Electromagnetic signal interference source positioning method, system and device based on neural network and medium

Abstract

The invention discloses an electromagnetic signal interference source positioning method, system and device based on a neural network and a medium, and the method comprises the steps: collecting original electromagnetic signals in a preset time window through a plurality of electromagnetic monitoring nodes, and constructing a first three-dimensional data matrix according to the number of the electromagnetic monitoring nodes, the time window and the original electromagnetic signals; preprocessing the first three-dimensional data matrix to obtain first spatial-temporal characteristic data; inputting the first spatial-temporal characteristic data into a double-branch neural network for processing to obtain a first interference source coordinate prediction result; and determining the geographic coordinates of the interference source according to the first interference source coordinate prediction result. By constructing the three-dimensional data matrix and the double-branch neural network, high-precision positioning of the interference source is realized, the robustness and reliability in a complex electromagnetic environment are improved, and the method can be widely applied to the technical field of electromagnetic monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic monitoring, and particularly to a method, system, device and medium for locating electromagnetic signal interference sources based on a neural network. Background Art

[0002] With the continuous improvement of the requirements for the stability of the electromagnetic environment in fields such as power systems, wireless communications, and rail transit, electromagnetic monitoring technology plays an increasingly important role in practical applications. By real-time monitoring and analyzing electromagnetic signals, abnormal interferences, signal conflicts or potential risk sources in system operation can be effectively identified, providing technical support for system operation and maintenance, fault troubleshooting and interference control.

[0003] However, in the prior art, there are still many limitations in the interference source location of electromagnetic monitoring systems. On the one hand, current electromagnetic monitoring means mainly focus on single-point monitoring or single-model analysis, making it difficult to achieve high-precision location in complex electromagnetic environments. On the other hand, although some solutions have tried to introduce convolutional neural networks to process monitoring data, their network structures are generally relatively simple, lacking the ability to jointly model multi-dimensional features, and the training process is highly dependent on the number of samples, with slow model convergence speed and unstable location accuracy. Summary of the Invention

[0004] An object of the present invention is to solve at least to some extent one of the technical problems existing in the prior art.

[0005] To this end, an object of an embodiment of the present invention is to provide a method for locating electromagnetic signal interference sources based on a neural network. By jointly modeling spatio-temporal features, constructing a three-dimensional data matrix and a double-branch neural network, high-precision location of interference sources is achieved, improving the robustness and reliability in complex electromagnetic environments, significantly superior to traditional single-sensor location methods, and particularly applicable to the monitoring scenario of tracking mobile interference sources in complex electromagnetic environments.

[0006] Another object of an embodiment of the present invention is to provide a system for locating electromagnetic signal interference sources based on a neural network.

[0007] To achieve the above technical objectives, the technical solutions adopted in the embodiments of the present invention include:

[0008] In the first aspect, an embodiment of the present invention provides a method for locating electromagnetic signal interference sources based on a neural network, including:

[0009] Collecting original electromagnetic signals through a plurality of electromagnetic monitoring nodes within a preset time window, and constructing a first three-dimensional data matrix according to the number of the electromagnetic monitoring nodes, the time window and the original electromagnetic signals;

[0010] Preprocess the first three-dimensional data matrix to obtain the first spatio-temporal feature data;

[0011] Input the first spatio-temporal feature data into a dual-branch neural network for processing to obtain the prediction result of the first interference source coordinates;

[0012] Determine the geographical coordinates of the interference source according to the prediction result of the first interference source coordinates.

[0013] Furthermore, the three-dimensional data matrix includes time dimension information, space dimension information, and feature dimension information. The construction of the three-dimensional data matrix according to the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals includes:

[0014] Perform time alignment on the original electromagnetic signals, divide a number of sampling time periods according to the time window, and perform feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data;

[0015] Determine the time dimension information according to the sampling time period, determine the space dimension information according to the number of electromagnetic monitoring nodes, and determine the frequency dimension information according to the frequency feature data, thereby obtaining the three-dimensional data matrix.

[0016] Furthermore, the frequency feature data includes the main frequency point amplitude, the total energy of the frequency band, and the spectral entropy. The performing feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data includes:

[0017] Perform short-time Fourier transform on the original electromagnetic signals corresponding to each sampling time period to obtain the spectrum corresponding to each sampling time period;

[0018] Determine the main frequency point amplitude, the total energy of the frequency band, and the spectral entropy according to the spectrum.

