Non-cooperative target beam signal arrival angle estimation method and device, medium and equipment

Abstract

The invention discloses a non-cooperative target beam signal arrival angle estimation method, a non-cooperative target beam signal arrival angle estimation device, a medium and equipment. The method comprises the following steps: acquiring a to-be-measured beam signal sent by a non-cooperative target; inputting the obtained to-be-measured beam signal into a pre-trained BD-DOA network to obtain an estimation parameter of the to-be-measured beam signal; calculating the loss between the estimation parameter of the to-be-measured beam signal and the estimation parameter of the sample incident signal of each region in the space, and outputting the angle of the region where the sample incident signal is located when the loss is minimum as the estimation angle of the arrival angle of the to-be-measured beam signal; wherein the estimation parameter of the sample incident signal of each area in the space is obtained by inputting the sample incident signal of each area in the space into a pre-trained BD-DOA network. According to the invention, the direction of the beam signal sent by the non-cooperative target can be accurately obtained.

Description

Technical Field

[0001] The invention relates to a method, a device, a medium and equipment for estimating the arrival angle of a non-cooperative target beam signal, and belongs to the technical field of photovoltaic power prediction. Background Art

[0002] Traditional spectrum sensing mainly detects omnidirectional signals. With the rise of multi-antenna and multiple-input multiple-output (MIMO) technology, a large number of spatial signals are generated through beamforming technology. These numerous beamforming signals have led to the increasing complexity of today's electromagnetic environment. Researchers' in-depth exploration of beamforming signals has expanded the focus of traditional spectrum sensing to the acquisition of multi-dimensional spatial information. This includes spectrum occupancy, spectrum radiation power, spectrum modulation schemes, spectrum access protocols, and detection of abnormal spectrum behavior. Therefore, the next generation of spectrum sensing technology has shifted from detecting omnidirectional signals to detecting directional signals, namely spatial spectrum sensing technology.

[0003] Beamforming technology has been widely used in various fields, effectively solving the problem of scarce spectrum resources. With the rapid development of the fifth generation of mobile communication technology (5G), beamforming signals can significantly increase the communication rate. In remote areas, the use of beamforming technology combined with airborne relays effectively ensures the quality of communication. In recent years, the emergence of reconfigurable smart surfaces has further expanded the application of beamforming technology. Radar also uses beamforming to detect unknown information.

[0004] However, the proliferation of beamforming signals has put forward more stringent requirements on spatial spectrum sensing technology. Detecting the presence of signals is the basis of spatial spectrum sensing, and determining the directional characteristics of target beamforming signals is another important part of this process. Therefore, in spatial spectrum sensing, it is necessary not only to sense the status of spectrum resources but also the angular information of the sensed signals. The first task of the perception stage is to detect the presence of target beamforming signals, while the localization stage involves estimating the directional information of these target beamforming signals.

[0005] The research on spatial spectrum sensing considers the detection and positioning stages as independent. In the prior art, researchers usually focus on the detection stage or the positioning stage. However, the positioning stage in spatial spectrum sensing is based on the results of the detection stage, and the performance of the detection stage significantly affects the estimation results of the positioning stage. At present, many researchers assume that the target signal has been accurately detected when studying the positioning stage, but this assumption is usually too idealistic. First, if there are errors in signal detection, the estimation results of the positioning stage will inevitably be affected; second, even if the signal detection is accurate, incomplete information about the target signal will significantly increase the difficulty of the positioning stage, which is particularly evident when using deep learning methods for beam estimation tasks. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method, device, medium and equipment for estimating the arrival angle of a non-cooperative target beam signal, which can accurately obtain the direction of the beam signal sent by the non-cooperative target. To achieve the above purpose, the present invention is implemented by adopting the following technical solutions:

[0007] In a first aspect, the present invention provides a method for estimating an arrival angle of a non-cooperative target beam signal, comprising:

[0008] Acquire the beam signal to be measured sent by the non-cooperative target;

[0009] The acquired beam signal to be measured is input into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured;

[0010] The loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space is calculated, and the angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

[0011] In combination with the first aspect, optionally, the pre-trained BD-DOA network is trained by the following steps:

[0012] Adding angle labels to the sample incident signals in each area in the space to obtain sample incident signals with angle labels; the angle label is the angle of the area where the sample incident signal is located;

[0013] The pre-built BD-DOANet network is trained using sample incident signals with angle labels, including:

[0014] Initialize the network parameters of the pre-built BD-DOANet network;

[0015] The sample incident signal with angle labels is input into the pre-built BD-DOANet network for forward propagation to obtain the estimated angle labels, and the loss function is used to calculate the loss between the estimated angle labels and the angle labels.

