Robust low-moment DOA estimation method in impulse noise environment

By using local information entropy to calculate optimization parameters in a uniform linear array structure, a robust low-order moment covariance matrix is ​​generated, and DOA estimation is used to use the MUSIC algorithm to perform DOA estimation accuracy in impulse noise scenarios is solved, and the problem of low estimation accuracy in impulse noise scenarios is realized. High-precision DOA estimation and noise outlier suppression are suitable for various impulse noise scenarios.

CN119024260APending Publication Date: 2024-11-26HANGZHOU DIANZI UNIV
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410973575.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The prior art is difficult to accurately perform DOA estimation in non-Gaussian noise environments, especially in impulse noise scenarios, where traditional methods fail and require prior noise knowledge.

Method used

The signal is received by a uniform linear array structure, and the optimization parameters are calculated through local information entropy, and a robust low-order moment covariance matrix is ​​generated. The MUSIC algorithm is used to perform DOA estimation, and the parameters are adaptively tuned to suppress impulse noise outliers to achieve robustness improvement .

Benefits of technology

The accuracy of DOA estimation is improved in the impulse noise environment, and can effectively suppress noise outliers without prior noise knowledge. It is suitable for any impulse noise scenarios. It is combined with spatial smoothing technology to be suitable for sparse array models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119024260A_ABST
    Figure CN119024260A_ABST
Patent Text Reader

Abstract

The invention discloses a robust low-moment DOA estimation method in an impulse noise environment. Signal information is received through a uniform linear array, a local information entropy is calculated according to the received signal information, then a tuning parameter b is calculated according to the local information entropy, a corresponding RLOM covariance matrix is further constructed, and finally accurate estimation of DOA is obtained in combination with a MUSIC algorithm. The invention provides a selection strategy of a self-adaptive tuning parameter b, the advantages of the FLOS-like method and the correlation entropy are combined, the self-adaptive parameter determined by the correlation entropy is adopted to automatically suppress the pulse noise power, the noise within the range of 0 < alpha < = 2 is effectively suppressed, and the noise suppression efficiency is improved. Therefore, a traditional subspace-based method can be effectively applied without any prior noise knowledge. The RLOM algorithm not only uses self-adaptive tuning parameters based on local information entropy, but also is suitable for various types of impulse noise. Compared with a traditional algorithm based on FLOS, the algorithm has obvious superiority under the conditions of strong impulse noise and low GSNR.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the estimation of direction of arrival (DOA) of a signal under impulse noise, and is applicable to the technical fields of radar, sonar positioning and the like. Background Art

[0002] DOA estimation has become a key area in array signal processing and has been widely used in wireless communications, radar, sonar, seismic detection, sound source localization, and other fields. In the past few decades, researchers have paid extensive attention to high-resolution methods, including multiple signal classification (MUSIC), estimation of signal parameters through rotational invariance techniques (ESPRIT), weighted subspace fitting (WSF), minimum variance distortionless response (MVDR), maximum likelihood (ML), etc. However, the traditional assumption of additive white Gaussian noise (AWGN) on which these methods are based is often unrealistic in practical scenarios, and will cause the algorithm to fail when the noise does not obey a Gaussian distribution. Therefore, there is an increasing need to develop robust DOA estimation methods that can accurately handle non-Gaussian noise sources. Recently, studies have shown that the α-stable distribution is a suitable noise model to describe this type of noise, which can be represented by S(α, β, γ, θ), including a series of distributions such as Gaussian distribution, Cauchy distribution, and Levy distribution. When different characteristic exponent parameters are selected, it can be considered as a maximum potential distribution representing various impulse noises. On this basis, scholars proposed a series of DOA estimation methods under pulse noise, such as fractional low-order moments, robust covariant moments, and phase fractional low-order moments. Summary of the invention

[0003] In order to solve the problems existing in the prior art, the present invention provides a robust low-order moment DOA estimation method in an impulse noise environment.

[0004] The present invention comprises the following steps:

[0005] S1: Receive signals through an array antenna with a uniform linear array structure to obtain received signal information X;

[0006] S2: Calculate the local information entropy H according to the received signal information using a calculation formula for the local information entropy;

[0007] S3: Substituting the obtained local information entropy into the optimization parameter formula to calculate the optimization parameter b;

[0008] S4: Calculate the robust low-order moment covariance matrix R according to the received signal information RLOM ;

[0009] S5: Use the MUSIC estimation algorithm to obtain R RLOM Accurate estimation of DOA: RLOMCharacteristic decomposition, obtain the noise subspace and construct the spatial spectrum function to find R through MUSIC spectrum peak search RLOM Peak value, thus obtaining R RLOM Accurate estimation of DOA.