[0019] Furthermore, the first spatio-temporal feature data includes a time feature matrix, a spatial topological relationship matrix, and a feature mean matrix. The preprocessing of the first three-dimensional data matrix to obtain the first spatio-temporal feature data includes:

[0020] Construct the time feature matrix according to the time dimension information and the feature dimension information of the first three-dimensional data matrix;

[0021] Construct the spatial topological relationship matrix according to the distances between the electromagnetic monitoring nodes;

[0022] Construct the feature mean matrix according to the space dimension information and the feature dimension information of the first three-dimensional data matrix.

[0023] Further, the dual-branch neural network includes a time feature extraction sub-network, a spatial feature extraction sub-network, a fusion layer, and an output layer. Processing the first spatio-temporal feature data through the dual-branch neural network to obtain a first interference source coordinate prediction result includes:

[0024] Inputting the time feature matrix into the time feature extraction sub-network for processing to obtain a time feature vector;

[0025] Inputting the spatial topological relationship matrix and the feature mean matrix into the spatial feature extraction sub-network for processing to obtain a spatial feature vector;

[0026] Inputting the time feature vector and the spatial feature vector into the fusion layer for processing to obtain a fusion feature vector;

[0027] Inputting the fusion feature vector into the output layer for processing to output the first interference source coordinate prediction result.

[0028] Further, the dual-branch neural network is trained through the following steps:

[0029] Obtaining historical monitoring data, constructing a second three-dimensional data matrix according to the historical monitoring data, and determining the true position coordinates of the interference source corresponding to the second three-dimensional data matrix;

[0030] Preprocessing the second three-dimensional matrix to obtain second spatio-temporal feature data;

[0031] Inputting the second spatio-temporal feature data into the dual-branch neural network for processing to obtain a second interference source coordinate prediction result;

[0032] Determining the loss value of the training according to the second interference source coordinate prediction result and the true position coordinates of the interference source;

[0033] Updating the parameters of the dual-branch neural network according to the loss value.

[0034] Further, determining the geographical coordinates of the interference source according to the first interference source coordinate prediction result includes:

[0035] Calibrating the geographical coordinates of each electromagnetic monitoring node;

[0036] Converting the first interference source coordinate prediction result into the geographical coordinates of the interference source according to the geographical coordinates of the electromagnetic monitoring node and a preset mapping relationship.

[0037] In a second aspect, an embodiment of the present invention provides an electromagnetic signal interference source positioning system based on a neural network, including:

[0038] A three-dimensional data matrix construction module, configured to collect original electromagnetic signals through a plurality of electromagnetic monitoring nodes within a preset time window, and construct a first three-dimensional data matrix according to the number of the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals;

[0039] A data preprocessing module, configured to preprocess the first three-dimensional data matrix to obtain first spatio-temporal feature data;

[0040] An interference source coordinate prediction module, configured to input the first spatio-temporal feature data into a dual-branch neural network for processing to obtain a first interference source coordinate prediction result;

[0041] An interference source positioning module, configured to determine the geographical coordinates of the interference source according to the first interference source coordinate prediction result.

[0042] In a third aspect, an embodiment of the present invention provides a device, including:

[0043] At least one processor;

[0044] At least one memory, configured to store at least one program;

[0045] When the at least one program is executed by the at least one processor, the at least one processor is caused to implement an electromagnetic signal interference source positioning method based on a neural network as described above.

[0046] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored, and the program executable by the processor is used to execute an electromagnetic signal interference source positioning method based on a neural network as described above when being executed by the processor.

[0047] The advantages and beneficial effects of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention:

[0048] In the embodiments of the present invention, a three-dimensional data matrix including a time dimension, a space dimension, and a feature dimension is constructed to uniformly model the monitoring data of multiple electromagnetic monitoring nodes, enabling subsequent data processing to simultaneously consider the temporal evolution characteristics and spatial distribution characteristics of electromagnetic signals, providing structured multi-dimensional input data for subsequent interference source localization tasks; through data preprocessing of the three-dimensional data matrix, three types of feature information, namely a time feature matrix, a spatial topology relationship matrix, and a feature mean matrix, are extracted to ensure that the neural network model can fully utilize the potential features of electromagnetic signals in the time and space dimensions, improving the interference source localization ability of the neural network model; by constructing a dual-branch neural network to process time features and spatial features respectively and output the predicted results of the interference source coordinates after fusion, the ability of the neural network model to capture temporal change features and global spatial features is further improved, which helps the neural network model maintain high interference source localization accuracy and robustness in a complex electromagnetic environment. Description of the Drawings