[0016] With the goal of minimizing the calculated loss, the network parameters of the BD-DOANet network are updated through back propagation;

[0017] In response to reaching a preset number of iterations or the calculated loss reaching a preset threshold, the weights and biases of the BD-DOANet network are output to obtain a pre-trained BD-DOA network.

[0018] In combination with the first aspect, optionally, the loss between the estimated angle label and the angle label is calculated using a loss function, which is expressed by the following formula:

[0019] ,

[0020] in, For the The estimated angle labels for the data, For the The angle label of the data, is the number of channels.

[0021] In combination with the first aspect, optionally, the pre-built BD-DOANet network includes a convolutional layer, a channel attention unit, two residual units, a fully connected layer and an activation function.

[0022] In combination with the first aspect, optionally, the convolution layer includes a convolution kernel and a convolution layer nonlinear activation function,

[0023] The input signal is input into the convolution kernel for feature extraction, and then the convolution layer nonlinear activation function is applied to obtain the first feature map, which is expressed by the following formula:

[0024] ,

[0025] in, is the first feature map, is the input signal, is the total number of convolution kernels, is the convolution kernel, is the offset, It is the operation of the nonlinear activation function of the convolution layer.

[0026] In combination with the first aspect, optionally, the residual unit includes a residual convolution layer, a batch normalization layer and a residual unit activation function,

[0027] The second feature map obtained by the channel attention unit is input into the residual convolution layer for feature extraction, and the obtained feature map is input into the batch normalization layer for standardization to obtain standardized data;

[0028] The standardization process is to calculate the mean and variance, which is expressed by the following formula:

[0029] ,

[0030] ,

[0031] ,

[0032] in, is the mean, is the size of the small batch of data in the second feature map, is a small batch of data in the second feature map of the input, is the serial number of the small batch of data in the second feature map; is the variance; To standardize the data, is a positive constant;

[0033] The standardized data is represented using learnable scaling parameters and offset parameters to obtain a standardized feature map, which is expressed as follows:

[0034] ,

[0035] in, is the standardized feature map, is the scaling parameter, is the offset parameter;

[0036] Applying the residual unit activation function to process the normalized feature map obtains the third feature map, which is expressed by the following formula:

[0037] ,

[0038] in, is the third characteristic graph, The operation of the residual unit activation function to take the maximum value.

[0039] In combination with the first aspect, optionally, the region where the sample incident signal is located is a region obtained by evenly dividing the space according to a step size of a preset angle, and the angle of the region is an angle range when dividing the region.

[0040] In a second aspect, the present invention provides a device for estimating an arrival angle of a non-cooperative target beam signal, comprising:

[0041] Acquisition module: used to acquire the beam signal to be measured sent by the non-cooperative target;

[0042] Estimation module: used to input the acquired beam signal to be measured into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured;

[0043] Output module: used to calculate the loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space. The angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

[0044] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the steps of the method for estimating the arrival angle of a non-cooperative target beam signal described in the first aspect.

[0045] In a fourth aspect, the present invention provides a computer device, characterized in that it includes:

[0046] Memory, for storing computer programs / instructions;

[0047] A processor is used to execute the computer program / instructions to implement the step of estimating the arrival angle of the non-cooperative target beam signal as described in the first aspect.

