Robust DOA estimation method based on full-connection neural network

Abstract

The invention particularly relates to a robust DOA estimation method based on a full-connection neural network, and the method comprises the steps: receiving far-field narrow-band signals through all array elements in a sensor array, and constructing an observation data vector based on the far-field narrow-band signals received by all array elements; determining a guiding vector according to the arrival angle of the far-field narrow-band signal; constructing a covariance matrix based on the observation data vector; taking the covariance matrix as a model output parameter, inputting the covariance matrix into a trained phase error estimation model based on a full-connection neural network, and obtaining a phase error estimation value output by the model; determining a corresponding phase error estimation matrix based on the phase error estimation value; compensating the guide vector by using the phase error estimation matrix to obtain a calibrated guide vector; constructing a MUSIC spatial spectrum function in combination with the calibrated guide vector and the noise subspace; and searching a spectrum peak of the MUSIC spatial spectrum function in the angle search range, and determining a DOA estimated value of the signal source according to a spectrum peak search result. According to the method, relatively high DOA estimation precision can be kept.

Description

Technical Field

[0001] The present invention relates to the technical field of array signal processing, and particularly to a robust DOA estimation method based on a fully connected neural network. Background Art

[0002] In the related art, direction of arrival (DOA) estimation is a major research content in the field of array signal processing. Common DOA estimation methods include: traditional beamforming methods, subspace methods, maximum likelihood methods, and array calibration methods, etc. Among them, the multiple signal classification (MUSIC) algorithm based on the subspace method has been widely studied due to its high resolution ability. The MUSIC algorithm performs eigenvalue decomposition on the covariance matrix of the array received signals, divides the eigenvector space into a signal subspace and a noise subspace according to the magnitudes of the eigenvalues, constructs a spatial spectrum based on their orthogonality, and then obtains the DOA estimation of the signal sources. This algorithm has excellent angle resolution ability under ideal conditions, but is sensitive to array errors. For example, when there are channel gain and phase errors, the performance will decrease significantly. To handle the DOA estimation problem in the presence of channel errors, Weiss and Friedlander proposed an array self-calibration method based on joint iteration (abbreviated as the WF algorithm). The WF algorithm simultaneously realizes the signal arrival direction estimation and array calibration by alternately updating the DOA estimation and array parameters. This algorithm can obtain good estimation performance when the array amplitude and phase errors are not too large, but may fall into a local optimum when the errors are large, resulting in a decrease in the estimation performance. With the development of deep learning technology, the DOA estimation method based on neural networks has received extensive attention due to its powerful feature extraction and non-linear mapping capabilities. Among them, the fully connected deep neural network (FCDNN) has been applied to DOA estimation due to its simple and effective characteristics. However, the existing neural network-based methods mainly focus on how to improve the accuracy of DOA estimation, and rarely consider the robustness of the algorithm in different amplitude and phase error scenarios.

[0003] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of the present invention, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0004] The present invention provides a robust DOA estimation method based on a fully connected neural network, a computer program product, and an electronic device, which can maintain a high DOA estimation accuracy and have good robustness, and thus can overcome the defects existing in the prior art to a certain extent.

[0005] Other features and advantages of the present invention will become apparent from the following detailed description, or will be learned in part through the practice of the present invention.

[0006] According to a first aspect of the present invention, there is provided a robust DOA estimation method based on a fully connected neural network, the method comprising:

[0007] Receiving far-field narrowband signals by each element in a sensor array, and constructing an observation data vector based on the far-field narrowband signals received by each element; and determining a corresponding steering vector according to the arrival angle of the far-field narrowband signals;

[0008] Constructing a covariance matrix based on the observation data vector;

[0009] Taking the covariance matrix as a model output parameter, inputting it into a trained phase error estimation model based on a fully connected neural network, and obtaining a phase error estimation value output by the model;

[0010] Determining a corresponding phase error estimation matrix based on the phase error estimation value; and compensating the steering vector using the phase error estimation matrix to obtain a calibrated steering vector;

[0011] Combining the calibrated steering vector and the noise subspace to construct a MUSIC spatial spectrum function; and searching for the spectral peak of the MUSIC spatial spectrum function within an angle search range to determine the DOA estimation value of the signal source according to the spectral peak search result.

