Space-time two-dimensional direction of arrival estimation method based on rank loss principle
Through the space-time two-dimensional wave direction method based on the rank loss principle, a conjugated augmented cross-correlation matrix and one-dimensional DOA estimator are constructed, which solves the source angle estimation problem in under-determined two-dimensional wave direction estimation, and realizes the High precision and low complexity estimation effect.
Patent Information
- Application Number
- CN202410992686.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
AI Technical Summary
The existing two-dimensional wave-drip direction estimation method is difficult to effectively solve the estimation problem of source angle in the case of underdetermined conditions, and most algorithms cannot handle underdetermined 2-D DOA estimation, and there are pairing errors and high complexity problems.
Using the space-time two-dimensional wave direction method based on the rank loss principle, the pitch angle and azimuth angle are estimated step by step by step by constructing a conjugated augmented cross-correlation matrix and a one-dimensional DOA estimator, and the complexity is reduced by using feature decomposition and decoupling techniques. .
High-precision parameter estimation in two-dimensional underdetermined DOA estimation is realized, which reduces complexity, is suitable for real-time estimation, and is superior to traditional methods in complexity.
Smart Images

Figure CN118938119A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of array signal processing, and in particular relates to a space-time two-dimensional direction of arrival estimation method based on the rank loss principle. Background Art
[0002] Two-dimensional direction of arrival (2-D DOA) estimation has been one of the important research directions in the field of array signal processing in the past decade, and is widely used in radar, sonar, communication, seismology and other fields. Compared with circular arrays and rectangular arrays, the L-shaped array consists of two uniform linear arrays, has good angle estimation performance, simple structure and easy implementation (Y.Hua, T.Sarkar, and D.Weiner, "An L-shaped array for estimating 2-D directions of wave arrival," IEEE Trans.Antennas Propag., vol.39, no.2, pp.143–146, Feb.1991.), so it is widely used in practice and is a commonly used two-dimensional direction finding array.
[0003] At present, researchers have proposed many methods for two-dimensional DOA estimation. For example:
[0004] (1) A CCM-ESPRIT algorithm that uses the cross-correlation matrix to construct a Toeplitz matrix to solve the pairing problem. However, this method still has the problem of pairing errors and is only effective for signals in certain incident directions (Kikuchi S, Tsuji H, Sano A. Pair-Matching Method for Estimating 2-D Angle of Arrival With a Cross-Correlation Matrix[J]. IEEE Antennas & Wireless Propagation Letters, 2006, 5:35-40.).
[0005] (2) A two-dimensional parameter joint estimation algorithm that does not require pairing has high estimation accuracy in certain situations, but its estimation performance is poor when the number of snapshots is small or the signal-to-noise ratio is low, and it does not fully utilize the cross-correlation matrix to eliminate the influence of additional noise (J. Liang and D. Liu, "Joint elevation and azimuth direction finding using L-shaped array," IEEE Trans. Antennas Propag., vol. 58, no. 6, pp. 2136–2141, Jun. 2010.).
[0006] (3) Based on the L array, a modified PM algorithm (MPM) is proposed for two-dimensional DOA estimation. Although this method reduces the amount of calculation, there is still a problem of pairing the pitch angle and azimuth angle (Tayem, Nizar, Kwon, et al. L-Shape2-Dimensional Arrival Angle Estimation With Propagator Method. [J]. IEEE Transactions on Antennas & Propagation, 2005.).
[0007] Among the commonly used high-precision estimation algorithms, the subspace-based two-dimensional MUSIC algorithm (M. Porozantzidou and M. Chryssomallis, “Azimuth and elevation an-glesestimationusing 2-DMUSIC algorithm with an L-shape antenna,” in Proc. IEEE APSURSI, 2010, pp. 1–4.) and the maximum likelihood algorithm (Chan A, Litva J. MUSIC and maximum likelihood techniques on two-dimensional DOA estimation with uniform circular array [J]. IEE Proceedings-Radar, Sonar and Navigation, 2002, 142 (3): 105-114.) can obtain better parameter estimates, but they require complex multi-dimensional spectral peak searches and cannot be used for real-time estimation.
