A method, system and terminal for estimating direction of arrival based on improved relative entropy

Abstract

The present invention belongs to the technical field of underwater acoustic array signal processing, and discloses a method, system and terminal for estimating the direction of arrival based on improved relative entropy. The method for estimating the direction of arrival includes: receiving a sound source signal; extracting the spectrum of the sound source signal, and obtaining the covariance matrix after the normalization of the frequency of the expected signal in the spectrum; evenly dividing K scanning grids in a preset space, and obtaining the weight vector of each scanning grid based on the frequency of the expected signal and array parameters; performing eigendecomposition on the matrix obtained by multiplying the inverse of the covariance matrix with the normalized weight matrix to obtain the maximum eigenvalue; configuring an improved relative entropy distance expression; substituting the maximum eigenvalue into the above expression to obtain an improved symmetric relative entropy distance; obtaining the normalized output power of each scanning grid based on the improved symmetric relative entropy distance; performing a one-dimensional spectrum peak search, and the scanning grid angle corresponding to the peak value is the direction of arrival of the sound source signal. The present invention has higher azimuth resolution and lower sidelobe level.

Description

Technical Field

[0001] The present invention relates to the technical field of underwater acoustic array signal processing, and in particular to a method, system and terminal for estimating a direction of arrival based on improved relative entropy. Background Art

[0002] Direction of Arrival (DOA) estimation is an important technology in signal processing. Its main goal is to determine the direction of the signal source by receiving the signal. It is widely used in radar, sonar, radio navigation, earthquake monitoring and wireless communication systems.

[0003] The traditional DOA estimation algorithm scans within the observation range by adjusting the value of θ in the weight vector, then calculates the correlation between the weight vector and the received signal, and finally outputs the spatial spectrum. The direction with the highest correlation corresponds to the highest output power, which is the signal direction.

[0004] The resolution of traditional beamforming technology is often limited by the aperture and signal-to-noise ratio of the array antenna. To overcome these limitations, existing technologies have proposed some high-resolution DOA estimation methods. These methods are mainly based on the statistical characteristics of the signal for azimuth estimation, and can provide higher angular resolution than traditional beamforming technology under conditions of smaller antenna arrays and lower signal-to-noise ratios. However, these methods do not deviate from the research framework of Euclidean geometry, and the calculation of the similarity between the received signal and the weight vector is limited to the Euclidean distance, which is not conducive to coping with complex environments and is prone to high sidelobes. Summary of the invention

[0005] The object of the present invention is to provide a wave direction estimation method, system and terminal based on improved relative entropy, so as to overcome the defects of the existing wave direction estimation method that the calculation of the similarity between the received signal and the weight vector is limited to the Euclidean distance, which is not conducive to coping with complex environments and is prone to higher side lobes.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for estimating a direction of arrival based on improved relative entropy, comprising the following steps:

[0008] S1. Use an array consisting of N array elements to receive the sound source signal, N ≥ 2;

[0009] S2. Extract the spectrum of the sound source signal, obtain the frequency of the desired signal included in the spectrum, calculate the covariance matrix of the frequency, normalize the covariance matrix, and obtain the normalized covariance matrix K;

[0010] S3. K scanning grids are evenly divided within the preset spatial range, and the orientation of the kth scanning grid is represented by θk , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements;

[0011] S4. The weight vector w(θ k ) conjugate multiply to obtain the weight matrix of each scanning grid, and normalize the weight matrix to obtain the normalized weight matrix W(θ k );

[0012] S5. The inverse of the covariance matrix K and the normalized weight matrix W(θ k ) multiplied by eigenvalues ​​to obtain the maximum eigenvalue λ max ;

[0013] S6. Configure the improved relative entropy distance expression;

[0014] S7. Set the maximum eigenvalue λ max Substitute the improved relative entropy distance expression to obtain the covariance matrix K and each weight matrix W(θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k ));

[0015] S8. Based on the improved symmetric relative entropy distance D SKL (K,W(θ k )) obtain the normalized output power of each scanning grid;

[0016] S9. Perform a one-dimensional spectrum peak search on the normalized output power. The scanning grid angle corresponding to the peak value is the arrival direction of the sound source signal.