[0049] Figure 1 It is a schematic diagram of the steps of a method for locating an electromagnetic signal interference source based on a neural network provided by an embodiment of the present invention;

[0050] Figure 2 It is a schematic diagram of a system for locating an electromagnetic signal interference source based on a neural network provided by an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of the structure of a device provided by an embodiment of the present invention. Detailed Embodiments

[0052] The embodiments of the present invention are described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention. For the step numbers in the following embodiments, they are only set for the convenience of description and explanation, and no limitation is imposed on the order between the steps. The execution order of each step in the embodiments can be adjusted adaptively according to the understanding of those skilled in the art.

[0053] In the description of the present invention, "a plurality" means two or more. If the first and second are described, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art of this technology.

[0054] Figure 1Schematic diagram of the steps of a method for locating electromagnetic signal interference sources based on a neural network provided by an embodiment of the present invention. Refer to Figure 1 An embodiment of the present invention provides a method for locating electromagnetic signal interference sources based on a neural network, including:

[0055] S101. Collect original electromagnetic signals through a number of electromagnetic monitoring nodes within a preset time window, and construct a first three-dimensional data matrix according to the number of electromagnetic monitoring nodes, the time window, and the original electromagnetic signals;

[0056] Specifically, in this embodiment, each electromagnetic monitoring node is distributed at different positions in the monitoring area. Each node includes a data acquisition and evaluation unit for collecting electromagnetic signals in the environment, which can be specifically an electromagnetic acquisition device or sensor such as a spectrum analyzer or a software-defined radio receiver. Each node performs continuous sampling at a unified sampling frequency within the preset time window length. Within this time window, the original electromagnetic signals collected by each monitoring node are arranged in chronological order, numbered and distinguished according to the identification information of different monitoring nodes, and at the same time, the characteristic data of the original electromagnetic signals are extracted at equal interval time steps, so as to obtain a three-dimensional data structure with a time dimension T, the number S of electromagnetic monitoring nodes, and a characteristic dimension F. This three-dimensional matrix can be used for subsequent feature extraction and serves as the data basis for the entire interference source location task.

[0057] In some alternative embodiments, the three-dimensional data matrix includes time dimension information, space dimension information, and characteristic dimension information. Constructing the three-dimensional data matrix according to the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals includes:

[0058] A1. Perform time alignment on the original electromagnetic signals, divide a number of sampling time periods according to the time window, and perform feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data;

[0059] A2. Determine the time dimension information according to the sampling time period, determine the space dimension information according to the number of electromagnetic monitoring nodes, and determine the frequency dimension information according to the frequency feature data, and then obtain the three-dimensional data matrix.

[0060] In some alternative embodiments, the frequency feature data includes the main frequency point amplitude, the total energy of the frequency band, and the spectrum entropy. Performing feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data includes:

[0061] B1. Perform short-time Fourier transform on the original electromagnetic signals corresponding to each sampling time period to obtain the spectrum corresponding to each sampling time period;

[0062] B2. Determine the main frequency point amplitude, the total energy of the frequency band, and the spectrum entropy according to the spectrum.

[0063] Specifically, since multiple electromagnetic monitoring nodes operate independently, there may be time deviations in the data they collect. First, it is necessary to align the time of the original electromagnetic signals collected by different nodes to ensure that the data of different nodes has a unified timestamp reference. Specifically, the timestamps of different nodes can be synchronized through time synchronization methods such as the Precision Time Protocol (PTP) or the Global Positioning System (GPS) timing mechanism. The sampling time period refers to dividing multiple time periods at fixed time intervals within this time window. Each time period corresponds to a segment of the electromagnetic signal after time alignment. Perform a short-time Fourier transform on each segment of the signal to obtain its spectrogram, and extract frequency feature data by analyzing the distribution of frequencies and amplitudes in the spectrogram. These features can specifically include the amplitudes of the main frequency points (such as the first 3 frequency points with the largest amplitudes), the total energy of the frequency band (such as the total energy within the range of 30 MHz to 300 MHz), and frequency domain features such as spectral entropy.