[0012] In some exemplary embodiments, constructing the observation data vector based on the far-field narrowband signals received by each element includes:

[0013] Combining the far-field narrowband signals received by each element to construct initial observation data;

[0014] Performing a transpose process on the initial observation data to obtain the observation data vector.

[0015] In some exemplary embodiments, constructing the covariance matrix based on the observation data vector includes:

[0016] Determining the covariance matrix based on the observation data vector and the conjugate transpose matrix of the observation data vector.

[0017] In some exemplary embodiments, the method further includes:

[0018] Taking the upper right corner element of the covariance matrix, arranging them to form a complex vector with a dimension of ;

[0019] Continue to splice the real part and the imaginary part of the complex vector to obtain an intermediate vector of dimension M(M - 1); configure the intermediate vector as the model input parameter.

[0020] In some exemplary embodiments, determining the covariance matrix based on the observation data vector and the conjugate transpose matrix of the observation data vector includes:

[0021] Construct the covariance matrix by combining the diagonal matrix with eigenvalues arranged in descending order, the signal subspace, the noise subspace, the eigenvector matrix, the signal eigenvalues, and the noise eigenvalues.

[0022] In some exemplary embodiments, the phase error matrix includes the phase errors corresponding to each array element; wherein, the first array element is configured as the reference array element.

[0023] In some exemplary embodiments, determining the DOA estimation value of the signal source according to the spectral peak search result includes:

[0024] Determine the DOA estimation value of the signal source according to the angle corresponding to the spectral peak.

[0025] In some exemplary embodiments, the phase error estimation model based on the fully connected neural network includes: an input layer, an intermediate layer, and an output layer; the intermediate layer includes a plurality of consecutive hidden layers and a dropout layer; wherein, each hidden layer is configured with a tanh non-linear activation function; the number of neurons in adjacent hidden layers is different.

[0026] In some exemplary embodiments, the method further includes: pre-training a phase error estimation model based on a fully connected neural network, including:

[0027] Utilize the sample far-field narrowband signals received by the sensor array in multiple different scenarios, and construct corresponding sample observation data vectors;

[0028] Based on the sample observation data vector and the actual phase error of the corresponding uncalibrated array element, construct sample data-label pairs; construct a training data set based on multiple sample data-label pairs;

[0029] Iteratively train the initial phase error estimation model using the training data set to obtain a trained phase error estimation model based on the fully connected neural network.

[0030] According to the second aspect of the present invention, there is provided a computer program product, on which a computer program is stored, and when the computer program is executed by a processor, it implements the above-mentioned robust DOA estimation method based on a fully connected neural network.

[0031] According to the third aspect of the present invention, there is provided an electronic device, including:

[0032] A processor; and

[0033] a memory for storing executable instructions of the processor;

[0034] wherein the processor is configured to implement the above-mentioned robust DOA estimation method based on a fully connected neural network when executing the executable instructions.

[0035] According to a fourth aspect of the present invention, there is provided a storage medium having stored thereon a computer program, which when executed by a processor implements the above-mentioned robust DOA estimation method based on a fully connected neural network.

[0036] The robust DOA estimation method based on a fully connected neural network provided by the embodiments of the present invention, under the condition that there are phase errors in the array channels, estimates and compensates for the phase errors through a phase error estimation model based on a fully connected neural network, and then performs MUSIC spatial spectrum search using the compensated array steering vector, thereby achieving high-precision DOA estimation. This method can still maintain a high DOA estimation accuracy when there are large phase errors in the array channels, demonstrating good robustness.