[0017] As a possible implementation method, the normalized covariance matrix K and the normalized weight matrix W(θ k ) are all symmetric positive definite matrices.

[0018] As a possible implementation method, the weight vector w(θ k ), specifically:

[0019]

[0020] Where N is the number of array elements, d is the array element spacing, λ is the signal wavelength, j is an imaginary number, and j 2 =-1.

[0021] As a possible implementation method, configuring an improved relative entropy distance expression includes the following steps:

[0022] S60. Define two symmetric positive definite matrices P and Q. The symmetric relative entropy distance between P and Q is expressed as:

[0023]

[0024] Where I is the identity matrix; tr means finding the trace of the matrix;

[0025] S61. Perform eigendecomposition on the matrix obtained by multiplying the inverse of the matrix P by the matrix Q:

[0026] (P -1 Q)=VΛV -1 ,

[0027] Where Λ=diag{λ 1 ,λ 2 ,...,λ n},λ i >0,i=1,2,...,n,Λ is a diagonal matrix,λ i represents the eigenvalue, V is the eigenvector matrix, and further we get:

[0028]

[0029] S62. Retain the maximum eigenvalue and obtain the improved relative entropy distance expression:

[0030]

[0031] As a possible implementation method, S7 is specifically:

[0032]

[0033] As a possible implementation, S8 includes:

[0034] S80. Solve for the improved relative entropy distance D SKL (K,W(θ k ))'s reciprocal, get S SKL (θ k ), determine S SKL (θ k ) in the maximum value max(S SKL (θ k ));

[0035] S81. Calculate the normalized output power of each scanning grid:

[0036] P(θ k )=10lg(S SKL (θ k ) / max(S SKL (θ k ))).

[0037] As a possible implementation method, the preset spatial range is [-90°, 90°], and K scanning grids are evenly divided in the spatial range of [-90°, 90°], specifically:

[0038] In the spatial range of [-90°, 90°], K scanning grids are evenly divided at intervals of preset angles, and the preset angles are any angles between 0.01° and 10°.

[0039] As a possible implementation manner, the array is a horizontal, inclined or vertical uniform linear array, and / or the array element spacing is half of the wavelength of the narrowband signal.

[0040] In a second aspect, the present invention provides a direction of arrival estimation system based on improved relative entropy, comprising:

[0041] The sound source signal acquisition unit uses an array composed of N array elements to receive the sound source signal, where N ≥ 2;

[0042] The frequency covariance matrix solving unit of the normalized desired signal extracts the spectrum of the sound source signal, obtains the frequency of the desired signal included in the spectrum, calculates the covariance matrix of the frequency, normalizes the covariance matrix, and obtains the normalized covariance matrix K;

[0043] The weight matrix solution unit after normalization of each scanning grid is evenly divided into K scanning grids in the spatial range of [-90°, 90°], and the orientation of the kth scanning grid is represented by θ k , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements; the weight vector w(θ k ) conjugate multiply to obtain the weight matrix of each scanning grid, and normalize the weight matrix to obtain the normalized weight matrix W(θ k );

[0044] The normalized output power solution unit of each scanning grid is the inverse of the covariance matrix K and the weight matrix W(θ k ) multiplied by eigenvalues ​​to obtain the maximum eigenvalue λ max , define the improved relative entropy distance expression, and change the maximum eigenvalue λ max Substitute the improved relative entropy distance expression to obtain the covariance matrix K and each weight matrix W(θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k )), based on the improved symmetric relative entropy distance D SKL (K,W(θk )) obtain the normalized output power of each scanning grid;

[0045] The direction of arrival estimation unit of the sound source signal performs a one-dimensional spectrum peak search on the normalized output power, and the scanning grid angle corresponding to the peak value is the direction of arrival of the sound source signal.

[0046] In a third aspect, the present invention provides a terminal, comprising a processor and a communication interface coupled to the processor, wherein the processor is used to run a computer program or instruction to implement the method for estimating a direction of arrival based on improved relative entropy provided in the first aspect.

