A coherent signal decorrelation method based on noise suppression, characterized in that: Include: Establish a received signal model; Based on the array interpolation idea, the virtual sensors filling the physical holes in the coprime array are equivalent to the interpolated virtual uniform array, and then the interpolated virtual uniform array receiving signal model is constructed. Initializing the interpolated virtual uniform array receiving signal model and obtaining the sampling covariance matrix of the initialized signal model; Decomposing the above sampling covariance matrix and replacing the diagonal elements to obtain the pre-whitened covariance matrix to suppress non-uniform noise; The first row of the pre-whitened covariance matrix is ​​selected to reconstruct the Toeplitz matrix. The low-rank property of the reconstructed Toeplitz matrix is ​​used to perform sparse recovery on the Toeplitz matrix to obtain Pair Matrix Perform eigendecomposition and use the MUSIC algorithm to estimate the target DOA.

2. The coherent signal decorrelation method based on noise suppression according to claim 1, characterized in that: Assume that K far-field narrowband coherent sources come from different directions θ k (k=1,2,...,K) is incident on a coprime array, and the array element position is represented by d0. Then the received signal model expression on the coprime array is: Where s1(t) is the original reference signal, ρ is a non-zero complex-valued constant coefficient representing the coherence coefficient, is the array flow matrix, is the non-uniform noise matrix on the coprime array.

3. The coherent signal decorrelation method based on noise suppression according to claim 2, characterized in that: Filling the physical holes in the coprime array with virtual sensors is equivalent to interpolating a virtual uniform array and Interpolate the uth virtual uniform array i The received signal of each array element is expressed as: Where λ is the incident wavelength; the interpolation virtual uniform array receiving signal model is expressed as: Among them, the array flow matrix is the non-uniform noise matrix on the interpolated virtual uniform array.

4. The coherent signal decorrelation method based on noise suppression according to claim 3, characterized in that: Interpolation Virtual Uniform Array Received Signal Model Initialize it as: in, Indicates that the lth array element is in the set But not in the collection If it is not in the set The signal received by the array element in is considered to be 0.

5. The coherent signal decorrelation method based on noise suppression according to claim 4, characterized in that: The sampling covariance matrix of the initialized signal model is obtained: The sampling covariance matrix of the interpolated virtual uniform array received signal model is expressed as follows: Where T represents the number of snapshots, (·) H represents the conjugate transpose of a matrix; Then, the signal model after initialization The sampling covariance matrix of is expressed as: Where (p,q) represents the matrix The element at row p and column q in .

6. The coherent signal decorrelation method based on noise suppression according to claim 5, characterized in that: The sampling covariance matrix The breakdown is as follows: in in, is a matrix with all 1s except the diagonal which is 0.

7. The coherent signal decorrelation method based on noise suppression according to claim 6, characterized in that: Replacing the diagonal elements to obtain the pre-whitened covariance matrix contains: Replace all diagonal elements in R2 with the smallest diagonal element<diag(R2)> min , the replaced R2 is expressed as: The covariance matrix after pre-whitening is expressed as:

8. The coherent signal decorrelation method based on noise suppression according to claim 1, characterized in that: Using the low-rank property of the reconstructed Toeplitz matrix, the Toeplitz matrix is ​​sparsely restored to obtain Include: Construct the optimization problem: Where C is a binary matrix with the same dimension as the Toeplitz matrix, whose value is 1 at the non-zero position of the Toeplitz matrix and 0 at the other positions. is the matrix of the Hermitian-Toeplitz structure, rank(·) is the rank of the matrix, represents the Hadamard product; Since the rank minimization problem is an NP-hard problem, the above equation is convexly relaxed to the following form: The final recovery is an estimate of the Toeplitz matrix, which implements pre-whitening of non-uniform noise and decorrelation of coherent signals.

Citation Information

Cited By

  • Noise covariance matrix updating method, electronic equipment and storage medium

    CN119255289A