Direction of arrival estimation method and system for non-uniform diagonal load shedding under impulse noise

Abstract

The present invention belongs to the field of signal estimation technology, and particularly relates to a method and system for estimating the direction of arrival based on non-uniform diagonal load shedding MVDR under impulse noise, wherein the array element receiving signal is obtained according to the array signal model, and the array element receiving signal is preprocessed using a nonlinear function to suppress the impulse noise; the array element receiving signal covariance matrix is ​​obtained from the preprocessed receiving signal, and the matrix is ​​inverted to obtain its inverse matrix; the covariance matrix is ​​diagonally unloaded using the inverse matrix and a preset unloading factor; the MVDR algorithm is used to estimate the direction of arrival of the target signal for the covariance matrix after the non-uniform diagonal unloading. The present invention can effectively improve the sound source direction resolution performance, and further verify through simulation experimental data that the algorithm of this case has a stronger ability to resolve multiple sound sources, and has a good application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal estimation, and in particular relates to a method and system for estimating a direction of arrival of a signal with non-uniform diagonal load shedding under impulse noise. Background Art

[0002] As a branch of signal processing, array signal processing has been widely used and developed rapidly in the fields of communication, radar, sonar, seismic exploration and radio astronomy. Underwater acoustic array signal processing arranges a group of hydrophones in different positions in space in a certain way to form a hydrophone array. Using a hydrophone array to receive spatial signals is equivalent to sampling spatially distributed signals to obtain spatial discrete observation data of the signal source; Direction of Arrival (DOA) estimation technology is an important part of underwater array signal processing technology, which realizes the azimuth estimation of spatial sound source signals through array signal processing. The conventional beamforming (CBF) algorithm and the minimum variance distortionless response (MVDR) algorithm are still the most commonly used DOA estimation algorithms in actual sonar processing systems. The main reason is that these two algorithms are not complex, can process data in real time, and do not need to estimate the number of signal sources. The MVDR algorithm is not limited by the "Rayleigh limit" and has a higher angular resolution than the CBF algorithm, but its performance is limited in low signal-to-noise ratio environments.

[0003] At the same time, underwater noise is usually assumed to obey Gaussian distribution, while impulse noise exists in shallow water and obeys Alpha stable distribution. Impulse noise has no finite second-order moment and high-order moment, so the performance of traditional DOA estimation method is reduced or even fails. Summary of the invention

[0004] To this end, the present invention provides a DOA estimation method and system with non-uniform diagonal load shedding under pulse noise. When the number of snapshots is sufficient, the incoherent noise is mainly superimposed on the main diagonal of the covariance matrix, and the sound source azimuth resolution performance can be improved by non-uniform diagonal load shedding.

[0005] According to the design scheme provided by the present invention, a method for estimating the direction of arrival of non-uniform diagonal load shedding under impulse noise includes the following contents:

[0006] The array element receiving signal is obtained according to the array signal model, and the array element receiving signal is preprocessed using a nonlinear function to transform the signal amplitude into a preset range to suppress the pulse noise;

[0007] The preprocessed received signal is expressed in a matrix form, the covariance matrix of the array element received signal is obtained, and the matrix is ​​inverted to obtain its inverse matrix;

[0008] The covariance matrix is ​​non-uniformly diagonally unloaded using the inverse matrix and the preset unloading factor to change its eigenvalue distribution;

[0009] The MVDR algorithm is used to estimate the target signal arrival direction based on the covariance matrix after non-uniform diagonal unloading.

[0010] As the DOA estimation method of the present invention with non-uniform diagonal load shedding under impulse noise, further, the process of preprocessing the received signal using a nonlinear function is expressed as: Wherein, x(t) is the received signal, and Δ is the preset value.

[0011] As a method for estimating the direction of arrival of non-uniform diagonal load shedding under impulse noise of the present invention, further, the matrix form of the received signal after preprocessing is expressed as: X(t)=A(Θ)S(t)+N(t), wherein A(Θ) is a flow matrix related to the array formation and signal characteristics of the received signal, S(t) represents the source signal vector, and N(t) represents the additive noise vector.