[0048] Compared with the prior art, the method, device, medium and equipment for estimating the arrival angle of a non-cooperative target beam signal provided by the embodiments of the present invention have the following beneficial effects:

[0049] The present invention obtains a beam signal to be measured sent by a non-cooperative target; inputs the obtained beam signal to be measured into a pre-trained BD-DOA network to obtain estimated parameters of the beam signal to be measured; calculates the loss between the estimated parameters of the beam signal to be measured and the estimated parameters of sample incident signals in various regions in the space, and outputs the angle of the region where the sample incident signal is located when the loss is minimum as the estimated angle of arrival of the beam signal to be measured; the estimated parameters of the sample incident signals in various regions in the space are obtained by inputting the sample incident signals in various regions in the space into a pre-trained BD-DOA network; the present invention can accurately obtain the direction of the beam signal sent by the non-cooperative target; the method for estimating the arrival angle of the non-cooperative target beam signal provided by the present invention can achieve better performance under various signal-to-noise ratio environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flowchart of a method for estimating the arrival angle of a non-cooperative target beam signal in Embodiment 1 of the present invention;

[0051] Figure 2 It is a schematic diagram of a use scenario of a method for estimating an arrival angle of a non-cooperative target beam signal in Embodiment 1 of the present invention;

[0052] Figure 3 It is a schematic diagram of the structure of a BD-DOANet network pre-constructed in a method for estimating the arrival angle of a non-cooperative target beam signal in Example 1 of the present invention;

[0053] Figure 4 is a schematic diagram showing that the sensing node in Embodiment 2 of the present invention completely captures the real signal;

[0054] Figure 5is a schematic diagram of a sensing node capturing a real signal at the beginning stage in Embodiment 2 of the present invention;

[0055] Figure 6 is a schematic diagram of a sensing node capturing a real signal at the end stage in Embodiment 2 of the present invention;

[0056] Figure 7 It is a schematic diagram showing that the sensing node in Embodiment 2 of the present invention completely captures the real signal, but the sensing period exceeds the total duration of the UAV beamforming signal;

[0057] Figure 8 is the accuracy curve of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in an ideal environment in Example 2 of the present invention;

[0058] Fig. 9 is a curve showing the change of mesh errors of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in an ideal environment in Example 2 of the present invention;

[0059] Fig.10 is a curve showing a change in mean square error of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in an ideal environment in Example 2 of the present invention;

[0060] Fig.11 is an accuracy curve of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in Example 2 of the present invention under a non-ideal environment;

[0061] Fig.12 is a curve showing the change of mesh errors of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in Example 2 of the present invention under a non-ideal environment;

[0062] Fig.13 It is a curve showing the variation of the mean square error of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in Example 2 of the present invention under a non-ideal environment. DETAILED DESCRIPTION

[0063] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0064] Embodiment 1:

[0065] like Figure 1 As shown, this embodiment provides a method for estimating the arrival angle of a non-cooperative target beam signal, including:

[0066] Acquire the beam signal to be measured sent by the non-cooperative target;

[0067] The acquired beam signal to be measured is input into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured;

[0068] The loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space is calculated, and the angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

[0069] The specific implementation steps are as follows.

[0070] Step 1: Describe the usage scenario and establish the problem of estimating the arrival angle of the non-cooperative target beam signal.

[0071] like Figure 2 As shown, the target ( Figure 2 When a target transmitter (in the middle) sends a beam signal to serve users in a certain area, the sensing node detects the beam signal. Figure 1 The illustrated method for estimating the angle of arrival of a non-cooperative target beam signal estimates the direction of the angle of arrival of the beam signal.

[0072] Step 1.1: Initialize the target antenna array as a linear array, then the beamforming angle is , the angle range is . If the path loss of all antennas of the target is initialized to be the same, then the path loss of the first antenna from the target The transmission signal of the antenna It is expressed as:

[0073] ,

[0074] in, is the transmission power, is the signal transmission distance, is the wavelength, and Is plural.

[0075] Step 1.2: A The sensing node composed of linear omnidirectional antenna array antenna receives a signal source , where the signals come from different directions. Describes the direction of the beam signal of the target in space. The sensing node is in the snapshot Output data at time It is expressed as:

[0076] ,

[0077] in, To transmit the signal, For noise.

[0078] Step 1.3: Spatial covariance of the beam signal sent by the non-cooperative target It is expressed as:

[0079] ,

[0080] in, For expectations, For the receiver to receive the signal, is the received signal angle, is the channel matrix expressed in angles, is the noise covariance, is the identity matrix. is the direction of the non-cooperative signal in the grid, .