[0037] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0039] Figure 1 A schematic diagram schematically showing a robust DOA estimation method based on a fully connected neural network according to an exemplary embodiment of the present invention;

[0040] Figure 2 A schematic diagram schematically showing the flow of a robust DOA estimation method based on a fully connected neural network according to an exemplary embodiment of the present invention;

[0041] Figure 3 A schematic diagram schematically showing the spatial spectra of DOA estimation when the number of signal sources is 1, 2, and 3 respectively according to an exemplary embodiment of the present invention;

[0042] Figure 4 A schematic diagram schematically showing the curve of the RMSE of DOA estimation varying with SNR when the number of signal sources is 1 according to an exemplary embodiment of the present invention;

[0043] Figure 5 Schematic diagram showing the variation curve of RMSE of DOA estimation with SNR when the number of signal sources in the exemplary embodiment of the present invention is 2;

[0044] Figure 6 Schematic diagram showing the variation curve of RMSE of DOA estimation with SNR when the number of signal sources in the exemplary embodiment of the present invention is 3;

[0045] Figure 7 Schematic diagram showing the composition of an electronic device in the exemplary embodiment of the present invention. Detailed implementation manners

[0046] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments.

[0047] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0048] In the related art, traditional DOA estimation algorithms have the problem that their performance significantly degrades when there are array channel phase errors.

[0049] In view of the disadvantages and deficiencies of the prior art, a robust DOA estimation method based on a fully connected neural network is provided in this example embodiment. Referring to Figure 1 as shown, the method may specifically include the following steps:

[0050] Step S11, receiving far-field narrowband signals by each element in the sensor array, constructing an observation data vector based on the far-field narrowband signals received by each element; and determining a corresponding steering vector according to the arrival angle of the far-field narrowband signals;

[0051] Step S12, constructing a covariance matrix based on the observation data vector;

[0052] Step S13: Use the covariance matrix as the model output parameter, input it into the trained phase error estimation model based on a fully connected neural network, and obtain the phase error estimation value output by the model;

[0053] Step S14: Determine the corresponding phase error estimation matrix based on the phase error estimation value; and use the phase error estimation matrix to compensate the steering vector to obtain the calibrated steering vector;

[0054] Step S15: Combine the calibrated steering vector and the noise subspace to construct a MUSIC spatial spectrum function; and search for the spectral peak of the MUSIC spatial spectrum function within the angle search range to determine the DOA estimation value of the signal source according to the spectral peak search result.

[0055] Next, each step of the robust DOA estimation method based on a fully connected neural network in this exemplary embodiment will be described in more detail with reference to the accompanying drawings and embodiments.

[0056] In step S11, use each element in the sensor array to receive far-field narrowband signals, and construct an observation data vector based on the far-field narrowband signals received by each element; and determine the corresponding steering vector according to the arrival angle of the far-field narrowband signal.

[0057] Exemplarily, the above method can be executed on an intelligent terminal device or a server side. For example, the data collected by the sensor array can be sent to the intelligent terminal device; the terminal device can create a processing task for the current data and execute the task to output the DOA estimation value of the signal source.

[0058] Among them, constructing the observation data vector based on the far-field narrowband signals received by each element includes: combining the far-field narrowband signals received by each element to construct initial observation data; performing a transpose process on the initial observation data to obtain the observation data vector.

[0059] Specifically, the sensor array is a uniform linear array composed of M sensors. K far-field and narrowband signal sources are incident on the array, and the sensor array receives K far-field and narrowband signals; the incident direction of the signal source is θ, and the arrival angle of the kth far-field narrowband signal is defined as θ k , k = 1, 2,..., K; the direction finding range of the sensor array is [θ min , θ max ), where θ min and θ max are the minimum angle and the maximum angle of the direction finding range respectively.

[0060] The signal received by the element is x m (t), m = 1, 2,..., M; the observation data vector of M elements is x(t) = [x1(t), x2(t),..., xM (t)] T ; where, (·) T represents the transpose operation.

[0061] For example, assume that the arrival angles of the K far-field narrowband signals received are θ1, θ2, …, θ K , s(t) is a vector composed of K signal sources s(t) = [s1(t), s2(t), …, s K (t)] T , A = [a(θ1), a(θ2), …, a(θ K ), a(θ i ) is the steering vector of the i-th angle. The steering vector is expressed as:

[0062]

[0063] where, d m is the distance between the m-th array element and the first array element, λ is the signal wavelength, and j is the imaginary unit.

[0064] When there is a phase error, a phase error matrix is constructed according to the phase errors of each array element, which is expressed as: where, is the phase error of the i-th array element Phase error ranges from [-L, L] degrees, and L is the known phase error range.