[0012] As a method for estimating the direction of arrival of non-uniform diagonal load shedding under impulse noise of the present invention, further, the non-uniform diagonal load shedding process is expressed as: Where λ is the preset load reduction factor, λ>0, R x =E[XX H ] is the covariance matrix, For R x The inverse matrix of .

[0013] As a method for estimating the direction of arrival of non-uniform diagonal load shedding under impulse noise of the present invention, the spatial spectrum is obtained based on the MVDR algorithm, and the spectral peak position with the largest spatial spectrum power is found. The direction of the target signal is determined according to the spectral peak position. The spatial spectrum output power is expressed as:

[0014] Furthermore, the present invention also provides a direction of arrival estimation system with non-uniform diagonal load shedding under impulse noise, comprising: a signal preprocessing module, a signal covariance matrix representation module, a matrix eigenvalue adjustment module and a direction of arrival estimation module, wherein:

[0015] A signal preprocessing module is used to obtain the array element receiving signal according to the array signal model, and preprocess the array element receiving signal using a nonlinear function to transform the signal amplitude into a preset range;

[0016] A signal covariance matrix representation module is used to represent the preprocessed received signal in a matrix form, obtain the array element received signal covariance matrix, and invert the matrix to obtain its inverse matrix;

[0017] A matrix eigenvalue adjustment module, used to change the eigenvalue of the covariance matrix by performing non-uniform diagonal unloading on the inverse matrix using a preset unloading factor;

[0018] The direction of arrival estimation module is used to estimate the direction of arrival of the target signal using the MVDR algorithm based on the covariance matrix after non-uniform diagonal load shedding.

[0019] Beneficial effects of the present invention:

[0020] The present invention aims at estimating the direction of arrival of underwater sound source signals under impulse noise, combines nonlinear preprocessing, non-uniform diagonal load shedding and MVDR algorithm, uses nonlinear preprocessing to suppress the impulse noise in the signal to obtain the array receiving signal covariance matrix, and subtracts the weight of its inverse matrix from the covariance matrix to effectively remove the noise component on the main diagonal of the covariance matrix, and finally estimates the direction of the sound source through the maximum peak position of the MVDR spatial spectrum, thereby improving the sound source resolution performance. Compared with the CBF algorithm, MVDR algorithm and uniform diagonal load shedding MVDR algorithm, the algorithm proposed in this case has stronger multi-sound source resolution capability and has good application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flowchart of a method for estimating the direction of arrival of non-uniform diagonal load shedding under impulse noise in an embodiment;

[0022] Figure 2 This is a schematic diagram of a normalized azimuth spectrum when a single sound source is used in the embodiment;

[0023] Figure 3 It is a schematic diagram of a normalized azimuth spectrum when there are multiple sound sources in the embodiment;

[0024] Figure 4 It is a partial enlarged schematic diagram of the normalized azimuth spectrum when there are multiple sound sources in the embodiment;

[0025] Figure 5 Schematic diagram of the comparison of multiple sound source resolution performance at different azimuth intervals in the embodiment;

[0026] Figure 6 Schematic diagram of the comparison of multi-source resolution performance under different mixed signal-to-noise ratios in the embodiment;

[0027] Figure 7 The figure is a comparison diagram of the multi-sound source resolution performance under different numbers of array elements in the embodiments. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below in conjunction with the accompanying drawings and technical solutions.

[0029] Aiming at the problem of estimating the direction of arrival of underwater sound source signals under impulse noise, an embodiment of the present invention provides a method for estimating the direction of arrival of underwater sound source signals under impulse noise with non-uniform diagonal load shedding. Figure 1 As shown, it contains the following contents:

[0030] S101, acquiring an array element receiving signal according to an array signal model, and preprocessing the array element receiving signal using a nonlinear function to transform the signal amplitude into a preset range to suppress the influence of underwater pulse noise;

[0031] S102, expressing the preprocessed received signal in a matrix form, obtaining the covariance matrix of the array element received signal, and inverting the matrix to obtain its inverse matrix;

[0032] S103, using the inverse matrix and a preset load reduction factor to perform non-uniform diagonal load reduction on the covariance matrix to change its matrix eigenvalues;

[0033] S104, using the MVDR algorithm to estimate the target signal arrival direction for the covariance matrix after the non-uniform diagonal load shedding.

