Manifold separation based dual-channel receiver direction finding method and system

Abstract

The application discloses a manifold separation-based double-channel receiver direction finding method and system, signal sources are prepared according to the characteristics of a scene to be measured and objects; the positions of the signal sources are arranged according to an array reference direction and a current observation round; the covariance matrix of a current incident direction signal is calculated; the obtained covariance matrix is subjected to eigenvalue decomposition and is spliced into an equivalent array manifold, a truncation transformation matrix of the equivalent array manifold of the array is constructed; the number of signal sources is estimated; the covariance matrix of a current unknown incident direction signal is calculated, the obtained covariance matrix is subjected to eigenvalue decomposition; a Vandermonde vector is constructed and a spatial spectrum is obtained by using a multiple signal classification algorithm, and the maximum value of the spatial spectrum is selected as an angle estimation value. The application can achieve super-resolution direction finding effect, considers errors caused by defects of the array itself in an offline process, and has higher robustness in an actual direction finding process.

Description

Technical Field

[0001] This invention relates to the field of antenna array signal processing technology, and in particular to a direction finding method and system for a dual-channel receiver based on manifold separation. Background Technology

[0002] Direction finding is the estimation of the angle of a radio transmitted signal. Common methods include phase difference direction finding, time difference direction finding, and multi-channel amplitude comparison direction finding. Amplitude comparison direction finding utilizes the directional characteristics of directional antenna arrays. The output amplitude of a directional antenna array has a definite correspondence with the direction of radio wave propagation. By comparing the amplitude, the azimuth of the target can be determined. A typical Watson-Watt direction finder in amplitude comparison uses two sets of orthogonal antennas and calculates the arctangent value to give the signal's direction of arrival. Time difference-based direction finding methods determine the signal by measuring the time difference between the arrival of the signal at different spatial antenna elements. Based on the linearity of electromagnetic wave propagation and its fixed propagation speed, the precise time difference between the arrival times of the wavefront at multiple receiving antennas is calculated to build a model and deduce the signal source's azimuth. Phase difference direction finding methods determine the direction of the incident signal by measuring the phase difference formed between different antenna elements of the incoming signal. Therefore, the number of channels used can be determined based on the method of obtaining the phase difference. In scenarios with insufficient channels, interferometer algorithms are generally used, while in scenarios with full channels, spatial spectrum algorithms are generally used.

[0003] Due to its compatibility with various array structures and deployment environments, phase difference direction finding remains the most widely used direction finding method. Phase difference direction finding is mainly divided into two technical routes.

[0004] One method is interferometric direction finding. By switching antennas, the phase information of the signal under test received by different antenna pairs in the antenna array is extracted, thereby obtaining the azimuth information of the signal under test. Phase interferometers directly calculate the phase difference between antenna pairs, while correlation interferometers obtain the angle by directly correlating the phase difference between antenna pairs with a sample library. Both methods feature simple algorithms, fast direction finding speed, strong environmental adaptability, and wide frequency coverage.

[0005] The second method is the spatial spectrum direction finding method, which is based on a multi-channel receiver. It quantizes the response of the incoming wave of the signal source into digital signals on each array element. By analyzing the spatial spectrum of the incoming wave, the spatial direction finding of the incoming wave can be obtained. This method can realize the simultaneous direction finding of multiple signals and super-resolution direction finding. It can relax the requirements of traditional direction finding methods on the measurement environment and the arrangement of direction finding antenna array elements, and has extremely high direction finding accuracy.

[0006] However, both phase difference-based direction finding schemes have different limitations. In the interferometric direction finding method, the phase interferometer needs to unambiguously resolve the phase difference obtained from the direction finding signal, while the correlation interferometer needs to establish a sample library based on different frequencies. In addition, when co-frequency interference exists, the direction finding performance of both the phase interferometer and the correlation interferometer is significantly reduced.

[0007] While traditional spatial spectrum estimation direction-finding methods offer considerable estimation accuracy, they also incur high implementation costs. This approach requires consistency between the number of channels and antennas, imposes strict requirements on the amplitude and phase consistency of the receiving channels, and places high demands on the sampling performance of the multi-channel receiver, particularly in large-scale antenna array scenarios. Furthermore, the computationally intensive operations such as matrix multiplication, inversion, and decomposition necessitate the use of high-performance processors. Summary of the Invention

[0008] Purpose of the invention: This invention provides a dual-channel receiver direction finding method and system based on manifold separation, which is more suitable for scenarios with limited hardware costs but high accuracy requirements.