[0081] Step 1.4: Simplify the expression of the spatial covariance of the equation to obtain:

[0082] ,

[0083] in, is a The vector of element is 1, and all other elements are 0; ; is the spatial spectrum, , which has non-zero elements only at actual angles. is a mathematical symbol, is the noise covariance.

[0084] Step 1.5: Define another new vector , expressed by the following formula:

[0085] ,

[0086] in, , .

[0087] Step 1.6: In a non-cooperative scenario, use snapshots to estimate the covariance matrix ,get:

[0088] ,

[0089] in, is the number of observed samples, is the nth observation sample.

[0090] Therefore, the estimated measurement vector It is expressed as:

[0091] ,

[0092] in, is the estimation error, .

[0093] Step 2: Obtain sample incident signals in various regions in the space.

[0094] Step 3: Build and train the network to obtain the pre-trained BD-DOA network.

[0095] like Figure 2 As shown, the area where the sample incident signal is located is an area obtained by evenly dividing the space according to a step size of a preset angle, and the angle of the area is the angle range when the area is divided.

[0096] In this embodiment, in order to effectively improve the accuracy of the estimation of the arrival angle, the spatial range is 0-180°, the step size of the preset angle is 0.5°, and the space is evenly divided into 361 areas.

[0097] This embodiment can control the average network error to 0.25.

[0098] Step 3.1: Build the BD-DOANet network.

[0099] like Figure 3 As shown in Figure 2, the pre-built BD-DOANet network consists of a convolutional layer, a channel attention unit, two residual units, a fully connected layer, and activation functions.

[0100] The convolution layer includes a convolution kernel and a convolution layer nonlinear activation function. The input signal is input into the convolution kernel for feature extraction, and then the convolution layer nonlinear activation function is applied to obtain the first feature map, which is expressed by the following formula:

[0101] ,

[0102] in, is the first feature map, is the input signal, is the total number of convolution kernels, is the convolution kernel, is the offset, It is the operation of the nonlinear activation function of the convolution layer.

[0103] In this embodiment, the nonlinear activation function of the convolutional layer adopts the ReLU activation function.

[0104] The channel attention unit is a unit used to improve the performance of deep neural networks. By learning the importance weight of each channel, the response of each channel in the feature map is dynamically adjusted to improve the feature representation ability. The channel attention unit includes a mechanism that can effectively capture and utilize the relationship between different channels, thereby enhancing the performance of the network in the task.

[0105] Residual units are commonly used in deep neural networks to solve the gradient vanishing and gradient exploding problems in deep network training. By introducing skip connections, information is allowed to pass directly through the network, effectively alleviating the gradient vanishing problem. This enables deeper networks to learn more complex features. Residual units can significantly improve the training effect and performance of the network.

[0106] In this embodiment, the residual unit includes a residual convolution layer, a batch normalization layer and a residual unit activation function.

[0107] The second feature map obtained by the channel attention unit is input into the residual convolution layer for feature extraction, and the obtained feature map is input into the batch normalization layer for standardization to obtain standardized data.

[0108] Standardized data can reduce changes in the input data distribution, accelerate the convergence process of the network and increase the training speed.

[0109] Specifically, the standardization process is to calculate the mean and variance, which is expressed by the following formula:

[0110] ,

[0111] ,

[0112] ,

[0113] in, is the mean, is the size of the small batch of data in the second feature map, is a small batch of data in the second feature map of the input, is the serial number of the small batch of data in the second feature map; is the variance; To standardize the data, is a positive constant.

[0114] The standardized data is represented using learnable scaling parameters and offset parameters to obtain a standardized feature map, which is expressed as follows:

[0115] ,

[0116] in, is the standardized feature map, is the scaling parameter, is the offset parameter;

[0117] Applying the residual unit activation function to process the normalized feature map obtains the third feature map, which is expressed by the following formula:

[0118] ,

[0119] in, is the third characteristic graph, The operation of the residual unit activation function to take the maximum value.

[0120] In this embodiment, the residual unit activation function uses an improved ReLU activation function.

[0121] Step 3.2: Train the pre-built BD-DOANet network.