[0034] In the embodiment of this case, nonlinear preprocessing, non-uniform diagonal load shedding and MVDR algorithm are combined. Nonlinear preprocessing is used to suppress the impulse noise in the signal to obtain the array receiving signal covariance matrix. The covariance matrix is ​​subtracted from the weighted inverse matrix to remove the noise component on the main diagonal of the covariance matrix. Finally, the maximum peak position of the MVDR spatial spectrum is used to estimate the direction of the sound source, thereby improving the sound source direction resolution performance.

[0035] Assume that the kth array element receives all sound source signals in the space at the same time, and its output can be expressed as:

[0036]

[0037] In the formula, g k (Θ i ) is the kth array element for Θ i The complex gain of the incident signal from the direction, τ k (Θ i ) is Θ i The time delay caused by the incident signal from the direction reaching the kth array element, n k (t) is the additive noise on the kth array element. Assuming that the directional response of each array element to all directions is the same and there is no influence of other factors, then g k (Θ i )=1.

[0038] , the spatial response of the entire array can be obtained by sampling the output signals on the M array elements in parallel, and the output signals are arranged into a column vector:

[0039]

[0040] Express (2) in matrix form:

[0041] X(t)=A(θ)S(t)+N(t) (3)

[0042] Where, X(t)=[x1(t) x2(t) … x M (t)] Τ represents the array output signal vector, S(t)=[s1(t) s2(t) … s D (t)] Τ represents the source signal vector, N(t)=[n1(t) n2(t) … n M (t)] Τ represents the additive noise vector. and A(Θ)=[a(Θ1) a(Θ2) … a(Θ D )] Τ are the array direction vector and array flow matrix respectively.

[0043] When using the MVDR algorithm for DOA estimation, the spatial spectrum is obtained by scanning the spatial direction, R x =E[XX H ]:

[0044]

[0045] Then, the sound source is estimated based on the position of the maximum spectrum peak of P(Θ).

[0046]

[0047] The minimum variance distortion-free response (MVDR) beamforming algorithm breaks through the limitation of the "Rayleigh limit" and has high angular resolution. However, its performance is limited in noisy environments. At the same time, since pulse noise has no finite second-order moment, the array receiving signal covariance matrix cannot be accurately obtained under pulse noise. When the MVDR algorithm is used directly, the performance is severely degraded or even fails. In order to estimate the DOA of underwater sound sources under pulse noise, in the embodiment of this case, the pulse noise is first suppressed by nonlinear preprocessing, so that the covariance matrix of the array receiving data after nonlinear preprocessing can be directly calculated. Then, the inverse matrix of the covariance matrix is ​​used to appropriately unload the covariance matrix diagonally to reduce the background noise component on its main diagonal. Finally, the spectral peak position with the largest spatial spectrum power of the MVDR is found through spatial scanning to determine the direction of the sound source signal.

[0048] The impulse noise suppression method based on nonlinear function processing uses a nonlinear function to transform the received signal amplitude into a certain range, which can smooth the impulse noise superimposed in the signal to a certain extent, so that the signal has a limited second-order moment function. In this embodiment, the nonlinear preprocessing can select the simplest nonlinear processing:

[0049]

[0050] x(t) is the received signal, and Δ is a small quantity to ensure that the denominator is not zero. Pulse noise appears as a spike pulse in the time domain. Through the processing of nonlinear functions, the spike pulses superimposed in the time domain can be effectively smoothed, thereby achieving the purpose of suppressing pulse noise, so that subsequent processing can be performed based on the Gaussian signal model.

[0051] Assume that the array additive incoherent background noise satisfies the isotropic Gaussian distribution, that is, R x is the array output signal covariance matrix, and its specific expression is shown in (7).

[0052]

[0053] in, is the signal power, a s is the array response corresponding to the signal direction, is the noise power, I is the unit matrix, and the incoherent background noise component is mainly concentrated in R x on the main diagonal.

[0054] Perform eigendecomposition on the covariance matrix:

[0055]

[0056] Where σ m (m=1,2,...,M) is the eigenvalue, which can be arranged in descending order as σ1≥σ2≥…≥σ M , Γ=diag{σ1,σ2,…,σ M} is a diagonal matrix composed of eigenvalues. m (m=1,2,...,M) is the eigenvector corresponding to the eigenvalue, U=[u1,u2,…,u M ] is the matrix composed of eigenvectors.