[0010] The design principle of the present invention is as follows:

[0011] 1. The uniform linear array receives the signal from the source and performs algorithmic processing on the received signal, which improves the accuracy of DOA estimation in pulse environments. In addition, a selection strategy for the adaptive tuning parameter b is proposed, which can effectively suppress the outliers of pulse noise, so that the traditional subspace-based method can be effectively applied. The advantage of this strategy is that it does not require any prior knowledge of noise, but is directly related to the data received by the array.

[0012] 2. For the method of the present invention, local information entropy is used to calculate the optimization parameters, and then the RFOM covariance matrix is ​​obtained. The generated covariance matrix is ​​subjected to eigenvalue decomposition, and then the classic MUSIC algorithm is used to perform accurate DOA estimation.

[0013] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0014] ① The uniform linear array receives the signal sent by the source. According to the robustness of the algorithm used, the accuracy of DOA estimation is improved, and it can be well combined with spatial smoothing technology to be applicable to sparse array models.

[0015] ② Compared with other DOA estimation algorithms in impulse noise environments, the use of adaptive tuning parameters can effectively suppress outliers of impulse noise without any prior knowledge and is suitable for any impulse noise scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic diagram of the uniform linear array structure involved in the present invention;

[0017] Figure 2 It is a schematic diagram of the performance of tuning parameters of the present invention;

[0018] Figure 3 is a performance comparison diagram of the method of the present invention and other algorithms under different noise coefficient conditions;

[0019] Figure 4 is a performance comparison chart of the method of the present invention and other algorithms under different generalized noise ratios and snapshot numbers;

[0020] Figure 5 is a performance comparison diagram of the method of the present invention with other corresponding algorithms in high-intensity impulse noise; DETAILED DESCRIPTION

[0021] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding 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 improper limitations on the present invention.

[0022] This embodiment provides a robust low-order moment DOA estimation method in an impulse noise environment based on the following principles. The specific method of this embodiment is:

[0023] 1. Uniform Linear Array Noise Model and Data Model

[0024] Noise model:

[0025] Traditional research methods for DOA estimation mostly use the second-order statistics of the Gaussian noise model. However, in the study of radar echoes, low-frequency atmospheric noise, and underwater acoustic signals, the noise is often composed of irregular pulses or noise spikes with short duration and large amplitude, that is, impulse noise. In recent years, SαS can well describe this type of random signal source with significant spike pulses. Therefore, the SαS noise model has better applicability, which can be expressed by the characteristic function φ(t):

[0026]

[0027] Among them, t is the variable of the characteristic function, α is the characteristic index, and its size can affect the pulse degree of this distribution, ranging from 0<α≤2. When α=2, the distribution is Gaussian distribution; j is an imaginary number, γ is the dispersion parameter, and its meaning is consistent with the variance of Gaussian distribution; β is the skewness parameter, δ is the location parameter, and the distribution when β=δ=0 is symmetric α-stable (SαS) distribution; SαS represents different probability density functions when the parameters are selected differently. The smaller α is in the SαS distribution, the more obvious the pulse is; conversely, the larger α is, the closer the noise is to Gaussian noise. The important difference between Gaussian distribution and SαS distribution is that Gaussian distribution has second-order statistics, while the second-order statistics of SαS distribution do not exist when 0<α<2, so the second-order covariance matrix is ​​inaccurate.

[0028] Data Model:

[0029] like Figure 1 The array antenna structure shown is composed of M array elements, with the first sensor being a reference element, the array element spacing k=λ2 between the array elements, and the wavelength being λ.

[0030] Assuming that K DOAs are narrowband incoherent signals with θk, k = 1, 2, ..., K incident on a uniform linear array (ULA) with M array elements, the complex envelope received by the mth sensor at the tth snapshot is:

[0031]

[0032] sk(t) represents the kth signal source, τmk=2π(m-1)sinθkd / λ is the phase delay of the kth signal propagation quantity received by the mth sensor relative to the reference element.

[0033] The received data model can be expressed as:

[0034] x(t)=Α(θ)s(t)+n(t) (5)

[0035] Α(θ)=[a(θ1),…,a(θ K )] is the direction matrix, where is the direction vector, s(t)=[s1(t),s2(t),…,s K (t)] T is the signal vector, is the noise vector.

[0036] 2. Angle Estimation Method

[0037] In this embodiment, the above noise model and data model are applied to the DOA estimation method of the present invention, that is, the robust low-order moment MUSICDOA estimation method, which specifically includes the following steps:

[0038] Step 1: Calculate the local information entropy H.