[0122] According to the arrival angle information and spatial features of the sample incident signal, the training overhead is reduced. In this embodiment, the channel characteristics of the signal are captured during the training phase. The usage scenario is considered constant, and the training samples are sample incident signals in various regions in the space, which are composed of data blocks in multiple regions, so that the estimated error of the arrival angle can be tracked in the data block with the richest features.

[0123] Specifically, an angle label is added to the sample incident signal of each area in the space to obtain the sample incident signal with the angle label. The angle label is the angle of the area where the sample incident signal is located.

[0124] The pre-built BD-DOANet network is trained using sample incident signals with angle labels, including:

[0125] Step 3.2.1: Initialize the network parameters of the pre-built BD-DOANet network.

[0126] Step 3.2.2: Input the sample incident signal with angle label into the pre-built BD-DOANet network for forward propagation to obtain the estimated angle label, and use the loss function to calculate the loss between the estimated angle label and the angle label.

[0127] The loss function is used to calculate the loss between the estimated angle label and the angle label, which is expressed as follows:

[0128] ,

[0129] in, For the The estimated angle labels for the data, For the The angle label of the data, is the number of channels.

[0130] Step 3.2.3: With the goal of minimizing the calculated loss, update the network parameters of the BD-DOANet network through back propagation.

[0131] Step 3.2.4: In response to reaching a preset number of iterations or the calculated loss reaching a preset threshold, the weights and biases of the BD-DOANet network are output to obtain a trained BD-DOA network.

[0132] The sample incident signals of each region in the space are input into the pre-trained BD-DOA network to obtain the estimated parameters of the sample incident signals of each region in the space.

[0133] Step 4: Acquire the beam signal to be measured sent by the non-cooperative target.

[0134] Step 5: Input the acquired beam signal to be measured into the pre-trained BD-DOA network obtained in step 3 to obtain the estimated parameters of the beam signal to be measured.

[0135] Step 6: Calculate the loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space. The angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of arrival of the beam signal to be measured.

[0136] This embodiment can accurately obtain the direction of the beam signal sent by the non-cooperative target. The method for estimating the arrival angle of the beam signal of the non-cooperative target provided by the present invention can achieve better estimation accuracy in various signal-to-noise ratio environments.

[0137] Embodiment 2:

[0138] This embodiment evaluates the BD-DOA network in the method for estimating the angle of arrival of a non-cooperative target beam signal in Embodiment 1.

[0139] The beam signals sent by the non-cooperative targets acquired by the sensing nodes include ideal environments and non-ideal environments.

[0140] like Figure 4 As shown, during the entire sensing cycle, the sensing node completely captured the real signal, which is an ideal environment.

[0141] like Figure 5 As shown in the figure, during the entire sensing cycle, the sensing node captures a large number of real signals and a small number of noise signals. The real signal (shown in the box) is captured at the beginning of the non-cooperative UAV beamforming signal, which is a non-ideal environment.

[0142] like Figure 6As shown in the figure, during the entire sensing cycle, the sensing node captured a large number of real signals and a small number of noise signals. The real signal (shown in the box) was captured at the end of the non-cooperative UAV beamforming signal, which is a non-ideal environment.

[0143] like Figure 7 As shown in the figure, during the entire sensing cycle, the sensing node captures a large amount of real signals and a small amount of noise signals. Although the sensing node captures the complete real signal (shown in the box), the sensing cycle exceeds the total duration of the drone beamforming signal, which is a non-ideal environment.

[0144] This embodiment uses mean error and mean square error for evaluation. Mean error emphasizes the average deviation between the predicted value and the actual value, providing a direct measure to help understand the overall accuracy of the prediction. Mean square error strictly measures the size of the prediction error by averaging the squared difference, thereby emphasizing the impact of large errors. Mean error and mean square error play different but complementary roles together for evaluating data models.

[0145] The mean error and mean square error consist of two parts: the predicted label error and the grid error.

[0146] Mean Error The calculation formula is as follows:

[0147] ,

[0148] Mean square error The calculation formula is as follows:

[0149] .

[0150] in, is the sample size, To estimate the angle, is the real angle, is the estimated angle label, is the actual angle label, is the grid error.