[0008] Underdetermined DOA estimation refers to estimating the angle of arrival of a source when the number of signals received in the array is less than the number of sources. Moreover, most algorithms cannot handle the problem of underdetermined 2-D DOA estimation. For example, the extended uniform circular array is transformed into a one-dimensional uniform linear virtual array, and the related one-dimensional subspace algorithm is used to achieve underdetermined DOA estimation (Cao MY, Huang L, Qian C, et al. Underdetermined DOA estimation of quasi-stationary signals via Khatri-Rao structure for uniform circular array [J]. Signal Processing, 2015, 106 (jan.): 41-48.). However, this method cannot be directly adapted to the two-dimensional underdetermined DOA estimation algorithm. There is also a 2D-MUSIC algorithm based on spatial smoothing, which realizes the 2-D DOA estimation in the underdetermined case (Pal P, Vaidyanathan P P. Nested Arrays in Two Dimensions, Part I: Geometrical Considerations [J]. IEEE Transactions on Signal Processing, 2012, 60 (9): 4694-4705., Pal P, Vaidyanathan P P. Nested Arrays in Two Dimensions, Part II: Application in Two Dimensional Array Processing [J]. IEEE Transactions on Signal Processing, 2012, 60 (9): 4706-4718.), but the two-dimensional search required by this algorithm is usually difficult to implement. Summary of the invention
[0009] In view of the deficiencies in the prior art, an object of the present invention is to provide a space-time two-dimensional direction of arrival method based on the rank loss principle, which solves the two-dimensional underdetermined 2-D DOA estimation problem.
[0010] The present invention is achieved through the following technical solutions.
[0011] A space-time two-dimensional direction of arrival method based on the rank loss principle comprises the following steps:
[0012] S1, assume that K narrowband far-field signal sources each transmit a narrowband far-field signal to an L-type sensor array and perform Nx snapshot sampling, where the L-type sensor array is composed of an x-subarray and a y-subarray, and the received signal vectors of the x-subarray and the y-subarray at the nth snapshot sampling are x(n) and y(n), respectively, where n is the number of sampling points of the snapshot sampling, n=1, ..., Nx;
[0013] x(n)=A x (α)s(n)+w x (n)
[0014] y(n)=A y (β)s(n)+w y (n)
[0015] Among them, the narrowband far-field signals emitted by the K narrowband far-field sources are mutually uncorrelated; the additive white Gaussian noise vector w received by the x subarray x (n) satisfies the mean value of 0, and the additive Gaussian white noise vector w received by the y subarray y (n) satisfies the mean value of 0;
[0016] s(n)=[s1(n),s2(n),···,s K (n)] T represents the incident signal vector of K narrowband far-field sources at the nth snapshot sampling, where s k (n) represents the incident signal value of the k-th narrowband far-field source at the n-th snapshot sampling, k = 1, 2, ... K;
[0017] A x (α)=[a x (α1),a x (α2),···,a x (α K )] and A y (β)=[a y (β1),a y (β2),···,a y (β K )] represent the steering vector matrices of the x sub-matrix and the y sub-matrix respectively, a x (α k ) is the steering vector of the x subarray of the kth narrowband far-field source, a y (β k ) is the steering vector of the y subarray of the kth narrowband far-field source.
[0018] S2, construct the first DOA estimator to obtain the estimated value of the pitch angle of each narrowband far-field signal source, where the estimated value of the pitch angle of the kth narrowband far-field signal source is The first DOA estimator is as follows:
[0019]
[0020] Among them, U n is the noise subspace, For U n The conjugate transpose operation of ;
[0021] D H (β k ) is the value of D(β k )'s conjugate transpose operation;
[0022] Get D(β k ) method comprises steps 1 to 3:
[0023] Step 1: Construct the conjugate augmented cross-correlation matrix
[0024]
[0025] in,
[0026] in, A y (β) conjugation; A x (α) conjugation; represents the variance of additive Gaussian white noise, r SS is the signal delay autocorrelation matrix of K narrowband far-field sources; R ω0 is a (2F-1)×(2F-1) matrix, R ω0 The element in the 1st row and the Fth column of is 1, and the rest of the elements are 0, where F is the number of frames obtained by dividing x(n) or y(n) using the maximum overlap principle in the time domain;
[0027] Step 2: Conjugate augmented cross-correlation matrix Expands to the following formula:
[0028]
[0029] For a y (β k ), For a x (α k )'s conjugation;
[0030] Step 3, Each column in To decouple, After decoupling, we obtain k D(β k), where each D(β k ) The steps of decoupling are as follows:
[0031]
[0032] Among them, I Nx is the identity matrix of dimension Nx, where Nx is the number of samples in the snapshot sampling;
[0033] S3, the result obtained by S2 Substitution β k and as
[0034] S4, construct a one-dimensional second DOA estimator:
[0035] Will Substitute into the second DOA estimator to obtain the azimuth estimate of the kth narrowband far-field source in, For The conjugate transpose of .