[0065] Define the noise vector as: n(t) = [n1(t), n2(t), …, n M (t)] T , and at this time the array received signal vector is expressed as: x(t) = ΓAs(t) + n(t).

[0066] In step S12, a covariance matrix is constructed based on the observation data vector.

[0067] Exemplarily, constructing a covariance matrix based on the observation data vector includes: determining the covariance matrix based on the observation data vector and the conjugate transpose matrix of the observation data vector.

[0068] Specifically, a covariance matrix is constructed based on the observation data vectors of each array element; the specific formula can include:

[0069]

[0070] where, (·) H represents the conjugate transpose operation, N is the number of samplings; x(t) is the observation data vector.

[0071] In step S13, the covariance matrix is used as a model output parameter and input into a pre-trained phase error estimation model based on a fully connected neural network to obtain the phase error estimation value output by the model.

[0072] Exemplarily, the phase error estimation model can be pre-trained on the terminal device side; this can be constructed based on a fully connected neural network. The covariance matrix can be used as a model output parameter to obtain the phase error estimation value output by the model.

[0073] In one exemplary embodiment, the method further includes: taking the upper right corner element of the covariance matrix and arranging it to form a complex vector with a dimension of ;

[0074] Continuing to splice the real part and the imaginary part of the complex vector to obtain an intermediate vector with a dimension of M(M - 1); configuring the intermediate vector as the model input parameter.

[0075] Specifically, the upper right triangular elements of the covariance matrix can be taken and arranged to form a -dimensional vector b cor , expressed as:

[0076] b cor = [R 1,2 , R 1,3 , …, R 1,M , R 2,3 , …, R 2,M , …, R M-1,M T .

[0077] Currently, b cor is a complex vector containing a real part and an imaginary part. It is necessary to split the real part and the imaginary part and splice them to obtain an M(M - 1)-dimensional vector as the intermediate vector, which can be specifically expressed as:

[0078] b = [Real{b cor T}, Imag{b cor T}]

[0079] where Real(·) is the operation of taking the real part, and Imag(·) is the operation of taking the imaginary part.

[0080] After that, b is normalized to obtain the vector:

[0081]

[0082] where mean(·) represents taking the mean, and ‖·‖2 represents the L2 norm.

[0083] Exemplarily, the method further includes: performing eigenvalue decomposition processing on the covariance matrix to obtain: signal eigenvalues, a diagonal matrix with eigenvalues arranged in descending order, a signal subspace, a noise subspace, an eigenvector matrix, signal eigenvalues, and noise eigenvalues to construct the covariance matrix.

[0084] Specifically, for the covariance matrix eigenvalue decomposition processing can be performed, and the corresponding formula can include:

[0085]

[0086] where, Λ = diag{λ1, λ2, …, λ m} is a diagonal matrix with eigenvalues arranged in descending order, and U is the corresponding eigenvector matrix; U s is the signal subspace, composed of the eigenvectors corresponding to the first K largest eigenvalues; U n is the noise subspace, composed of the eigenvectors corresponding to the remaining M - K eigenvalues; Λ s = diag{λ1, λ2, …, λ K} are the signal eigenvalues, and Λ n = diag{λ K+1 , λ K+2 , …, λ M} are the noise eigenvalues.

[0087] In step S14, a corresponding phase error estimation matrix is determined based on the phase error estimation value; and the steering vector is compensated using the phase error estimation matrix to obtain a calibrated steering vector.

[0088] Exemplarily, the phase error matrix includes the phase errors corresponding to each array element; wherein, the first array element is configured as the reference array element.

[0089] Specifically, based on the phase error estimation value output by the current model a phase error estimation matrix can be constructed, and the formula can be expressed as:

[0090]

[0091] where, is the phase error of the m-th array element, is the reference array element.

[0092] Based on the obtained phase error estimation matrix the steering vector can be compensated to obtain a compensated steering vector, realizing the calibration of the steering vector, and the formula can be expressed as:

[0093]

[0094] In step S15, a MUSIC spatial spectrum function is constructed by combining the calibrated steering vector and the noise subspace; and the spectral peak of the MUSIC spatial spectrum function is searched within the angle search range to determine the DOA estimation value of the signal source according to the spectral peak search result.