[0057] Array receiving the inverse of the covariance matrix It can be expressed as:

[0058]

[0059] In the formula, σ min is the minimum eigenvalue, that is, σ min =σM .

[0060] In the MVDR algorithm, the optimal weight vector is Will Substituting in and omitting the denominator scalar yields:

[0061]

[0062] Therefore, the beam response corresponding to the weighted vector is:

[0063]

[0064] μ m =a H (Θ0)u m represents the static beam weight vector a H The correlation between (Θ0) and the mth eigenvector, is the beam response obtained by taking the mth eigenvector as the weighting vector, that is, the mth eigenbeam.

[0065] If the covariance matrix R x If there is no error in the estimation of , then the eigenvalues ​​corresponding to the white noise are all equal:

[0066] σ D+1 =σ D+2 =…=σ M (12)

[0067] At this time, the third term on the right side of equation (11) is 0, and the white noise component has no effect on the MVDR beamforming method.

[0068] When the beam direction is different from the direction of the sound source signal, the signal can be regarded as interference, and the characteristic beam corresponding to the white noise is all 0, then the beam response in the direction not coming from the signal is:

[0069]

[0070] That is, the beam response of MVDR in the non-signal direction is equal to the conventional beamforming response minus the weighted sum of the interference beam, and its weighting coefficient is:

[0071]

[0072] According to formula (13), the MVDR beam response in the non-signal direction decreases compared with the conventional beam response, forming a null. The depth of the null is related to the weighting coefficient a. m is positively correlated, that is The smaller the interference beam weighting coefficient a is, the m The larger the value, the deeper the null, and the smaller the MVDR beam response in the non-signal direction.

[0073] Diagonal load shedding improves the weighting coefficient of the interference beam direction in equation (14) by changing the eigenvalue of the covariance matrix, thereby enhancing the sound source azimuth resolution capability of the MVDR algorithm in a low signal-to-noise ratio environment. The non-uniform diagonal load shedding MVDR algorithm proposed in this chapter has a covariance matrix The expression is as shown in formula (15), where λ is the load reduction factor, λ>0.

[0074]

[0075] It can be seen from formula (15) that the covariance matrix after non-uniform diagonal unloading is Its eigenvalue becomes smaller, from the original σ m White becomes To ensure the non-negative definiteness of the covariance matrix after unloading:

[0076]

[0077] Available When λ = 0, it degenerates into the MVDR algorithm. In order to achieve a good spatial spectrum output effect in the actual engineering application, and to avoid the negative definiteness of the covariance matrix due to excessive unloading factor, the unloading factor can be selected as When the mixed signal-noise ratio is low, the covariance matrix contains more noise, and noise becomes the main factor restricting the sound source resolution performance of the MVDR algorithm. At this time, λ tends to The noise components on the main diagonal of the covariance matrix are removed as much as possible, and diagonal load reduction can effectively improve the performance of the algorithm. When the mixed signal-to-noise ratio is high, that is, the signal in the covariance matrix is ​​the main component, the sound source resolution ability of the traditional MVDR algorithm is sufficient, λ tends to 0, and the IDR-MVDR algorithm gradually degenerates into the traditional MVDR algorithm.

[0078] After non-uniform diagonal load shedding, the spatial spectrum output obtained by the MVDR algorithm is:

[0079]

[0080] In summary, the steps of the proposed algorithm are shown in Table 1.

[0081] Table 1 DOA estimation algorithm steps under impulse noise

[0082]

[0083]

[0084] Further, based on the above method, an embodiment of the present invention also provides a direction of arrival estimation system with non-uniform diagonal load shedding under impulse noise, comprising: a signal preprocessing module, a signal covariance matrix representation module, a matrix eigenvalue adjustment module and a direction of arrival estimation module, wherein:

[0085] A signal preprocessing module is used to obtain the array element receiving signal according to the array signal model, and preprocess the array element receiving signal using a nonlinear function to transform the signal amplitude into a preset range;

[0086] A signal covariance matrix representation module is used to represent the preprocessed received signal in a matrix form, obtain the array element received signal covariance matrix, and invert the matrix to obtain its inverse matrix;

[0087] A matrix eigenvalue adjustment module, used to change the eigenvalue of the covariance matrix by performing non-uniform diagonal unloading on the inverse matrix using a preset unloading factor;

[0088] The direction of arrival estimation module is used to estimate the direction of arrival of the target signal using the minimum variance distortion-free response algorithm based on the covariance matrix after non-uniform diagonal unloading.