[0039] According to the data model, the received signal measurement information can be obtained and the local information entropy H can be calculated:

[0040] The local information entropy H can be expressed as:

[0041]

[0042] Where M represents the number of array elements, T represents the number of snapshots, and g m,t is the ratio of the data received for a single array element to the entire array information, that is:

[0043]

[0044] Where T is the number of snapshots, xm(t) is the data received by the mth array element, x(t) is the information of the entire array, and ||x(t)||1 is the 1-norm of x(t). The local information entropy H is calculated by these two formulas.

[0045] Step 2: Calculate the optimization parameter b.

[0046] Substitute the local information entropy H obtained in step 1 into the formula:

[0047] b=1 / (5+e -H ) (8)

[0048] For a complex number we have:

[0049]

[0050] Where b = 1 / (5 + e -H ) is the optimization parameter, z * is the conjugate of the complex number z, and Here, based on the local information entropy H obtained previously, the optimization parameter b can be obtained.

[0051] Step 3: Calculate the robust low-order moment covariance matrix R according to the received signal information and the obtained optimization parameters. RLOM ;

[0052] For the signal model of the mth array element RLOM matrix R RLOM The (i,j)th element of is:

[0053]

[0054] Among them, x i (t), x j (t) are the i-th and j-th rows of the signal model x(t) at time t, respectively. The optimization parameter b represents the order of the moment. According to:

[0055]

[0056] x(t)=Α(θ)s(t)+n(t) (12)

[0057] Matrix R RLOM Represented as a specific expression:

[0058] R RLOM =AΦ s A+κI M (13)

[0059] Φ s is a diagonal matrix, κ is a constant, A is the steering matrix defined in the above formula, I M is the M×M identity matrix, R RLOM is an M×M matrix, whose (i,j)th item is:

[0060]

[0061] Among them, δ i,j is the Kronecker function, is the direction vector, a * (θ k ) represents a(θ k ) is the conjugate of kkrepresents a diagonal matrix.

[0062] According to the above equations (10) and (14), we can calculate

[0063] Step 4: Use the MUSIC estimation algorithm to obtain R RLOM Accurate estimation of DOA: RLOM Perform eigendecomposition to obtain the noise subspace:

[0064]

[0065] Among them, [u k+1 ,u k+2 ,...,u M ] corresponds to the eigenvector of the noise subspace, K is the number of signal sources, and M is the number of array elements.

[0066] And construct the spatial spectrum function:

[0067]

[0068] Among them, U is the noise subspace, U H is the conjugate transpose of U, a(θ) is the direction vector, a H (θ) is the conjugate transpose of a(θ).

[0069] The peak value is found through MUSIC spectrum peak search, thus obtaining an accurate estimate of DOA.

[0070] 3. Performance Analysis and Experimental Analysis

[0071] 1. Complexity analysis

[0072] We use the number of complex multiplications of real (or complex) numbers as the criterion for computational complexity. The complexity of computing the correlation entropy H is O{2MT}, and the complexity of computing the RLOM matrix is ​​O{3M 2 T}, for R RLOM The complexity of eigendecomposition is O{M 3}, the total computational complexity of the MUSIC algorithm is O{M 2 (MK)+2M 2 n s}, where n s is the number of searches. Therefore, the total computational complexity of the proposed RLOM algorithm is:

[0073] O{M 3 +3M 2 T+M 2 (MK)+2M 2 n s +2MT}

[0074] 2. Experimental analysis

[0075] In order to verify the effect of the above method, multiple simulation experiments were conducted in this embodiment, and the experimental performance was analyzed, as follows:

[0076] (1) Experimental performance evaluation indicators

[0077] In an impulse noise environment, the generalized signal-to-noise ratio is defined as:

[0078]

[0079] The performance estimation criterion is the joint root mean square error (RMSE) defined as:

[0080]

[0081] in, is the accurate estimate of the DOA of the kth source in the jth Monte Carlo process, K represents the number of sources, and MC represents the number of Monte Carlo trials.

[0082] (2) Experimental effect diagram

[0083] Figure 2 This is an analysis chart of the optimal selection range of the tuning parameter b in the method of the present invention. Through 500 MC tests to test the performance of the proposed RLOM algorithm in three different SαS noise scenarios, it can be seen that for the lowest RMSE, the value of b is between 0.15 and 0.20, so we let b = 1 / (5+e -H ) to satisfy the value of b between 0.16 and 0.20.