[0009] Technical solution: The present invention provides a dual-channel receiver direction finding method based on manifold separation, comprising the following steps:

[0010] Step 1: Prepare the signal source according to the characteristics of the scene and object to be measured;

[0011] Step 2: Pre-set the incident positions of Q signal sources in the 360° angular domain. The positions of the signal sources are deployed according to the array reference direction and the current observation round. The array reference direction is the direction pointed to by the x-axis in its three-dimensional relative coordinate system.

[0012] Step 3: Calculate the covariance matrix of the signal in the current incident direction. For different positions in the covariance matrix, switch different antenna pairs. After the antenna pair switching is completed, perform a Fast Fourier Transform (FFT) on the time-domain signal received by the antenna pair to obtain the frequency-domain signal, and extract the elements of the corresponding frequency points in the frequency domain. Correlate the elements of the corresponding frequency points between the antennas and add them to the corresponding positions in the covariance matrix to complete the calculation of the covariance matrix of the current incident signal.

[0013] Step 4: Perform eigenvalue decomposition on the covariance matrix obtained in Step 3, and insert the eigenvector corresponding to the largest eigenvalue into the equivalent array manifold. Construct a truncation transformation matrix for the equivalent array manifold of the array.

[0014] Step 5: Estimate the number of information sources;

[0015] Step 6: Calculate the covariance matrix of the signal with the unknown incident direction, perform eigenvalue decomposition on the obtained covariance matrix, and set the minimum... The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. ,in and These represent the number of array elements and the number of incident signals, respectively.

[0016] Step 7: Construct the Vandermonde vector. Project the array of any two-dimensional planar structure onto the virtual linear array using the truncation transformation matrix. Obtain the spatial spectrum using the MUSIC algorithm and search for the largest value in the spatial spectrum. Each spectral peak yields a corresponding angle estimate.

[0017] Furthermore, in step 1, the characteristics of the scene and object to be measured include the signal bandwidth, intermediate frequency, and sampling rate.

[0018] Furthermore, in step 3, let the number of array elements be... , for the composition The structure of the complex covariance matrix of dimension Second observation, A single observation contains The first snapshot captures the nth observation. In a quick snapshot, , The first switch in the antenna pair switches to the position corresponding to the nth observation in the covariance matrix. One antenna, And obtain the corresponding time-domain signal. The second switch in the antenna pair switches to the position corresponding to the nth observation in the covariance matrix. One antenna, And obtain the corresponding time-domain signal. , respectively and Perform FFT transformation to obtain and ;extract and The values ​​at corresponding frequency points are correlated, and the correlated values ​​are added to the first value of the covariance matrix. On each element; when continuously adding to the corresponding position in the covariance matrix... After obtaining the correlation values, the elements corresponding to the covariance matrix are obtained; after N observations, when both switches have completed a full round of switching, the covariance matrix corresponding to the current deployment direction is finally obtained.

[0019] Furthermore, to ensure the Hermitian property of the covariance matrix, a conjugate average is performed on the diagonally symmetric matrix elements after the calculation. This results in the covariance matrix after conjugate averaging. Represented as:

[0020]

[0021] in Representation matrix The Line number List, express conjugate, and They represent and The specific measured values.

[0022] Furthermore, in step 4, during each calculation, the eigenvector corresponding to the maximum eigenvalue of each preset grid in step 2 is obtained, and then... After calculating each incident direction individually, they are spliced ​​together to form an equivalent array manifold. .

[0023] Furthermore, in step 4, a Discrete Inverse Fourier Transform (IDFT) is performed on each row of the equivalent array manifold to obtain the complete transformation matrix. Using the transformation center of the IDFT as a reference, the matrix is ​​extended to the left and right. By truncating the points, a complete symmetric truncation matrix is ​​obtained. For the cut-off length It must comply with the following constraints:

[0024]

[0025] in As an empirical upper bound, in which For the corresponding frequency point, Where is the radius of the circular array. Wavefront velocity, Indicates the number of array elements. Indicates the number of observations.