[0047] Compared with the prior art, the present invention has the following technical effects:

[0048] The method for estimating the direction of arrival based on improved relative entropy provided by the present invention deduces, from the perspective of non-Euclidean geometry principles, that DOA estimation can be described by comparing the similarities between matrices. Thus, a weight matrix is ​​defined. Considering that the weight matrix and the data covariance matrix are both symmetric positive definite matrices, according to the principle of Riemannian geometry, the symmetric positive definite matrix is ​​abstracted into points distributed on a Riemann manifold, and then similarities between matrices can be compared on the manifold. Based on this, a measure of matrix similarity based on improved relative entropy is proposed. Simulation and experimental results show that the present invention has the following advantages: (1) The side lobes of the present invention are lower, the main lobe width is narrower, and the spatial spectrum performance is better; (2) The present invention has higher azimuth resolution for two spatially adjacent signals; (3) The present invention can be applied to complex environments, and the resolution is not affected by the signal distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings described herein are used to provide further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0050] Figure 1 A flow chart of a method for estimating direction of arrival using improved relative entropy provided in an embodiment of the present invention;

[0051] Figure 2 The relationship between the maximum sidelobe level and the number of snapshots in the spatial spectrum of the CBF processor, the MVDR processor, the MCE processor and the improved relative entropy SKL processor proposed in the present invention when the signal-to-noise ratio is fixed at 5dB in the simulation of a single signal in the embodiment of the present invention;

[0052] Figure 3 In the case of simulating a single signal in an embodiment of the present invention, when the signal-to-noise ratio is fixed at 5 dB, the relationship between the main lobe width and the number of snapshots in the spatial spectrum of the CBF processor, the MVDR processor, the MCE processor and the improved relative entropy SKL processor proposed by the present invention;

[0053] Figure 4 To simulate a single signal, the mean square error performance in the range of -20dB to 10dB for a fixed number of snapshots at 12;

[0054] Figure 5 To simulate a single signal, the mean square error performance in the range of -20dB to 10dB for a fixed number of snapshots at 40;

[0055] Figure 6 To simulate the relationship between the maximum sidelobe level and the number of snapshots in the spatial spectrum of the CBF processor, the MVDR processor, the MCE processor and the SKL processor proposed by the present invention when the signal-to-noise ratio is fixed at 5dB under four signal conditions;

[0056] Figure 7 To simulate the relationship between the main lobe width and the number of snapshots in the spatial spectrum of the CBF processor, MVDR processor, MCE processor and the SKL processor proposed by the present invention when the signal-to-noise ratio is fixed at 5dB under four signal conditions;

[0057] Figure 8 To simulate the spatial spectra of the CBF processor, MVDR processor, MCE processor, and SKL processor, the signal-to-noise ratio is fixed at 5 dB for the four signal cases;

[0058] Fig. 9 To simulate the four signal case, when the number of snapshots is fixed at 40, the mean square error (MSE) performance in the signal-to-noise ratio range of -20dB to 10dB. DETAILED DESCRIPTION

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

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

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

[0062] The DOA estimation methods provided by the existing technologies mainly estimate the azimuth based on the statistical characteristics of the signal, and can provide higher angular resolution than traditional beamforming technology under the conditions of smaller antenna arrays and lower signal-to-noise ratio. However, these methods have not deviated from the research framework of Euclidean geometry, and the calculation of the similarity between the received signal and the weight vector is limited to the Euclidean distance, which is not conducive to coping with complex environments and is prone to high side lobes.

[0063] To solve the above problems, the present invention proposes a direction of arrival estimation method, system and terminal based on improved relative entropy, which have higher azimuth resolution capability and lower sidelobe level.

[0064] In a first aspect, the present invention provides a method for estimating a direction of arrival based on improved relative entropy, see Figure 1 , including the following steps:

[0065] S1. Use an array consisting of N array elements to receive the sound source signal, N ≥ 2;

[0066] As a possible implementation manner, the array is a horizontal, inclined or vertical uniform linear array, and / or the array element spacing is half of the wavelength of the narrowband signal.

[0067] S2. Extract the spectrum of the sound source signal, obtain the frequency of the desired signal included in the spectrum, calculate the covariance matrix of the frequency, normalize the covariance matrix, and obtain the normalized covariance matrix K; at this time, the normalized covariance matrix K is a symmetric positive definite matrix.