[0036] In the above technical solution, the conjugate augmented cross-correlation matrix is calculated The covariance matrix of
[0037]
[0038] Where E{·} represents the expected operation function, For The conjugate transpose operation of .
[0039] In the above technical solution, the covariance matrix Perform eigendecomposition to obtain U n .
[0040] In the above technical solution, the covariance matrix Methods for performing eigendecomposition include:
[0041]
[0042] Among them, U s is the signal subspace, U n is the noise subspace, Σ s is the eigenvalue matrix of the signal subspace, Σ n is the eigenvalue matrix of the noise subspace.
[0043] The space-time two-dimensional direction of arrival estimation method of the present invention can be adapted to a two-dimensional underdetermined DOA estimation algorithm, has high parameter estimation accuracy, and has lower complexity compared with direct two-dimensional search. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 is obtained by the space-time two-dimensional direction of arrival method in Example 2 Comparison between estimated value and actual value;
[0045] Figure 2 This is a complexity comparison between the space-time two-dimensional direction of arrival method and the 2D-MUSIC method in Example 3. DETAILED DESCRIPTION
[0046] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.
[0047] Example 1
[0048] A space-time two-dimensional direction of arrival method based on the rank loss principle comprises the following steps:
[0049] S1, assume that K narrowband far-field signal sources each transmit narrowband far-field signals to the L-type sensor array and perform snapshot sampling, where the L-type sensor array consists of an x-subarray and a y-subarray and has a total of M array elements, and the x-subarray has M array elements on the x-axis. x The uniform linear array, the y subarray is the array element number M on the y axis y The total number of elements of the L-shaped sensor array is M = M x +M y -1; the received signal vectors of the x-array and y-array at the nth snapshot sampling are x(n) and y(n) respectively, and the dimensions of x(n) and y(n) are M respectively. x ×1 and M y ×1, n is the number of sampling points for snapshot sampling, n=1, ..., Nx.
[0050] x(n)=A x (α)s(n)+w x (n)
[0051] y(n)=A y (β)s(n)+w y (n)
[0052] Among them, the narrowband far-field signals emitted by K narrowband far-field sources are are mutually uncorrelated and have the same wavelength of λ, k=1,2,...K; the additive white Gaussian noise vector w received by the x subarray x (n) satisfies the mean value of 0 and is consistent with the narrowband far-field signal Uncorrelated, the additive white Gaussian noise vector w received by the y subarray y(n) satisfies the mean value of 0 and is consistent with the narrowband far-field signal Not relevant;
[0053] s(n)=[s1(n),s2(n),···,s K (n)] T represents the incident signal vector of K narrowband far-field sources at the nth snapshot sampling, where s k (n) represents the incident signal value of the k-th narrowband far-field source at the n-th snapshot sampling. x (α)=[a x (α1),a x (α2),···,a x (α K )] and A y (β)=[a y (β1),a y (β2),···,a y (β K )] represent the steering vector matrices of the x sub-matrix and the y sub-matrix respectively, a x (α k ) is the steering vector of the x subarray of the kth narrowband far-field source, a y (β k ) is the steering vector of the y subarray of the kth narrowband far-field source.
[0054] S2, construct a one-dimensional first DOA estimator to obtain the estimated value of the pitch angle of each narrowband far-field source, where the estimated value of the pitch angle of the kth narrowband far-field source is The first DOA estimator is as follows:
[0055]
[0056] Among them, U n is the noise subspace, U n 2M x M y ×(2M x M y -K)-dimensional matrix; For U n The conjugate transpose operation of Perform eigendecomposition to obtain U n .
[0057] D H (β k ) is the value of D(β k )'s conjugate transpose operation;
[0058] D(β k ) is of dimension 2M x My ×2 matrix, obtain D(β k ) methods include:
[0059] Step 1: Construct the conjugate augmented cross-correlation matrix (Dong Yangyang. Research on simultaneous multi-signal passive direction finding technology.