[0151] The grid error depends on the grid spacing, and the best way to reduce the grid error is to increase the division of angle category labels. For example, in Example 1, a step size of 0.5° is used to divide the angle range of 0-180° into 361 angle intervals, thereby controlling the average network error to 0.25.

[0152] This embodiment uses a test set under an ideal environment to evaluate the BD-DOA network (BD-DOANet) in the method for estimating the angle of arrival of a non-cooperative target beam signal in Embodiment 1.

[0153] The other networks for comparison are convolutional neural network (CNN), deep convolutional neural network (DCN) and deep neural network (DNN).

[0154] like Figure 8 The figure shows the accuracy (ACC) curves of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios under an ideal environment. The predicted label error is closely related to the accuracy of the network estimation. Figure 8 , the BD-DOA network in the method for estimating the angle of arrival of a non-cooperative target beam signal in Example 1 can achieve better accuracy in various signal-to-noise ratio environments. Although the performance of the deep neural network (DNN) DNN is close to that of the BD-DOA network in Example 1 in a high signal-to-noise ratio environment, its performance is significantly inferior to that of the BD-DOA network in Example 1 in a low signal-to-noise ratio environment.

[0155] like Fig. 9 The figure shows the change curve of mean error (ME) of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios under an ideal environment. The BD-DOA network in Example 1 has the smallest mean error (ME) under various signal-to-noise ratio environments and has significantly superior performance.

[0156] Fig.10 The figure shows the change curve of the mean square error (MSE) of the BD-DOA network provided in Example 1 and other networks under different signal-to-noise ratios under an ideal environment. The BD-DOA network in Example 1 has the smallest mean square error (MSE) under various signal-to-noise ratio environments and has significantly superior performance.

[0157] This embodiment uses a test set in a non-ideal environment to evaluate the BD-DOA network (BD-DOANet) in the method for estimating the angle of arrival of a non-cooperative target beam signal in Embodiment 1.

[0158] Other networks for comparison are a BD-DOA network in a method for estimating the angle of arrival of a non-cooperative target beam signal in an embodiment, obtained by removing the channel attention unit (BD-DOANet without CBU), a deep convolutional neural network (DCN), and a deep neural network (DNN).

[0159] like Fig.11 The figure shows the accuracy (ACC) curves of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in a non-ideal environment. After removing the channel attention unit, the accuracy performance of the BD-DOA network in Example 1 is significantly reduced. However, even without the channel attention unit, the accuracy of the BD-DOA network in Example 1 is still better than that of other networks.

[0160] like Fig.12 The figure shows the change curve of mean error (ME) of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in a non-ideal environment. The BD-DOA network in Example 1 has the smallest mean error (ME) under various signal-to-noise ratio environments and has significantly superior performance.

[0161] like Fig.13 The figure shows the change curve of the mean square error (MSE) of the BD-DOA network and other networks provided in Example 1 under different signal-to-noise ratios in a non-ideal environment. The BD-DOA network in Example 1 has the smallest mean square error (MSE) under various signal-to-noise ratio environments and has significantly superior performance.

[0162] In summary, the method for estimating the arrival angle of a non-cooperative target beam signal provided in Example 1 can achieve better performance in various signal-to-noise ratio environments.

[0163] Embodiment 3:

[0164] This embodiment provides a device for estimating an arrival angle of a non-cooperative target beam signal, including:

[0165] Acquisition module: used to acquire the beam signal to be measured sent by the non-cooperative target;

[0166] Estimation module: used to input the acquired beam signal to be measured into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured;

[0167] Output module: used to calculate the loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space. The angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

[0168] Embodiment 4:

[0169] This embodiment provides a computer-readable storage medium having a computer program / instruction stored thereon, characterized in that when the computer program / instruction is executed by a processor, the steps of the method for estimating the arrival angle of a non-cooperative target beam signal described in Embodiment 1 are implemented.

[0170] Embodiment 5:

[0171] This embodiment provides a computer device, characterized by comprising:

[0172] Memory, for storing computer programs / instructions;

[0173] A processor is used to execute the computer program / instructions to implement the steps of the method for estimating the arrival angle of the non-cooperative target beam signal described in Example 1.