[0068] As an example, a frequency domain processing technique such as short-time Fourier transform or spectrum estimation is used to extract the signal spectrum, the frequency point with the highest power in the spectrum is selected as the frequency of the sound source signal, the covariance matrix of the frequency is calculated and normalized by the Frobenius norm, and the normalized covariance matrix is ​​expressed as follows:

[0069]

[0070] Among them, K 0 is the unnormalized covariance matrix, K is the covariance matrix obtained after Frobenius norm normalization, ||·|| F represents the Frobenius norm, k m,c Denotes the matrix K 0 The element in the mth row and cth column of 0 Dimension.

[0071] S3. K scanning grids are evenly divided within the preset spatial range, and the orientation of the kth scanning grid is represented by θ k , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements;

[0072] As a possible implementation method, the preset spatial range is [-90°, 90°], and K scanning grids are evenly divided in the spatial range of [-90°, 90°], specifically:

[0073] K scanning grids are evenly divided in the spatial range of [-90°, 90°] at intervals of preset angles, where the preset angles are any angles between 0.01° and 10°.

[0074] As a possible implementation method, the weight vector w(θ k ), specifically:

[0075]

[0076] Where N is the number of array elements, d is the array element spacing, λ is the signal wavelength, j is an imaginary number, and j 2 =-1.

[0077] S4. The weight vector w(θ k ) conjugate multiply to obtain the weight matrix of each scanning grid, and normalize the weight matrix to obtain the normalized weight matrix W(θ k );

[0078] As an example, let the weight vector w(θ k ) conjugate multiply to obtain the weight matrix on each scanning grid: W 0 =w(θ k ) H (θ k ), perform Frobenius norm normalization on the weight matrix to obtain the normalized weight matrix:

[0079]

[0080] Among them, w m,c Represents the matrix W 0 The element in the mth row and cth column of 0 dimension, at this time, the normalized weight matrix W(θ k ) is a symmetric positive definite matrix.

[0081] S5. The inverse of the covariance matrix K and the normalized weight matrix W(θ k ) multiplied by eigenvalues ​​to obtain the maximum eigenvalue λ max ;

[0082] As an example, the inverse of the covariance matrix K and the normalized weight matrix W(θ k ) is the matrix K - 1 W(θ k ), for the matrix K -1 W(θ k ) is decomposed into features, where there is a maximum eigenvalue λ containing the signal components max , and is much larger than the other eigenvalues.

[0083] S6. Configure the improved relative entropy distance expression;

[0084] As a possible implementation method, configuring an improved relative entropy distance expression includes the following steps:

[0085] S60. Define two symmetric positive definite matrices P and Q. The symmetric relative entropy distance between P and Q is expressed as:

[0086]

[0087] Where I is the identity matrix; tr means finding the trace of the matrix;

[0088] S61. Perform eigendecomposition on the matrix obtained by multiplying the inverse of the matrix P by the matrix Q:

[0089] (P -1 Q)=VΛV -1 ,

[0090] Where Λ=diag{λ 1 ,λ 2 ,...,λ n},λ i >0,i=1,2,...,n,Λ is a diagonal matrix,λ i represents the eigenvalue, V is the eigenvector matrix, and further we get:

[0091]

[0092] S62. Retain the maximum eigenvalue and obtain the improved relative entropy distance expression:

[0093]

[0094] S7. The maximum eigenvalue λ obtained in S5 max Substituting the above improved relative entropy distance expression, we can obtain the covariance matrix K and each weight matrix W(θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k )),get:

[0095]

[0096] Wherein, k=1...K represents the sequence numbers of different grid points, and K is the number of scanning grids.

[0097] S8. Based on the improved symmetric relative entropy distance D SKL (K,W(θ k )) obtain the normalized output power of each scanning grid;

[0098] As a possible implementation method, S8 includes the following steps:

[0099] S80. Solve for the improved relative entropy distance D SKL (K,W(θ k ))'s reciprocal, get S SKL (θ k ),Right now:

[0100] S SKL (θ k )=1 / D SKL (K,W(θ k ))

[0101] Determine S SKL (θ k ) in the maximum value max(S SKL (θ k ));

[0102] S81. Calculate the normalized output power of each scanning grid:

[0103] P(θ k )=10lg(S SKL (θ k ) / max(S SKL (θ k ))).