[0060] (Doctoral dissertation, Xidian University):
[0061]
[0062] in,
[0063] in, A y (β) conjugation; A x (α) conjugation; represents the variance of additive Gaussian white noise, R SS is the signal delay autocorrelation matrix of K narrowband far-field sources; R ω0 is a (2F-1)×(2F-1) matrix, R ω0 The element of the first row and the Fth column is 1, and the rest of the elements are 0, where F is the number of frames obtained by dividing x(n) or y(n) using the maximum overlap principle in the time domain; J (2F-1) As shown below:
[0064]
[0065] Step 2: According to the relationship between Khatri-Rao product and Kronecker product, the conjugate augmented cross-correlation matrix Expands to the following formula:
[0066]
[0067] a x (α k ) is the steering vector of the x subarray of the kth narrowband far-field source, a y (β k ) is the steering vector of the y subarray of the k-th narrowband far-field source, For a y (β k ), For a x (α k )'s conjugation;
[0068] Step 3: According to the properties of Kronecker product, Each column in To decouple, After decoupling, we obtain k D(β k ), where each D(β k ) The steps of decoupling are as follows:
[0069]
[0070] Among them, I Nx is the identity matrix of dimension Nx, where Nx is the number of samples in snapshot sampling.
[0071] S3, the result obtained by S2 Substitution β k and will As
[0072] S4, construct a one-dimensional second DOA estimator:
[0073] Will Substitute into the second DOA estimator to obtain the azimuth estimate of the kth narrowband far-field source in, For The conjugate transpose of .
[0074] In the above technical solution, the conjugate augmented cross-correlation matrix is calculated The covariance matrix of
[0075]
[0076] Where E{·} represents the expected operation function, For The conjugate transpose operation of .
[0077] In the above technical solution, the covariance matrix Perform eigendecomposition:
[0078]
[0079] Among them, U s is the signal subspace, U s 2M x M y A matrix of dimension ×K;
[0080] U n is the noise subspace, U n 2M x M y ×(2M x M y -K)-dimensional matrix;
[0081] Σ s is the eigenvalue matrix of the signal subspace, Σ s is a matrix of K×K dimensions;
[0082] Σ n is the eigenvalue matrix of the noise subspace, Σ n (2M x M y -K)×(2M x M y -K)-dimensional matrix.
[0083] Example 2
[0084] A space-time two-dimensional direction of arrival method based on the rank loss principle, based on Example 1, assuming that the power of all narrowband far-field sources is equal, the signal-to-noise ratio (SNR) is defined as in, represents the signal power of the narrowband far-field source. Assume that there are 6 uncorrelated narrowband far-field sources incident on a total of 5 array elements (i.e., M x 3 and M y The L-type sensor array of 3) is set so that the actual azimuth angle of the L-type sensor array and the K narrow-band far-field signal sources is α = [20°, 50°, 80°, 110°, 140°, 160°], and the actual elevation angle of the L-type sensor array and the K narrow-band far-field signal sources is β = [25°, 50°, 80°, 105°, 125°, 155°]. The array element spacing is set to 0.5λ (λ = 1), the signal-to-noise ratio SNR is set to 25dB, the number of snapshots is set to 500 (Nx), and the number of frames is set to 100 (F).
[0085] Actual azimuth and elevation estimates for six narrowband far-field sources like Figure 1 As shown, from Figure 1 It can be seen that in the underdetermined case, the space-time two-dimensional direction of arrival method based on the rank loss principle of the present invention can successfully identify 6 signals.
[0086] Example 3
[0087] The space-time two-dimensional direction of arrival method based on the rank loss principle is compared with the 2D-MUSIC method (M. Porozantzidou and M. Chryssomallis, "Azimuth and elevation angles estimation using 2-D MUSIC algorithm with an L-shape antenna," in Proc. IEEE APSURSI, 2010, pp. 1-4) in terms of complexity. The parameters are based on Example 2, with the total number of array elements, the number of snapshots, and the number of frames fixed at M = 5 and F = 100 respectively, the signal-to-noise ratio is set to 5 dB, and the number of snapshots Nx varies from 200 to 2000.
[0088] The complexity of the space-time two-dimensional direction of arrival method and the 2D-MUSIC method of the present invention is as follows: Figure 2 As shown, from Figure 2 It can be seen that the ordinate in the figure is complexity and the abscissa is the number of snapshots; Figure 2 This shows that the space-time two-dimensional direction of arrival method of the present invention is superior to the 2D-MUSIC algorithm in terms of complexity.