[0089] In order to verify the effectiveness of this solution, the following is a further explanation based on simulation test data:

[0090] In the experiment, the CBF algorithm, MVDR algorithm, uniform diagonal load reduction MVDR algorithm and the algorithm of this case were compared based on uniform linear array to determine the azimuth resolution of single and multiple sound sources, as well as the probability of successful azimuth resolution under different conditions. When comparing the algorithms, nonlinear preprocessing was performed to suppress impulse noise.

[0091] 1. Algorithm effectiveness analysis

[0092] Experiment 1 investigates the performance of the algorithm for sound source azimuth estimation in the case of a single sound source. A sound source 2FSK signal is incident on the base array from the 0° direction. Its center frequency is 7.5kHz, the frequency interval is 200Hz, the code rate is 200Baud, the system sampling rate is 20kHz, and 2000 snapshots are observed. The array is an 8-element uniform linear array, and the ratio of the array element spacing to the signal wavelength is 0.5. The environmental noise is an impulse noise with a characteristic index of 1.5 and a mixed signal-to-noise ratio of -10dB. The normalized azimuth spectra obtained by the four algorithms are shown in the following figure. Figure 2 shown.

[0093] It can be seen that all four algorithms can correctly estimate the azimuth of the sound source, but it is obvious that the main lobe width of the CBF algorithm is the widest. When its spectral peak power drops to -1dB, the beam width is about 14.9°, and there are obvious side lobes. Its normalized azimuth spectrum noise background level is about -2.1dB. When the spectral peak power of the MVDR algorithm drops to -1dB, the beam width is about 10.8°, the spectral peak is sharper, the side lobes are fewer, and the noise background level is comparable to that of the CBF algorithm. When the spectral peak of the uniform diagonal load reduction MVDR algorithm and the algorithm of this case drops to -1dB, the beam width is comparable, about 8.2°, which is smaller than the MVDR algorithm and the CBF algorithm. At the same time, the noise background level of the uniform diagonal load reduction MVDR algorithm is about 0.6dB lower than that of the MVDR algorithm, while the noise background level of the algorithm of this case is about 0.8dB lower than that of the MVDR algorithm.

[0094] Experiment 2 considers the case of multiple sound sources. It is assumed that there are three sound sources with azimuth angles of -15°, 15° and 27° respectively. The array receives 2FSK signals with center frequencies of 7kHz, 7.5kHz and 8kHz respectively. The frequency interval is 200Hz and the symbol rate is 200Baud. Other parameters are the same as those in Experiment 1. Let the minimum eigenvalue of the covariance matrix after uniform load shedding and non-uniform load shedding be equal. Figure 3 The normalized azimuth spectra of the four algorithms are given. Figure 4 This is a partial enlargement for easy observation.

[0095] It can be clearly seen that the CBF algorithm and the MVDR algorithm can no longer distinguish the two sound source signals at 15° and 27°. The beam widths of the uniform diagonal load shedding MVDR algorithm and the algorithm in this case are smaller than those of the CBF and MVDR algorithms, and the normalized azimuth spectrum background noise level is lower, so the azimuth of the two sound source signals can be distinguished. It is precisely because both algorithms appropriately reduce the noise components on the covariance matrix through diagonal load shedding, making the output spatial spectrum sidelobes lower and improving the sound source resolution. At the same time, it can be seen from the figure that when the minimum eigenvalue of the algorithm in this case is equal to that of the uniform diagonal load shedding MVDR algorithm, the background noise level is lower and the signal direction resolution effect is better. This is because the algorithm in this case is non-uniform load shedding. While effectively removing the noise components on the diagonal of the covariance matrix, it retains the signal components as much as possible, and maximizes the depth of the non-signal upward zero sink, so the sound source resolution is better and the noise background level is lower.