[0026] Furthermore, in step 6, the calculated covariance matrix is ​​expressed as follows:

[0027]

[0028] Where E() represents taking the expectation. and They are respectively the front eigenvalues ​​and The matrix composed of eigenvectors corresponding to each eigenvalue corresponds to the signal subspace and the noise subspace. For a former A diagonal matrix with non-zero diagonal elements. This represents the matrix of the Vandermonde structure spanned by vectors representing the Vandermonde structures corresponding to the angles of incidence. Let be the autocorrelation matrix of the information source, which ideally has a diagonal matrix structure. For noise power, Represents the identity matrix. Represents the received signal matrix. Representation matrix The conjugate transpose of . This is the truncation transformation matrix.

[0029] Furthermore, in step 7, the guide vector of the Vandermonde structure is constructed:

[0030]

[0031] Indicates the number of array elements. Indicates the angle of the incident direction.

[0032] Furthermore, in step 7, the MUSIC algorithm is used to search for the largest [value] in the spatial spectrum. Each spectral peak yields a corresponding estimate, and its cost function is expressed as:

[0033]

[0034] As the guide vector, Guide vector The conjugate transpose of . To truncate the conjugate transpose of the matrix, For the front A matrix consisting of the eigenvectors corresponding to the remaining eigenvalues. for The conjugate transpose of .

[0035] Correspondingly, the system for a dual-channel receiver direction-finding method based on manifold separation includes: a signal source processing module, a covariance matrix calculation module, a truncated transform matrix construction module, and an online direction-finding module. The signal source processing module prepares the signal source and determines its parameters based on the characteristics of the signal to be measured, and deploys the signal source according to the absolute direction of the first array element and the current observation round. The covariance matrix calculation module calculates the covariance matrix of the signal in the current incident direction. The truncated transform matrix construction module performs eigenvalue decomposition on the obtained covariance matrix and incorporates it into the equivalent array manifold, constructing a truncated transform matrix on the equivalent array manifold. The online direction-finding module estimates the number of signal sources, calculates the covariance matrix of the signal in the current unknown incident direction, performs eigenvalue decomposition on the obtained covariance matrix, and sets the minimum... The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. Construct Vandermonde vectors and use the MUSIC algorithm to obtain the spatial spectrum, then search for the largest spatial spectrum. Each value represents a selected angle estimate.

[0036] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: (1) In the traditional spatial spectrum method, there are enough channels, so a batch of time domain data can be collected first, and the vectors of the corresponding frequency points can be extracted after the time domain data is FFTed. Repeating this process multiple times will splice the vectors into a matrix. The first step to obtain the covariance matrix is ​​to perform autocorrelation on the obtained matrix. In the present invention, due to the limitation of the number of channels, the corresponding frequency point elements can only be extracted after a batch of time domain data is collected, and the two elements are correlated and placed in the corresponding position of the covariance matrix. In order to minimize the influence of noise, the sampling-FFT-frequency point extraction-cross-correlation calculation-placing operation needs to be repeated multiple times at a certain position of the covariance matrix. (2) The present invention can achieve super-resolution direction finding effect. Higher resolution can be set only during the online process to achieve higher measurement accuracy. (3) Errors caused by array defects can be considered during the offline process, such as non-uniform noise, amplitude and phase errors and estimation errors caused by mutual coupling effects, thus having higher robustness in the actual direction finding process. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the array structure corresponding to the present invention.

[0038] Figure 2 This is a schematic diagram of the method flow of the present invention.

[0039] Figure 3 This is a schematic diagram of the data acquisition process of the present invention.

[0040] Figure 4This is a schematic diagram of the covariance matrix calculation process of the present invention.

[0041] Figure 5 This is the omnidirectional response diagram of the relevant interferometer.

[0042] Figure 6 This is the omnidirectional response diagram of the traditional spatial spectrum method.

[0043] Figure 7 This is an omnidirectional response diagram of the method proposed in this invention.

[0044] Figure 8 This is a multi-source response diagram using the traditional spatial spectrum method.

[0045] Figure 9 This is a multi-source response diagram of the method proposed in this invention.

[0046] Figure 10 This is a comparison diagram of the spatial response of the three methods under a single incident source according to the present invention.

[0047] Figure 11 This is a comparison diagram of the spatial response of the two methods under multiple incident sources according to the present invention. Detailed Implementation

[0048] like Figure 1 and Figure 2 As shown, a direction-finding method for a dual-channel receiver based on manifold separation includes the following steps:

[0049] Step 1: Prepare the signal source according to the characteristics of the scene and object to be measured;

[0050] Step 2: Pre-set the incident positions of Q signal sources in the 360° angular domain. The positions of the signal sources are deployed according to the array reference direction and the current observation round. The array reference direction is the direction pointed to by the x-axis in its three-dimensional relative coordinate system.