[0104] S9. Perform a one-dimensional spectrum peak search on the normalized output power. The scanning grid angle corresponding to the peak value is the arrival direction of the sound source signal.

[0105] In one possible implementation, 16 sensors are used to form a uniform linear array, with the array element spacing equal to the half wavelength of the narrowband signal. The acquired sound source signal is a single far-field narrowband signal. The scanning grid is evenly divided at intervals of 0.1° within the spatial range of [-90°, 90°]. The simulation results are shown in Figures 2 to 5 .

[0106] Figure 2 To simulate the relationship between the maximum sidelobe level and the number of snapshots in the spatial spectrum of the conventional beamforming (CBF) processor, the minimum variation distortionless response (MVDR) processor, the multi-criteria evaluation (MCE) processor and the improved relative entropy (Kullback-Leibler, SKL) processor proposed in the present invention when the signal-to-noise ratio is fixed at 5dB under the condition of a single signal; Figure 3 To simulate a single signal, when the signal-to-noise ratio is fixed at 5dB, the relationship between the main lobe width and the number of snapshots in the spatial spectrum of the CBF processor, MVDR processor, MCE processor and SKL processor is shown. It can be seen that the SKL processor proposed in the present invention has the lowest side lobe, the narrowest main lobe width and the best performance.

[0107] Figure 4 To simulate a single signal, the mean square error performance in the range of -20dB to 10dB for a fixed number of snapshots at 12; Figure 5 For the simulation of a single signal, the mean square error performance in the range of -20dB to 10dB of signal-to-noise ratio when the number of snapshots is fixed at 40. It can be seen that the mean square error performance of the SKL processor proposed in the present invention is the same as that of the MVDR and MCE methods, and is close to that of the CBF method.

[0108] In another possible implementation, 16 sensors are used to form a uniform linear array, and the array element spacing is equal to the half wavelength of the narrowband signal. The acquired sound source signals are four far-field narrowband signals, and the azimuths are θ 1 =0°,θ 2 =5°,θ 3 =-45°,θ 4 =30°. The scanning grid is evenly divided at intervals of 0.1° in the spatial range of [-90°, 90°]. For simulation results, see Figures 6 to 9 .

[0109] Figure 6To simulate the relationship between the maximum sidelobe level and the number of snapshots in the spatial spectrum of the CBF processor, the MVDR processor, the MCE processor and the SKL processor proposed by the present invention when the signal-to-noise ratio is fixed at 5dB under four signal conditions; Figure 7 To simulate the relationship between the main lobe width and the number of snapshots in the spatial spectrum of the CBF processor, MVDR processor, MCE processor and the SKL processor proposed in the present invention when the signal-to-noise ratio is fixed at 5dB under four signal conditions. It can be seen that in the case of multiple signals, the SKL processor proposed in the present invention still maintains good spatial spectrum performance. Figure 8 To simulate the spatial spectra of the CBF processor, MVDR processor, MCE processor, and SKL processor, the signal-to-noise ratio is fixed at 5 dB for the four signal cases; Fig. 9 For the simulation of four signals, the mean square error (MSE) performance in the range of -20dB to 10dB for a signal-to-noise ratio when the number of snapshots is fixed at 40. It can be seen that under high signal-to-noise ratio, the MSE performance of the SKL method provided by the present invention is significantly better than that of the CBF method.