[0095] Exemplarily, determining the DOA estimation value of the signal source according to the spectral peak search result includes: determining the DOA estimation value of the signal source according to the angle corresponding to the spectral peak.

[0096] Specifically, according to the compensated steering vector and the extracted noise subspace U n , a MUSIC spatial spectrum function is constructed, and the formula includes:

[0097]

[0098] where is the steering vector at the search angle θ after compensating for the phase error, and U n is the noise subspace.

[0099] The DOA estimation value of the signal source is obtained by searching for the spectral peak of the MUSIC spatial spectrum function within the angle search range

[0100] Exemplarily, the phase error estimation model based on the fully connected neural network includes: an input layer, an intermediate layer, and an output layer; the intermediate layer includes a plurality of consecutive hidden layers and a dropout layer; wherein, each hidden layer is configured with a tanh non-linear activation function; the number of neurons in adjacent hidden layers is different.

[0101] Specifically, the fully connected neural network is trained through a regression training task. For example, the model can be composed of an input layer, two hidden layers, and an output layer. Among them, the number of neurons in the input layer is M×(M - 1), and the number of neurons in the hidden layers is 70 and 30 respectively; there is a tanh non-linear activation function behind each hidden layer, and a dropout is added behind the second hidden layer to prevent overfitting during the training process; the number of neurons in the output layer is the same as the number of array elements with phase errors. For example, the output layer includes 2 neurons. A corrected spatial spectrum is constructed through the MUSIC algorithm, and the DOA estimation value of the target signal is obtained by searching for the spectral peak of the spatial spectrum.

[0102] Exemplarily, the method further includes: pre-training a phase error estimation model based on a fully connected neural network, including:

[0103] Step S21, using the sample far-field narrowband signals received by the sensor array in multiple different scenarios, and constructing corresponding sample observation data vectors;

[0104] Step S22: Based on the sample observation data vectors and the actual phase errors of the corresponding uncalibrated array elements, construct sample data-label pairs; construct a training data set based on multiple sample data-label pairs.

[0105] Step S23: Use the training data set to iteratively train the initial phase error estimation model to obtain a trained phase error estimation model based on a fully connected neural network.

[0106] For example, the sensor array can be a uniform linear array consisting of 6 array elements; the target signal is a far-field, narrowband signal; the signal-to-noise ratio is the power of the target signal, is the power of the noise, and training data is generated under Gaussian noise conditions; define the number of Monte-Carlo experiments as 100 times; the direction finding range is [θ min =-60°, θ max =60°).

[0107] Exemplarily, when constructing the training data set, data simulation can be carried out in a two-target scenario, phase errors and random noise are added to each set of target angles, and the data is labeled as y i , where the subscript i is the data serial number, and the corresponding output label is z i . Since the deep neural network has good non-linear mapping ability, it can effectively improve the phase error estimation accuracy in the two-target scenario. After correction combined with the MUSIC algorithm, the DOA estimation performance can be further improved.

[0108] Consider training the network using data in different signal-to-noise ratio scenarios. The signal-to-noise ratio values are [-4dB, -2dB, 0dB]. Such values can not only improve the robustness of the algorithm under low signal-to-noise ratio conditions but also generalize well to high signal-to-noise ratio scenarios. At the same time, it avoids introducing a large amount of high signal-to-noise ratio training data with low marginal benefits. The number of snapshots is 400. The two-target angle search interval is [-60°, 60°), and the set of two-target angle intervals is Δ={3°, 6°, …, 60°}. The range of the first target angle θ1∈[-60°, -60°+Δ s , where Δ s represents the selected two-target interval angle in the Δ set, and the second angle θ2 = θ1 + Δ s . Taking the first array element as the reference array element, assuming that array elements 2 and 3 are uncalibrated array elements, phase errors are applied to array elements 2 and 3 respectively, and the range of the phase error is [-40°, 40°], stepping at intervals of 2°. By calculating the covariance matrix, the upper right corner element is extracted as the input feature, and the corresponding phase error is used as the output label to form a data-label set {(y1, z1), (y2, z2), …, (y F,z F )}。

[0109] During model training, the covariance matrix samples are input into the initial model to obtain the phase error estimation value of the initial model. Calculate the difference between the phase error estimation value and the phase error label value, and based on this difference, adjust the model hyperparameters and perform backpropagation training on the model. Iteratively train until the preset number of times or the model converges to obtain the trained phase error estimation model.