[0051] Step 3: Calculate the covariance matrix of the signal in the current incident direction. For different positions in the covariance matrix, switch different antenna pairs. After the antenna pair switching is completed, perform a Fast Fourier Transform (FFT) on the time-domain signal received by the antenna pair to obtain the frequency-domain signal, and extract the elements of the corresponding frequency points in the frequency domain. Correlate the elements of the corresponding frequency points between the antennas and add them to the corresponding positions in the covariance matrix to complete the calculation of the covariance matrix of the current incident signal.

[0052] Step 4: Perform eigenvalue decomposition on the covariance matrix obtained in Step 3, and insert the eigenvector corresponding to the largest eigenvalue into the equivalent array manifold. Construct a truncation transformation matrix for the equivalent array manifold of the array.

[0053] Step 5: Estimate the number of information sources;

[0054] Step 6: Calculate the covariance matrix of the signal with the unknown incident direction, perform eigenvalue decomposition on the obtained covariance matrix, and set the minimum... The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. ,in and These represent the number of array elements and the number of incident signals, respectively.

[0055] Step 7: Construct the Vandermonde vector. Project the array of any two-dimensional planar structure onto the virtual linear array using the truncation transformation matrix. Obtain the spatial spectrum using the MUSIC algorithm and search for the largest value in the spatial spectrum. Each spectral peak yields a corresponding angle estimate.

[0056] like Figure 3 and Figure 4 As shown, in step 3, let the number of array elements be... , for the composition The structure of the complex covariance matrix of dimension Second observation, A single observation contains The first snapshot captures the nth observation. In a quick snapshot, , The first switch in the antenna pair switches to the position corresponding to the nth observation in the covariance matrix. One antenna, And obtain the corresponding time-domain signal. The second switch in the antenna pair switches to the position corresponding to the nth observation in the covariance matrix. One antenna, And obtain the corresponding time-domain signal. , respectively and Perform FFT transformation to obtain and ;extract and The values ​​at corresponding frequency points are correlated, and the correlated values ​​are added to the first value of the covariance matrix. On each element; when continuously adding to the corresponding position in the covariance matrix... After obtaining the correlation values, the elements corresponding to the covariance matrix are obtained; after N observations, when both switches have completed a full round of switching, the covariance matrix corresponding to the current deployment direction is finally obtained.

[0057] To ensure the Hermitian property of the covariance matrix, a conjugate average is performed on the diagonal-symmetric elements after the calculation. This results in the covariance matrix after the conjugate average. It can be represented as:

[0058]

[0059] in Representation matrix The Line number List, express . conjugate.

[0060] In step 4, during each calculation, the eigenvector corresponding to the maximum eigenvalue of each preset grid in step 2 is obtained, and then... After calculating each incident direction individually, they are spliced ​​together to form an equivalent array manifold. .

[0061] Performing an IDFT transform on each row of the equivalent array manifold yields the complete transform matrix. Extending the matrix to the left and right, with the transform center of the IDFT as the reference,... By truncating the point, a complete symmetric truncation calibration matrix is ​​obtained. For the cut-off length It must comply with the following constraints:

[0062]

[0063] in As an empirical upper bound, in which For the corresponding frequency point, Where is the radius of the circular array. This represents the wavefront velocity.

[0064] In step 6, the calculated covariance matrix is ​​expressed as:

[0065]

[0066] Where E() represents taking the expectation. and They are respectively the front eigenvalues ​​and The matrix composed of eigenvectors corresponding to each eigenvalue corresponds to the signal subspace and the noise subspace. For a former A diagonal matrix with non-zero elements. This represents the matrix of the Vandermonde structure spanned by vectors representing the Vandermonde structures corresponding to the angles of incidence. For noise power, Represents the identity matrix. This represents the received signal matrix.

[0067] In step 7, the guide vector of the Vandermonde structure is constructed:

[0068]

[0069] Use the MUSIC algorithm to search for the largest in the spatial spectrum. Each spectral peak yields a corresponding estimate, and its cost function is expressed as:

[0070]

[0071] The calibration matrix can project an array of arbitrary two-dimensional planar structures onto a virtual linear array, thereby obtaining the spatial spectrum using the MUSIC algorithm and searching for the largest value in the spatial spectrum. Each spectral peak yields a corresponding estimated value.