[0110] In a second aspect, the present invention provides a direction of arrival estimation system based on improved relative entropy, comprising:

[0111] The sound source signal acquisition unit uses an array composed of N array elements to receive the sound source signal, where N ≥ 2;

[0112] The frequency covariance matrix solving unit of the normalized desired signal extracts the spectrum of the sound source signal, obtains the frequency of the desired signal included in the spectrum, calculates the frequency covariance matrix and normalizes the covariance matrix to obtain the normalized covariance matrix K;

[0113] The weight matrix solution unit after normalization of each scanning grid is evenly divided into K scanning grids in the spatial range of [-90°, 90°], and the orientation of the kth scanning grid is represented by θ k , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements; the weight vector w(θ k ) conjugate multiply to obtain the weight matrix of each scanning grid, and normalize the weight matrix to obtain the normalized weight matrix W(θ k );

[0114] The normalized output power solution unit of each scanning grid is the inverse of the covariance matrix K and the weight matrix W(θ k ) multiplied by eigenvalues ​​to obtain the maximum eigenvalue λ max , configure the improved relative entropy distance expression, and change the maximum eigenvalue λ maxSubstitute the improved relative entropy distance expression to obtain the covariance matrix K and each weight matrix W(θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k )), based on the improved symmetric relative entropy distance D SKL (K,W(θ k )) obtain the normalized output power of each scanning grid;

[0115] The direction of arrival estimation unit of the sound source signal performs a one-dimensional spectrum peak search on the normalized output power, and the scanning grid angle corresponding to the peak value is the direction of arrival of the sound source signal.

[0116] In a third aspect, the present invention provides a terminal, comprising a processor and a communication interface coupled to the processor, wherein the processor is used to run a computer program or instruction to implement the method for estimating a direction of arrival based on improved relative entropy provided in the first aspect.

[0117] Although the present invention is described herein in conjunction with various embodiments, in the process of implementing the claimed invention, those skilled in the art may understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the drawings, etc. In the specification, the word "comprising" does not exclude other components or steps, and "one" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the specification. Certain measures are recorded in different embodiments, but this does not mean that these measures cannot be combined to produce good results.

[0118] Although the present invention has been described in conjunction with specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present invention. Accordingly, this specification and the accompanying drawings are merely exemplary illustrations of the present invention and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present invention. Obviously, those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, the present invention is intended to include such modifications and variations if they fall within the scope of the present invention and its equivalents.

Claims

1. A method for estimating direction of arrival based on improved relative entropy, characterized in that: The steps include: S1. Use an array consisting of N array elements to receive the sound source signal, N ≥ 2; S2. extracting the spectrum of the sound source signal, obtaining the frequency of the desired signal included in the spectrum, calculating the covariance matrix of the frequency, normalizing the covariance matrix, and obtaining the normalized covariance matrix K; S3. K scanning grids are evenly divided within the preset spatial range, and the orientation of the kth scanning grid is represented by θ k , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements; S4. The weight vector w(θ k ) are conjugate multiplied to obtain the weight matrix of each scanning grid, and the weight matrix is ​​normalized to obtain the normalized weight matrix W(θ k );The normalized covariance matrix K and the normalized weight matrix W(θ k ) are all symmetric positive definite matrices; S5. The inverse of the covariance matrix K and the normalized weight matrix W (θ k ) is multiplied by each other to obtain the matrix, and the eigenvalues ​​are obtained by eigendecomposition, and the maximum eigenvalue λ among the eigenvalues ​​is obtained. max ; S6. Configuring an improved relative entropy distance expression includes the following steps: S60. Define two symmetric positive definite matrices P and Q. The symmetric relative entropy distance between P and Q is expressed as: Where I is the identity matrix; tr means finding the trace of the matrix; S61. Perform eigendecomposition on the matrix obtained by multiplying the inverse of the matrix P by the matrix Q: (P -1 Q)=VΛV -1 , Among them, Λ=diag{λ1, λ2,…,λ n },λ i >0,i=1,2,…,n,Λ is a diagonal matrix,λ i represents the eigenvalue, V is the eigenvector matrix, and further we get: S62. Retain the maximum eigenvalue and obtain the improved relative entropy distance expression: S7. The maximum eigenvalue λ max Substitute the improved relative entropy distance expression to obtain the covariance matrix K and each weight matrix W (θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k )); S8. Based on the improved symmetric relative entropy distance D SKL (K,W(θ k )) obtaining the normalized output power of each of the scanning grids; S9. Perform a one-dimensional spectrum peak search on the normalized output power, and the scanning grid angle corresponding to the peak value is the wave arrival direction of the sound source signal.