[0110] During model testing, refer to Figure 2 as shown, obtain the input data that is not in the training dataset input it into the trained fully connected neural network to obtain the output phase error estimation value

[0111] Construct a phase error matrix based on the obtained phase error estimation Invert it and compensate the steering vector, and the calibrated steering vector becomes Use the calibrated steering vector and the MUSIC algorithm to obtain the spatial spectrum and perform spectral peak search to obtain K spectral peak data, and the corresponding angles are the estimated source directions.

[0112] Exemplarily, the data label corresponding to the fully connected neural network is a vector z of dimension i , where is the number of array elements, so Thus, a data-label pair (y i ,z i ) is obtained, where the subscript i is the serial number, and z i is the true phase error of the uncalibrated array element.

[0113] Obtain the observed data x(t) under different scenarios, and according to the corresponding and obtain the data-label pair (y i ,z i ), and finally obtain the training data-label set of the fully connected neural network composed of F groups of data-label pairs {(y1,z1),(y2,z2),…,(y F ,z F )}.

[0114] Use the training data-label set {(y1,z1),(y2,z2),…,(y F ,z F )} to train the fully connected neural network. Assume that the number of network layers of the fully connected neural network is L layers (excluding the input layer), and use net l to represent the output of the l-th layer, net l =W l,l-1h l-1 +b l , l = 1, 2, …, L, where W l,l-1 represents the weight matrix between the (l - 1)-th layer and the l-th layer, and b l represents the bias vector of the l-th layer, and h l-1 represents the activation value of the (l - 1)-th layer; for each intermediate layer, its output is: h l = g l (net l ), l = 1, 2, …, L - 1, where g(·) is the activation function; for the output layer z out = net L .

[0115] Exemplarily, this embodiment examines the performance of the method of the present invention (i.e., FCDNN-MUSIC), and compares it with the traditional DOA estimation methods MUSIC algorithm and WF algorithm. The performance metrics are the spatial spectrum and the root mean square error (RMSE) of DOA estimation.

[0116] This embodiment examines the DOA estimation performance of the method under the conditions of K = 1, 2, 3. Unless otherwise specified, the phase errors of antenna 2 and antenna 3 are -25.5° and 30.5° respectively, the number of snapshots N = 400, the signal-to-noise ratio SNR = 10 dB, and the spatial spectrum and RMSE of the algorithm are examined; when K = 1, there is a single target. Unless otherwise specified, the single target DOA is θ1 = -20°; when K = 2, there are two targets. Unless otherwise specified, the two target DOAs are (θ1 = -20°, θ2 = 12°); when K = 3, there are three targets. Unless otherwise specified, the three target DOAs are (θ1 = -20°, θ2 = 0°, θ3 = 12°).

[0117] It can be seen from the above experimental results the effects of the method of the present invention under different experimental conditions:

[0118] The spatial spectra of DOA estimation when K = 1, 2, 3 are as Figure 3As shown, when the number of signal sources is 1, both the method of the present invention (FCDNN-MUSIC) and the MUSIC algorithm can estimate the DOA of the signal source well. Since the WF algorithm needs to use the steering vectors of multiple signal sources to construct a reversible optimization matrix Q to estimate the phase error during the calibration process, this matrix Q is irreversible in the single-signal-source scenario, resulting in the failure of the WF algorithm. Therefore, the spatial spectrum shows severe distortion. When the number of signal sources is 2, all three algorithms can effectively estimate the signal source direction, and both the method of the present invention and the WF algorithm show narrower spectral peaks and lower sidelobe levels. When the number of signal sources is 3, the angle resolution ability of the traditional MUSIC algorithm significantly decreases, while the method of the present invention and the WF algorithm still maintain good resolution performance and can clearly distinguish the directions of the three signal sources. It should be noted that the WF algorithm has an estimation deviation at θ1 = -20° and θ2 = 0°, while the method of the present invention can accurately estimate the angles of the three targets.