[0072] This invention can be divided into two processes: offline and online. First, calibration is performed by truncating the calibration matrix during the offline process, so that the incident signal can be measured during the online process.

[0073] Offline process:

[0074] Step 1: Prepare the signal source and determine its parameters based on the characteristics of the signal to be measured. First, determine the bandwidth, intermediate frequency (IF), and sampling rate of the preset signal. For example, if the signal bandwidth is 20MHz, the IF is 2.4GHz, the sampling rate is 50MHz, the FFT points are 8192, and the grid resolution for offline calculation and online scanning is 1°, then the required number of offline covariance matrix calculations for constructing the calibration matrix is ​​360. During each covariance matrix calculation, the number of snapshots for offline observation and online observation are 1000 and 100 respectively. Simultaneously, determine the position of the first array element and the numbering of each element to facilitate subsequent modeling. To prevent unknown co-channel interference, the offline observation scenario is set on an open rooftop, and it has been confirmed beforehand that there are no drones or other unknown interference sources operating on the same frequency band nearby. During the offline observation phase, the narrowband signal generated by the up-converted signal source is used as the calibration signal. After setting the signal parameters, receive the IQ signal from the air interface and verify whether there are any problems with the signal source's transmission and the antenna array's reception.

[0075] Step 2: Deploy the signal source according to the absolute direction of the first array element and the position of the current observation round.

[0076] Step 3: Calculate the covariance matrix of the signal in the current incident direction. Using 8192 sampling points as one snapshot, after receiving a sufficient number of sampling points, perform an FFT transformation on these 8192 sampling points. The conjugate product of the corresponding frequencies of the two antennas after the transformation is the corresponding element of the covariance matrix under the current snapshot, and accumulate it at the corresponding positions. Performing the above steps to accumulate the covariance matrix 1000 times yields the covariance matrix calculated for the current direction.

[0077] Step 4: Perform eigenvalue decomposition on the obtained covariance matrix and concatenate it into the equivalent array manifold. Construct a truncation transformation matrix for the equivalent array manifold. During each observation, obtain the eigenvectors corresponding to the maximum eigenvalue of each preset grid and concatenate them into the equivalent array manifold. The covariance matrix decomposition obtained from the nth observation is used to concatenate the eigenvectors corresponding to the largest eigenvalues ​​of the decomposition into the equivalent array. In the list.

[0078] First, perform an IDFT transform on each row of the equivalent array manifold to obtain the complete transform matrix. Finally, using the transform center of the IDFT as the reference, extend the matrix to the left and right. The point is truncated to obtain a complete symmetric truncated transformation matrix. The range of truncation orders given in this invention, based on the array structure and the incident source frequency, is clearly... Here, the cut-off length is set to 35.

[0079] Online process:

[0080] Step 5: Roughly estimate the number of information sources based on source estimation algorithms such as the Akaike Information Criterion (AIC) and the Minimum Redundancy Length (MDL) method. .

[0081] Step 6: Calculate the covariance matrix of the signal with the unknown incident direction, perform eigenvalue decomposition on the obtained covariance matrix, and set the minimum... The eigenvectors corresponding to the eigenvalues ​​are noise subspace vectors. In the online measurement scenario, the number of information sources and the number of array elements are known to be 1 and 7 respectively. Therefore, the eigenvectors corresponding to the 6 smallest eigenvalues ​​are taken as the noise subspace vectors.

[0082] Step 7: Construct the Vandermonde vector and use the MUSIC algorithm to obtain the spatial spectrum, then search for the vector with the largest spatial spectrum. Each value represents a selected angle estimate. First, the guide vector of the Vandermonde structure needs to be constructed:

[0083]

[0084] Use the MUSIC algorithm to search for the largest in the spatial spectrum. The estimated value can be obtained by identifying each spectral peak, and its cost function can be expressed as:

[0085]

[0086] The calibration matrix can project an array of arbitrary two-dimensional planar structures onto a virtual linear array, thereby obtaining the spatial spectrum using the MUSIC algorithm and searching for the largest value in the spatial spectrum. Each spectral peak yields a corresponding estimated value.