[0096] 2. Algorithm performance analysis under different conditions

[0097] By designing a Monte Carlo simulation experiment, the multi-sound source azimuth resolution capabilities of the four algorithms are examined. Assuming that the true azimuth angles of the two sound source signals are θ1 and θ2, for a certain simulation experiment, if the azimuth estimation value obtained by the algorithm is and If equation (17) is satisfied, it is considered that the two sound source directions in this experiment have been successfully resolved.

[0098]

[0099] Let the number of correct identifications be L t , the number of Monte Carlo experiments is L, then the success rate of resolution is P t The expression of is shown in (18), and 500 Monte Carlo experiments are performed.

[0100]

[0101] Experiment 3 compares the algorithm resolution performance of two incoherent underwater sound source signals at different azimuth intervals. Assume a 16-element uniform linear array, the ratio of array spacing to signal wavelength is 0.5, there are two underwater sound sources, sending 2FSK signals with center frequencies of 7kHz and 7.5kHz, respectively, with a frequency interval of 200Hz and a symbol rate of 200Baud. The two incident signals have the same power, the system sampling rate is 20kHz, the number of snapshots is 2000, and the environmental noise is a pulse noise with a characteristic index of 1.5 and a mixed signal-to-noise ratio of -10dB. The azimuth interval of the two signals changes from 4° to 10° with an interval of 0.5°. Figure 5 The resolution success rate curves of the four algorithms are given.

[0102] As can be seen from the figure, when the signal azimuth interval is less than 4.5°, the four algorithms cannot correctly distinguish the signal direction. When the signal azimuth interval is greater than 8°, the MVDR algorithm, the uniform diagonal load shedding MVDR algorithm and the algorithm in this case all achieve a successful resolution of 100%, while the CBF algorithm needs to reach a 100% resolution success rate when the azimuth interval is greater than 8.5°. In the range of signal azimuth interval greater than 4.5° and less than 8°, it can be seen that among the four algorithms, the azimuth success resolution of the algorithm in this case is the highest. For example, when the azimuth interval is 6°, its resolution success rate reaches 76.5%, while the uniform diagonal load shedding MVDR algorithm, MVDR algorithm and CBF algorithm are 66%, 45.5% and 9% respectively. The success resolution of the algorithm in this case is significantly higher than that of the other three algorithms.

[0103] Experiment 4 compares the resolution performance of the algorithms under different mixed signal-to-noise ratios for two incoherent underwater sound source signals. The azimuth interval is set to 5°, the ambient noise is an impulse noise with a characteristic index of 1.5, and the mixed signal-to-noise ratio varies from -20dB to 10dB in 2dB intervals. Other simulation experiment parameters are the same as those in Experiment 3. The resolution success rate curves of the four algorithms under different mixed signal-to-noise ratios are shown in Figure 2. Figure 6 shown.

[0104] As can be seen from the figure, except for the CBF algorithm, the resolution success rates of the other three algorithms increase with the increase of the mixed signal-to-noise ratio. Since CBF is restricted by the "Rayleigh limit", its beam width is greater than the azimuth interval of the two sound sources under the current simulation conditions, so it is impossible to successfully distinguish the two signals, and the successful resolution is 0. When the mixed signal-to-noise ratio is equal to 6dB, the successful resolution of MVDR, uniform diagonal load reduction MVDR and the algorithm in this case reaches 100%, and when the mixed signal-to-noise ratio is lower than 6dB, the probability of successful resolution of the algorithm in this case is greater than that of the MVDR algorithm and the uniform diagonal load reduction MVDR algorithm, and it has better sound source direction resolution capability. Therefore, in a low mixed signal-to-noise ratio environment, the algorithm in this case has a better sound source direction resolution capability.

[0105] Experiment 5 compares and analyzes the multi-source resolution performance of the four algorithms under different numbers of array elements. The number of array elements varies from 10 to 30 with an interval of 2. The impulse noise mixed signal-to-noise ratio is -10dB, the azimuth interval is 5°, and other simulation experiment parameters are the same as those in Experiment 3. Figure 7 The successful resolution of the four algorithms with different array element numbers is demonstrated.