[0119] When K = 1, in the case of a single target, as Figure 4 shown, the RMSE of both the method of the present invention and the MUSIC algorithm is less than 1°, and the RMSE of the method of the present invention is slightly lower than that of the MUSIC algorithm, showing higher estimation accuracy. The WF fails in the single-signal-source scenario, resulting in its RMSE remaining at about 15°.

[0120] When K = 2, in the case of two targets, as Figure 5 shown, the RMSE of all three algorithms shows a downward trend as the SNR increases, and the RMSE of the method of the present invention is always the lowest.

[0121] When K = 3, in the case of three targets, as Figure 6 shown, under the condition of low signal-to-noise ratio (-5 dB), the RMSE of the method of the present invention is about 2.5°, which is close to the RMSE of the WF algorithm. As the SNR increases, the performance of all three algorithms improves, but the improvement degrees are different: the RMSE of the method of the present invention rapidly decreases and stabilizes below 0.2° after SNR > 5 dB, the RMSE of the WF algorithm stabilizes at about 1.2°, and the performance of the MUSIC algorithm improves, with its RMSE decreasing from 9° to about 0.8° and then reaching stability.

[0122] The method provided by the embodiments of the present invention estimates the phase error of the array through a fully connected neural network FCDNN and compensates it, and then obtains an accurate DOA estimation value according to the MUSIC method, and can still maintain a high estimation accuracy under different phase error scenarios. The present invention aims to improve the DOA estimation accuracy and robustness of the algorithm at the same time, and provides a more reliable solution for the DOA estimation problem in practical applications. Under the condition that there is a phase error in the array channel, this method estimates and compensates the phase error through a fully connected neural network, and then uses the compensated array steering vector to perform MUSIC spatial spectrum search, so as to achieve high-precision DOA estimation. This method can still maintain a high DOA estimation accuracy in the case of a large phase error in the array channel, showing good robustness.

[0123] It should be noted that the above-mentioned drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present invention, rather than for limiting purposes. It is easy to understand that the processes shown in the above-mentioned drawings do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be executed synchronously or asynchronously in, for example, multiple modules.

[0124] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present invention, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0125] Figure 7 The schematic diagram of the electronic device suitable for implementing the embodiments of the present invention is shown.

[0126] It should be noted that Figure 7 The shown electronic device 1000 is only an example, and should not bring any limitation to the functions and usage scope of the embodiments of the present invention.

[0127] Such as Figure 7As shown, the electronic device 1000 includes a Central Processing Unit (CPU) 1001, which can perform various appropriate actions and processes according to the program stored in the Read-Only Memory (ROM) 1002 or the program loaded from the storage section 1008 into the Random Access Memory (RAM) 1003. In the RAM 1003, various programs and data required for system operation are also stored. The CPU 1001, ROM 1002, and RAM 1003 are connected to each other via a bus 1004. An Input / Output (I / O) interface 1005 is also connected to the bus 1004.

[0128] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including, for example, a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc. and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed so that a computer program read from it can be installed into the storage section 1008 as needed.

[0129] In particular, according to an embodiment of the present invention, the processes described below with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a storage medium, and the computer program contains program codes for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from the network via the communication section 1009, and / or installed from the removable medium 1011. When the computer program is executed by the Central Processing Unit (CPU) 1001, various functions defined in the system of the present application are executed.

[0130] It should be noted that the storage medium shown in the embodiments of the present invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of a computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium can also be any storage medium other than a computer-readable storage medium, and this storage medium can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the storage medium can be transmitted by any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0131] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, and the above module, program segment, or part of code contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks can occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0132] The units involved in the embodiments of the present invention can be implemented in software or in hardware, and the described units can also be provided in a processor. Among them, the names of these units do not constitute a limitation to the units themselves in some cases.

[0133] It should be noted that, on the other hand, the present application also provides a storage medium, which can be included in an electronic device; or can exist alone without being assembled into the electronic device. The above storage medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device is caused to implement the methods described in the following embodiments. For example, the described electronic device can implement each step of the method as Figure 1 shown.