[0089] The calculation formula of the space-time two-dimensional direction of arrival method complexity of the present invention is:
[0090] The complexity calculation formula of the 2D-MUSIC method is:
[0091] Wherein, Δα and Δβ represent the search steps of the elevation angle and azimuth angle, respectively, both of which are 0.1°.
[0092] In summary, the space-time two-dimensional direction of arrival method of the present invention is less complex than direct two-dimensional search.
Claims
1. A space-time two-dimensional direction of arrival method based on the rank loss principle, characterized in that: The following steps are involved: S1, assume that K narrow-band far-field signal sources each transmit a narrow-band far-field signal to an L-type sensor array and perform Nx snapshot sampling, where the L-type sensor array consists of an x-subarray and a y-subarray. The received signal vectors of the x-subarray and the y-subarray at the nth snapshot sampling are x(n) and y(n), respectively, and n is the number of sampling points of the snapshot sampling. n=1, ..., Nx; x(n)=A x (a)s(n)+w x (n) y(n)=A y (β)s(n)+w y (n) Among them, the narrowband far-field signals emitted by the K narrowband far-field sources are mutually uncorrelated; the additive white Gaussian noise vector w received by the x subarray x (n) satisfies the mean value of 0, and the additive Gaussian white noise vector w received by the y subarray y (n) satisfies the mean value of 0; s(n)=[s1(n),s2(n),···,s K (n)] T represents the incident signal vector of K narrowband far-field sources at the nth snapshot sampling, where s k (n) represents the incident signal value of the k-th narrowband far-field source at the n-th snapshot sampling, k = 1, 2, ... K; A x (α)=[a x (α1),a x (α2),···,a x (α K )] and A y (β)=[a y (β1),a y (β2),···,a y (β K )] represent the steering vector matrices of the x sub-matrix and the y sub-matrix respectively, a x (α k ) is the steering vector of the x subarray of the kth narrowband far-field source, a y (β k ) is the steering vector of the y subarray of the k-th narrowband far-field source; S2, construct the first DOA estimator to obtain the estimated value of the pitch angle of each narrowband far-field signal source, where the estimated value of the pitch angle of the kth narrowband far-field signal source is The first DOA estimator is as follows: Among them, U n is the noise subspace, For U n The conjugate transpose operation of ; D H (β k ) is the value of D(β k )'s conjugate transpose operation; Get D(β k ) method comprises steps 1 to 3: Step 1: Construct the conjugate augmented cross-correlation matrix in, in, A y (β) conjugation; A x (α) conjugation; represents the variance of additive Gaussian white noise, R SS is the signal delay autocorrelation matrix of K narrowband far-field sources; R ω0 is a (2F-1)×(2F-1) matrix, R ω0 The element in the 1st row and the Fth column of is 1, and the rest of the elements are 0, where F is the number of frames obtained by dividing x(n) or y(n) using the maximum overlap principle in the time domain; Step 2: Conjugate augmented cross-correlation matrix Expands to the following formula: For a y (β k ), For a x (α k )'s conjugation; Step 3, Each column in To decouple, After decoupling, we obtain k D(β k ), where each D(β k ) The steps of decoupling are as follows: Among them, I Nx is the identity matrix of dimension Nx, where Nx is the number of samples in the snapshot sampling; S3, the result obtained in S2 Substitution β k and as S4, construct a one-dimensional second DOA estimator: Will Substitute into the second DOA estimator to obtain the azimuth estimate of the kth narrowband far-field source in, For The conjugate transpose of .
2. The space-time two-dimensional direction of arrival method according to claim 1, characterized in that: Compute the conjugate augmented cross-correlation matrix The covariance matrix of Where E{·} represents the expected operation function, For The conjugate transpose operation of .
3. The space-time two-dimensional direction of arrival method according to claim 1, characterized in that: Covariance matrix Perform eigendecomposition to obtain U n .
4. The space-time two-dimensional direction of arrival method according to claim 3, characterized in that: Covariance matrix Methods for performing eigendecomposition include: Among them, U s is the signal subspace, U n is the noise subspace, Σ s is the eigenvalue matrix of the signal subspace, Σ n is the eigenvalue matrix of the noise subspace.
Citation Information
Cited By
Distributed single-snapshot direction-of-arrival estimation method based on two-dimensional Hankel matrix completion
CN121027976A