2. The method for estimating the direction of arrival based on improved relative entropy according to claim 1, characterized in that: The weight vector w(θ) of each scanning grid is obtained based on the frequency of the desired signal and the array parameters. k ), specifically: Where N is the number of array elements, d is the array element spacing, λ is the signal wavelength, j is an imaginary number, and j 2 =-1.

3. The method for estimating direction of arrival based on improved relative entropy according to claim 1, characterized in that: The S7 is specifically:

4. The method for estimating direction of arrival based on improved relative entropy according to claim 1, characterized in that: The S8 includes: S80. Solve for the improved relative entropy distance D SKL (K,W(θ k ))'s reciprocal, get S SKL (θ k ), determine S SKL (θ k ) in the maximum value max(S SKL (θ k )); S81. Calculate and obtain the normalized output power of each scanning grid: P(θ k )=10lg(S SKL (θ k ) / max(S SKL (θ k )))。 5. The method for estimating direction of arrival based on improved relative entropy according to claim 1, characterized in that: The preset spatial range is [-90°, 90°], and K scanning grids are evenly divided within the spatial range of [-90°, 90°], specifically: K scanning grids are evenly divided in the spatial range of [-90°, 90°] at intervals of preset angles, where the preset angles are any angles between 0.01° and 10°.

6. The method for estimating direction of arrival based on improved relative entropy according to claim 1, characterized in that: The array is a horizontal, inclined or vertical uniform linear array, and / or the array element spacing is half of the wavelength of the narrowband signal.

7. A direction of arrival estimation system based on improved relative entropy, characterized in that: include: The sound source signal acquisition unit uses an array composed of N array elements to receive and acquire the sound source signal, where N ≥ 2; A frequency covariance matrix solving unit for the normalized desired signal extracts the spectrum of the sound source signal, obtains the frequency of the desired signal included in the spectrum, calculates the covariance matrix of the frequency and normalizes the covariance matrix to obtain a normalized covariance matrix K; The weight matrix solution unit after normalization of each scanning grid is evenly divided into K scanning grids in the spatial range of [-90°, 90°], and the orientation of the kth scanning grid is represented by θ k , based on the frequency of the desired signal and the array parameters, the weight vector w(θ k ), the array parameters are the number of array elements and the spacing between array elements; the weight vector w(θ k ) are conjugate multiplied to obtain the weight matrix of each scanning grid, and the weight matrix is ​​normalized to obtain the normalized weight matrix W(θ k ), the normalized covariance matrix K and the normalized weight matrix W(θ k ) are all symmetric positive definite matrices; The normalized output power solving unit of each scanning grid is used to calculate the inverse of the covariance matrix K and the weight matrix W(θ k ) is multiplied by each other to obtain the matrix, and the eigenvalues ​​are obtained by eigendecomposition, and the maximum eigenvalue λ among the eigenvalues ​​is obtained. max , configure the improved relative entropy distance expression, including the following steps: First, define two symmetric positive definite matrices P and Q. The symmetric relative entropy distance between P and Q is expressed as: Where I is the identity matrix; tr means finding the trace of the matrix; Secondly, perform eigendecomposition on the matrix obtained by multiplying the inverse of matrix P by matrix Q: (P -1 Q)=VΛV -1 , Among them, Λ=diag{λ1, λ2,...,λ n },λ i >0,i=1,2,...,n,Λ is a diagonal matrix,λ i represents the eigenvalue, V is the eigenvector matrix, and further we get: Finally, the maximum eigenvalue is retained to obtain the improved relative entropy distance expression: The maximum eigenvalue λ max Substitute the improved relative entropy distance expression to obtain the covariance matrix K and each weight matrix W (θ k ) of the improved symmetric relative entropy distance D SKL (K,W(θ k )), based on the improved symmetric relative entropy distance D SKL (K,W(θ k )) obtaining the normalized output power of each of the scanning grids; The direction of arrival estimation unit of the sound source signal performs a one-dimensional spectrum peak search on the normalized output power, and the scanning grid angle corresponding to the peak value is the direction of arrival of the sound source signal.

8. A terminal comprising a processor and a communication interface coupled to the processor, wherein the processor is configured to run a computer program or instruction to implement the method for estimating a direction of arrival based on improved relative entropy according to any one of claims 1 to 6.

Images