[0134] In one embodiment, the present application provides a computer program product, including a computer program, which when executed by a processor implements the steps in the above method embodiments.

[0135] In addition, the above drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present invention, rather than for limiting purposes. It is easy to understand that the processes shown in the above drawings do not indicate or limit the chronological order of these processes. Additionally, it is also easy to understand that these processes can be executed synchronously or asynchronously in, for example, multiple modules.

[0136] Those skilled in the art will readily think of other embodiments of the present invention after considering the specification and practicing the invention herein. The present application is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include the common general knowledge or conventional technical means in the technical field not disclosed in the present invention. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present invention are pointed out by the claims.

[0137] It should be understood that the present invention is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.

Claims

1. A robust DOA estimation method based on a fully connected neural network, characterized in that, The method includes: Receiving far - field narrow - band signals by each element in a sensor array, and constructing an observation data vector based on the far - field narrow - band signals received by each element; and determining a corresponding steering vector according to the arrival angle of the far - field narrow - band signals; Constructing a covariance matrix based on the observation data vector; Taking the covariance matrix as a model output parameter, inputting it into a trained phase error estimation model based on a fully - connected neural network, and obtaining the phase error estimation value output by the model; Determining a corresponding phase error estimation matrix based on the phase error estimation value; and compensating the steering vector using the phase error estimation matrix to obtain a calibrated steering vector; Combining the calibrated steering vector and the noise subspace to construct a MUSIC spatial spectrum function; and searching for the spectral peak of the MUSIC spatial spectrum function within the angle search range to determine the DOA estimation value of the signal source according to the spectral peak search result.

2. The method according to claim 1, characterized in that, Constructing the observation data vector based on the far - field narrow - band signals received by each element includes: Combining the far - field narrow - band signals received by each element to construct initial observation data; Performing a transpose operation on the initial observation data to obtain the observation data vector.

3. The method according to claim 1, wherein Constructing the covariance matrix based on the observation data vector includes: Determining the covariance matrix based on the observation data vector and the conjugate transpose matrix of the observation data vector.

4. The method according to claim 1 or 3, characterized in that, The method further includes: Take the upper right corner element of the covariance matrix and arrange them to form a complex vector with a dimension of ; Continuing to splice the real part and the imaginary part of the complex vector to obtain an M(M - 1) - dimensional intermediate vector; configuring the intermediate vector as the model input parameter.

5. The method according to claim 1, wherein The method further includes: performing eigenvalue decomposition on the covariance matrix to obtain: signal eigenvalues, a diagonal matrix with eigenvalues arranged in descending order, a signal subspace, a noise subspace, an eigenvector matrix, signal eigenvalues, and noise eigenvalues to construct the covariance matrix.

6. The method according to claim 1, characterized in that The phase error matrix includes the phase errors corresponding to each element; where the first element is configured as a reference element.

7. The method according to claim 1, characterized in that, Determining the DOA estimation value of the signal source according to the spectral peak search result includes: Determining the DOA estimation value of the signal source according to the angle corresponding to the spectral peak.

8. The method according to claim 1, wherein The phase error estimation model based on a fully - connected neural network includes: an input layer, an intermediate layer, and an output layer; the intermediate layer includes a continuous plurality of hidden layers and a dropout layer; where each hidden layer is configured with a tanh non - linear activation function; the number of neurons in adjacent hidden layers is different.

9. The method according to claim 1 or 8, characterized in that, The method further includes: pre - training a phase error estimation model based on a fully - connected neural network, including: Using the sample far - field narrow - band signals received by the sensor array in multiple different scenarios, and constructing corresponding sample observation data vectors; Based on the sample observation data vectors and the actual phase errors of the corresponding uncalibrated elements, constructing sample data - label pairs; constructing a training data set based on multiple sample data - label pairs; Using the training data set to iteratively train an initial phase error estimation model to obtain a trained phase error estimation model based on a fully - connected neural network.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the robust DOA estimation method based on a fully - connected neural network according to any one of claims 1 to 9.

Images