[0174] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0175] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0176] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0177] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0178] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A method for estimating the angle of arrival of a non-cooperative target beam signal, characterized in that: include: Acquire the beam signal to be measured sent by the non-cooperative target; The acquired beam signal to be measured is input into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured; The loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space is calculated, and the angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

2. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 1, characterized in that: The pre-trained BD-DOA network is trained by the following steps: Adding angle labels to the sample incident signals in each area in the space to obtain sample incident signals with angle labels; the angle label is the angle of the area where the sample incident signal is located; The pre-built BD-DOANet network is trained using sample incident signals with angle labels, including: Initialize the network parameters of the pre-built BD-DOANet network; The sample incident signal with angle labels is input into the pre-built BD-DOANet network for forward propagation to obtain the estimated angle labels, and the loss function is used to calculate the loss between the estimated angle labels and the angle labels. With the goal of minimizing the calculated loss, the network parameters of the BD-DOANet network are updated through back propagation; In response to reaching a preset number of iterations or the calculated loss reaching a preset threshold, the weights and biases of the BD-DOANet network are output to obtain a pre-trained BD-DOA network.

3. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 2, characterized in that: The loss function is used to calculate the loss between the estimated angle label and the angle label, which is expressed by the following formula: ， in, For the The estimated angle labels for the data, For the The angle label of the data, is the number of channels.

4. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 2, characterized in that: The pre-built BD-DOANet network includes a convolutional layer, a channel attention unit, two residual units, a fully connected layer, and activation functions.

5. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 4, characterized in that: The convolution layer includes a convolution kernel and a convolution layer nonlinear activation function. The input signal is input into the convolution kernel for feature extraction, and then the convolution layer nonlinear activation function is applied to obtain the first feature map, which is expressed by the following formula: ， in, is the first feature map, is the input signal, is the total number of convolution kernels, is the convolution kernel, is the offset, It is the operation of the nonlinear activation function of the convolution layer.

6. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 4, characterized in that: The residual unit includes a residual convolution layer, a batch normalization layer and a residual unit activation function, The second feature map obtained by the channel attention unit is input into the residual convolution layer for feature extraction, and the obtained feature map is input into the batch normalization layer for standardization to obtain standardized data; The standardization process is to calculate the mean and variance, which is expressed by the following formula: ， ， ， in, is the mean, is the size of the small batch of data in the second feature map, is a small batch of data in the second feature map of the input, is the serial number of the small batch of data in the second feature map; is the variance; To standardize the data, is a positive constant; The standardized data is represented using learnable scaling parameters and offset parameters to obtain a standardized feature map, which is expressed as follows: ， in, is the standardized feature map, is the scaling parameter, is the offset parameter; Applying the residual unit activation function to process the normalized feature map obtains the third feature map, which is expressed by the following formula: ， in, is the third characteristic graph, The operation of the residual unit activation function to take the maximum value.

7. The method for estimating the arrival angle of a non-cooperative target beam signal according to claim 1, characterized in that: The region where the sample incident signal is located is a region obtained by evenly dividing the space according to a step size of a preset angle, and the angle of the region is the angle range when the region is divided.

8. A device for estimating the angle of arrival of a non-cooperative target beam signal, characterized in that: include: Acquisition module: used to acquire the beam signal to be measured sent by the non-cooperative target; Estimation module: used to input the acquired beam signal to be measured into the pre-trained BD-DOA network to obtain the estimated parameters of the beam signal to be measured; Output module: used to calculate the loss between the estimated parameters of the beam signal to be measured and the estimated parameters of the sample incident signals in each area in the space. The angle of the area where the sample incident signal is located when the output loss is minimum is the estimated angle of the arrival angle of the beam signal to be measured; wherein the estimated parameters of the sample incident signals in each area in the space are obtained by inputting the sample incident signals in each area in the space into a pre-trained BD-DOA network.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for estimating the arrival angle of a non-cooperative target beam signal described in any one of claims 1-7 are implemented.

10. A computer device, characterized in that: include: Memory, for storing computer programs / instructions; A processor, configured to execute the computer program / instructions to implement the steps of the method for estimating the angle of arrival of a non-cooperative target beam signal according to any one of claims 1 to 7.

Images