[0087] like Figure 5-7 As shown, the antenna array used in the simulation is a 7-element antenna array with a radius of 0.18m and a signal-to-noise ratio of 2dB. The carrier frequency of the incident signal is 3GHz. Figure 5-7 The omnidirectional scanning capabilities of the relevant interferometer algorithm, the traditional spatial spectrum algorithm, and the proposed method are demonstrated respectively, with the incident angle ranging from -175°. Scan to 175 Compared to the correlation interferometer algorithm, this invention has a narrower main lobe while maintaining omnidirectional accurate estimation capability. Compared to the traditional spatial spectrum method, this invention has a lower sidelobe level.

[0088] like Figure 8-9 As shown, the three incident angles are respectively -265° -175 With -85 Scan to 85 175 With 265 The antenna array structure, signal-to-noise ratio, and carrier frequency of the incident signal used in the simulation are all... Figure 5-7 The simulation results are consistent. Compared to traditional spatial spectrum methods, this invention reduces the number of channels to two while still achieving lower sidelobes.

[0089] This invention considers the measurement effect of the incident source direction under two scenarios. For example... Figure 10 As shown, a comparison diagram of the spatial response of the proposed method, traditional spatial spectrum methods, and related interferometer methods is presented, where the incident direction is 40°. The sidelobe level of the method proposed in this invention is reduced by about 10dB compared to the original multi-channel array spatial spectrum. Compared with the interferometer direction finding method, it has more stable sidelobes and a narrower main lobe, which shows that the present invention has more obvious distinguishability of incident signals in the direction finding process of single incident signal scenario.

[0090] like Figure 11 As shown, a comparison diagram of the spatial response of the proposed method, traditional spatial spectrum methods, and related interferometer methods is presented, with incident directions of -55° and 125°, respectively. The sidelobe level of the proposed method is reduced by about 5dB compared to the original multi-channel array spatial spectrum, and it has a narrower main lobe and lower side lobes, demonstrating that the proposed method has a more obvious distinguishability of incident signals in the direction finding process of multi-incident signal scenarios.

[0091] The process of an array receiving spatial signals can be viewed as performing a linear transformation on the received signal. Similarly, the process of a uniform linear array with a Vandermonde structure and a steering vector receiving signals can be considered a discrete Fourier transform. This invention, based on the connection between a uniform circular array and a virtual uniform linear array, approximates the wavefield manifold received by the array through orthogonal decomposition into a matrix and another Vandermonde structure vector (steering vector) that is only affected by the incident source. This improves the fitting quality of the spatial spectrum in subsequent angle estimation. Likewise, addressing the limitations of dual-channel arrays in terms of channel number and the difficulty of array calibration in spatial spectrum algorithms, an equivalent array manifold is constructed using the eigenvector corresponding to the largest eigenvalue in the covariance matrix obtained by antenna pair switching, replacing the original array manifold. This effectively considers array errors while avoiding the complexity of obtaining the array manifold during the calculation of the calibration matrix.

[0092] Accordingly, a dual-channel receiver direction-finding system based on manifold separation includes: a signal source processing module, a covariance matrix calculation module, a truncated transform matrix construction module, and an online direction-finding module. The signal source processing module prepares the signal source and determines its parameters based on the characteristics of the signal to be measured, and deploys the signal source according to the absolute direction of the first array element and the current observation round. The covariance matrix calculation module calculates the covariance matrix of the signal in the current incident direction. The truncated transform matrix construction module performs eigenvalue decomposition on the obtained covariance matrix and incorporates it into the equivalent array manifold, constructing a truncated transform matrix on the equivalent array manifold. The online direction-finding module estimates the number of signal sources, calculates the covariance matrix of the signal in the current unknown incident direction, performs eigenvalue decomposition on the obtained covariance matrix, and sets the minimum... The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. Construct Vandermonde vectors and use the MUSIC algorithm to obtain the spatial spectrum, then search for the largest spatial spectrum. Each value represents a selected angle estimate.