[0106] It can be seen that as the number of array elements increases, the multi-source resolution capabilities of the four algorithms are enhanced. The increase in the number of array elements makes the beam width narrower. Therefore, when the signal azimuth interval is constant, the narrower the beam, the easier it is to distinguish the sound sources in two different azimuths. At the same time, it can be seen that when the current signal azimuth interval is 5°, if the number of array elements is less than 14, the four algorithms cannot successfully distinguish the sound source signals at two different azimuths. When the number of array elements is greater than 14, the sound source resolution capability of the algorithm in this case is significantly stronger than that of the other three algorithms.

[0107] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0108] Based on the above method and / or system, an embodiment of the present invention also provides a server, including: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the above method.

[0109] Based on the above method and / or system, an embodiment of the present invention further provides a computer-readable medium having a computer program stored thereon, wherein the above method is implemented when the program is executed by a processor.

[0110] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limiting, and thus other examples of the exemplary embodiments may have different values.

[0111] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0112] 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 method for estimating direction of arrival with non-uniform diagonal load shedding under impulse noise, characterized in that: Contains the following: The array element receiving signal is obtained according to the array signal model, and the array element receiving signal is preprocessed using a nonlinear function to suppress the influence of pulse noise; The preprocessed received signal is expressed in a matrix form, the covariance matrix of the array element received signal is obtained, and the matrix is ​​inverted to obtain its inverse matrix; The covariance matrix is ​​diagonally unloaded using the preset unloading factor and the inverse matrix, where the non-uniform diagonal unloading process is expressed as: λ is the load reduction factor, λ>0, R x =E[XX Η ] is the covariance matrix of the received signal X, is the inverse matrix of the covariance matrix; For the covariance matrix after non-uniform diagonal unloading, according to the covariance matrix after diagonal unloading The MVDR algorithm is used to obtain the spatial spectrum output power, and the target signal arrival direction is estimated by the maximum peak position of the spatial spectrum graph. The spatial spectrum output power acquisition process is expressed as: a(Θ0) is the array direction vector of the incident signal in the Θ0 direction.

2. The method for estimating the direction of arrival with non-uniform diagonal load shedding under impulse noise according to claim 1, characterized in that: The process of preprocessing the received signal using a nonlinear function is expressed as: Wherein, x(t) is the received signal, and Δ is a preset small value.

3. The method for estimating the direction of arrival with non-uniform diagonal load shedding under impulse noise according to claim 1 or 2, characterized in that: The matrix form of the received signal after preprocessing is expressed as: X(t)=A(Θ)S(t)+N(t), where A(Θ) is a flow matrix related to the array formation and signal characteristics of the received signal, S(t) represents the source signal vector, N(t) represents the additive noise vector, and X(t) is the array output signal vector.

4. A DOA estimation system with non-uniform diagonal load shedding under impulse noise, characterized in that: It includes: signal preprocessing module, signal covariance matrix representation module, matrix eigenvalue adjustment module and wave direction estimation module, among which, A signal preprocessing module is used to obtain the array element receiving signal according to the array signal model, and preprocess the array element receiving signal using a nonlinear function to transform the signal amplitude into a preset range to suppress impulse noise; The signal covariance matrix representation module is used to represent the preprocessed received signal in matrix form, obtain the array element received signal covariance matrix, and invert the matrix to obtain its inverse matrix; The matrix eigenvalue adjustment module is used to perform non-uniform diagonal load shedding on the signal covariance matrix using the inverse matrix and the preset load shedding factor to change its eigenvalue distribution, wherein the non-uniform diagonal load shedding process is expressed as: λ is the load reduction factor, λ>0, R x =E[XX Η ] is the covariance matrix of the received signal X, is the inverse matrix of the covariance matrix; The direction of arrival estimation module is used to estimate the covariance matrix after non-uniform diagonal load shedding. The MVDR algorithm is used to obtain the spatial spectrum output power, and the target signal arrival direction is estimated by the maximum peak position of the spatial spectrum graph. The spatial spectrum output power acquisition process is expressed as: a(Θ0) is the array direction vector of the incident signal in the Θ0 direction.

Images