[0084] Figure 3 This is the RMSE graph of the compared algorithms under the scenario of 0<α<2, GSNR=0dB, snapshot T=300, MC=500. Figure 3 It can be seen from the middle left figure that the RMSE of all methods gradually decreases with the increase of α, but the phase fractional low-order moment still has a large error compared with other methods, and the performance of the algorithm of the method of the present invention is the best. When 0<α<1, the generalized autocorrelation entropy method fails because it cannot guarantee the boundedness of the covariance matrix. Therefore, the symbol covariance matrix, phase fractional low-order moment, correlation entropy infinite norm and the method of the present invention are considered for comparison. Figure 3 The results shown in the middle right figure clearly show that the performance of the method of the present invention is better than the symbol covariance matrix, phase fractional low-order moment and correlation entropy infinite norm method. In addition, the method does not need to deliberately determine the tuning parameter b, because the local correlation entropy adaptively determines its value.

[0085] Figure 4In the case of α = 1.2, snapshot T = 300, GSNR = -2dB. Figure 4 As can be seen from the middle left figure, the performance of the phase fractional low-order moment algorithm is not ideal, while the root mean square error of the other five algorithms decreases rapidly with the increase of GSNR. When GSNR = -2dB, we Figure 4 The middle right figure tests the dependence of these algorithms on the number of snapshots, and it can be seen that the method of the present invention has the best performance.

[0086] Figure 5 is the simulation result under high intensity impulse noise scenario (α = 0.4), and the other parameters are the same as Figure 4 Since the generalized autocorrelation entropy method fails in the range of 0<α<1, only the performance of the symbol covariance matrix, phase fractional low-order moment, correlation entropy infinite norm and the method of the present invention is compared. Figure 5 It can be seen that the RLOM algorithm proposed in this paper performs best in the high-intensity impulse noise environment, followed by the phase fractional low-order moment algorithm, then the correlation entropy infinite norm algorithm, and finally the symbol covariance matrix algorithm, which further highlights the superiority of the method of the present invention in the high-intensity impulse noise scenario.

[0087] In summary, the performance of the present invention is better than other methods in both high impulse noise scenarios and medium intensity impulse noise scenarios. It is also better than other algorithms in low signal-to-noise ratio and low snapshot number scenarios.

Claims

1. A robust low-order moment DOA estimation method in an impulse noise environment, characterized by: The steps include: S1: Receive signals through the array antenna of the uniform linear array to obtain the received signal information X; S2: According to the received signal information X, calculate the local information entropy H using the calculation formula of the local information entropy; S3: Substituting the obtained local information entropy H into the optimization parameter formula to calculate the optimization parameter b; S4: Calculate the robust low-order moment covariance matrix R according to the received signal information X RLOM ; S5: Use the MUSIC estimation algorithm to obtain R RLOM Accurate estimation of DOA: RLOM Characteristic decomposition, obtain the noise subspace and construct the spatial spectrum function to find R through MUSIC spectrum peak search RLOM Peak value, thus obtaining R RLOM Accurate estimation of DOA.

2. The robust low-order moment DOA estimation method in an impulse noise environment according to claim 1, characterized in that: In step S1, K incoherent narrowband sources are incident on a uniform linear array with M antenna elements, the first sensor is a reference element, the spacing between array elements is λ / 2, and the carrier wavelength is λ.

3. The robust low-order moment DOA estimation method in an impulse noise environment according to claim 1, characterized in that: The local information entropy H in step S2 is expressed as: Among them, g m,t is the ratio of the data received for a single array element to the entire array information, that is: Where T is the number of snapshots, x m (t) is the data received by the mth array element, x(t) is the information of the entire array; ||x(t)||1 is the 1-norm of x(t); the local information entropy H is calculated by these two formulas.

4. The robust low-order moment DOA estimation method in an impulse noise environment according to claim 1, characterized in that: The optimization parameter formula in step S3 is used to calculate the optimization parameter b, which is specifically: Where b = 1 / (5 + e -H ) is the optimization parameter, z * is the conjugate of the complex number z, and Here, based on the local information entropy H obtained previously, the optimization parameter b can be obtained.

5. The robust low-order moment DOA estimation method in an impulse noise environment according to claim 1, characterized in that: Generate a robust low-order moment covariance matrix as follows: For the signal model, assume that the kth DOA is θ k , k=1,2,…,K, then the complex envelope received by the mth sensor at the tth snapshot is: Among them, s k (t) represents the kth signal source at time t, is the phase delay of the kth signal propagation quantity received by the mth sensor relative to the reference element; RLOM matrix R RLOM The (i,j)th element of is: Among them, x i (t), x j (t) are the i-th and j-th rows of the signal model x(t) at time t, E represents the mathematical expectation, and b represents the order; in addition, For any i and j, it is bounded; the generated RLOM covariance matrix is ​​eigen-decomposed and the azimuth is estimated using the MUSIC estimation algorithm.

Citation Information

Cited By

  • Absolute polynomial correlation entropy DOA estimation method in impulse noise environment

    CN121477119A

  • A method for estimating DOA of absolute polynomial correlation entropy in pulse noise environment

    CN121477119B