Claims

1. A direction-finding method for a dual-channel receiver based on manifold separation, characterized in that, Includes the following steps: Step 1: Prepare the signal source according to the characteristics of the scene and object to be measured; Step 2: Pre-set the incident positions of Q signal sources in the 360° angular domain. The positions of the signal sources are deployed according to the array reference direction and the current observation round. The array reference direction is the direction pointed to by the x-axis in its three-dimensional relative coordinate system. Step 3: Calculate the covariance matrix of the signal in the current incident direction; switch different antenna pairs for different positions in the covariance matrix; After the antenna pair switching is completed, a Fast Fourier Transform (FFT) is performed on the time-domain signal received by the antenna pair to obtain the frequency-domain signal, and the elements of the corresponding frequency points are extracted. The elements of the corresponding frequency points between the antennas are correlated and added to the corresponding positions of the covariance matrix, thus completing the calculation of the covariance matrix of the current incident signal. Step 4: Perform eigenvalue decomposition on the covariance matrix obtained in Step 3, and insert the eigenvector corresponding to the largest eigenvalue into the equivalent array manifold. Construct a truncation transformation matrix for the equivalent array manifold of the array. Step 5: Estimate the number of information sources; Step 6: Calculate the covariance matrix of the signal with the unknown incident direction, perform eigenvalue decomposition on the obtained covariance matrix, and set the minimum... The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. ,in and These represent the number of array elements and the number of incident signals, respectively. Step 7: Construct the Vandermonde vector. Project the array of any two-dimensional planar structure onto the virtual linear array using the truncation transformation matrix. Obtain the spatial spectrum using the MUSIC algorithm and search for the largest value in the spatial spectrum. Each spectral peak yields a corresponding angle estimate.

2. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, In step 1, the characteristics of the scene and object to be measured include the signal bandwidth, intermediate frequency, and sampling rate.

3. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, To ensure the Hermitian property of the covariance matrix, the elements of the diagonally symmetric matrix are averaged after the calculation.

4. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, In step 4, during each calculation, the eigenvector corresponding to the maximum eigenvalue of each preset grid is obtained, and then... After calculating each incident direction individually, they are spliced ​​together to form an equivalent array manifold. .

5. The dual-channel receiver direction finding method based on manifold separation as described in claim 4, characterized in that, In step 4, a Discrete Inverse Fourier Transform (IDFT) is performed on each row of the equivalent array manifold to obtain the complete transformation matrix. The matrix is ​​then extended to the left and right, using the transform center of the IDFT as the reference. By truncating the points, a complete symmetric truncation matrix is ​​obtained. For the cut-off length It must comply with the following constraints: in As an empirical upper bound, in which For the corresponding frequency point, Where is the radius of the circular array. Wavefront velocity, Indicates the number of array elements. Indicates the number of observations.

6. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, In step 6, the calculated covariance matrix is ​​expressed as follows: Where E() represents taking the expectation. and They are respectively the front eigenvalues ​​and The matrix composed of eigenvectors corresponding to each eigenvalue corresponds to the signal subspace and the noise subspace. For a former A diagonal matrix with non-zero diagonal elements. This represents the matrix of the Vandermonde structure spanned by vectors representing the Vandermonde structures corresponding to the angles of incidence. Let be the autocorrelation matrix of the information source, which ideally has a diagonal matrix structure. For noise power, Represents the identity matrix. Represents the received signal matrix. Representation matrix The conjugate transpose of . This is the truncation transformation matrix.

7. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, In step 7, the guide vector of the Vandermonde structure is constructed: Indicates the number of array elements. Indicates the angle of the incident direction.

8. The dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, In step 7, the MUSIC algorithm is used to search for the largest value in the spatial spectrum. Each spectral peak yields a corresponding estimate, and its cost function is expressed as: As the guide vector, Guide vector The conjugate transpose of . To truncate the conjugate transpose of the matrix, For the front A matrix consisting of the eigenvectors corresponding to the remaining eigenvalues. for The conjugate transpose of .

9. A system for implementing the dual-channel receiver direction finding method based on manifold separation as described in claim 1, characterized in that, include: Signal source processing module, covariance matrix calculation module, truncation transformation matrix construction module, and online direction finding module; The signal source processing module prepares the signal source and determines the parameters according to the characteristics of the signal to be measured, and places the signal source according to the absolute direction of the first array element and the current observation round; the covariance matrix calculation module calculates the covariance matrix of the signal in the current incident direction; the truncation transformation matrix construction module performs eigenvalue decomposition on the obtained covariance matrix and splices it into the equivalent array manifold, and constructs the truncation transformation matrix on the equivalent array manifold of the array. The online direction-finding module estimates the number of signal sources, calculates the covariance matrix of the signal from the currently unknown incident direction, performs eigenvalue decomposition on the obtained covariance matrix, and sets the minimum eigenvalue. The eigenvectors corresponding to the eigenvalues ​​are matrices spanned by the vectors of the noise subspace. Construct Vandermonde vectors and use the MUSIC algorithm to obtain the spatial spectrum, then search for the largest spatial spectrum. Each value represents a selected angle estimate.

Images