[0064] In this embodiment, the time dimension information is the number T of sampling time periods in each time window, the space dimension information is the number S of electromagnetic monitoring nodes participating in signal collection, and the feature dimension information is the number F of frequency feature data corresponding to the electromagnetic signals collected by each electromagnetic monitoring node in each sampling time period. Based on the above three types of dimension information, a three-dimensional data matrix X can be constructed, and its structure is X ∈ R^(T×S×F). The specific organization method of this three-dimensional data matrix is as follows:

[0065] The first dimension T: represents the time axis, and each layer corresponds to a sampling time period;

[0066] The second dimension S: represents the space axis, and each layer corresponds to an electromagnetic monitoring node;

[0067] The third dimension F: represents the feature axis, and each layer corresponds to a type of frequency feature data;

[0068] That is, in the three-dimensional data matrix X, X[t][s][f] represents the value of the f-th type of frequency feature corresponding to the electromagnetic signal collected by the s-th electromagnetic monitoring node in the t-th sampling time period. Specifically, if the time window is set to 60 seconds, the sampling time period is 1 second, a total of 4 electromagnetic monitoring nodes are deployed, and 3 frequency feature data are extracted from the electromagnetic signals corresponding to each sampling time period, then the dimension of the finally constructed three-dimensional data matrix is 60×4×3.

[0069] It can be recognized that in this embodiment, by constructing a three-dimensional data matrix, the evolutionary characteristics of electromagnetic monitoring data in the time dimension, the distribution characteristics in the space dimension, and the signal characteristics in the frequency dimension are unifiedly modeled, providing a multi-dimensional and structured input data basis for completing the interference source location task.

[0070] S102. Preprocess the first three-dimensional data matrix to obtain first spatio-temporal feature data;

[0071] Specifically, in this embodiment, to obtain input data that is convenient for subsequent processing by the double-branch neural network model, it is necessary to preprocess the three-dimensional data matrix. The preprocessing operations may include conventional operations such as data cleaning and data augmentation, and may also include extracting data that can represent the time evolution characteristics and spatial distribution characteristics of the original electromagnetic signal from the three-dimensional data matrix to construct spatio-temporal feature data, so as to improve the accuracy of interference source localization through multi-dimensional feature extraction.

[0072] In some alternative embodiments, the first spatio-temporal feature data includes a time feature matrix, a spatial topological relationship matrix, and a feature mean matrix. Preprocessing the first three-dimensional data matrix to obtain the first spatio-temporal feature data includes:

[0073] C1. Construct a time feature matrix according to the time dimension information and feature dimension information of the first three-dimensional data matrix;

[0074] C2. Construct a spatial topological relationship matrix according to the distances between each electromagnetic monitoring node;

[0075] C3. Construct a feature mean matrix according to the feature dimension information of the first three-dimensional data matrix.

[0076] Specifically, in this embodiment, the time feature matrix is used to reflect the frequency feature evolution of the electromagnetic signals collected by each electromagnetic monitoring node in different sampling time periods. The construction method is: extract the frequency feature data of each electromagnetic monitoring node in each sampling time period in chronological order to form multiple time feature matrices. The structure of each time feature matrix is X∈R^(T×F), where T is the number of sampling time periods, F is the number of different frequency feature data, and the number of time feature matrices is the same as the number of electromagnetic monitoring nodes.

[0077] The spatial topological relationship matrix is used to reflect the spatial distribution relationship between different electromagnetic monitoring nodes. The construction method is: construct a spatial topological relationship matrix A∈R^(S×S) according to the physical position relationship between all electromagnetic monitoring nodes, where S is the number of electromagnetic monitoring nodes, and the element A ij can be the reciprocal of the physical distance between the i-th electromagnetic monitoring node and the j-th electromagnetic monitoring node;

[0078] The feature mean matrix is used to reflect the statistical information of the frequency feature data collected by each electromagnetic monitoring node within the entire time window. Its construction method is as follows: According to the feature dimension information in the three-dimensional data matrix, calculate the average value of each frequency feature data corresponding to each electromagnetic monitoring node within the entire time window, and construct the feature mean matrix μ ∈ R^S×F, where F is the number of different frequency feature data, and S is the number of electromagnetic monitoring nodes.

[0079] It can be recognized that in this embodiment, by constructing the time feature matrix to extract the temporal features of the electromagnetic signal, it can provide discriminant information in the time dimension for subsequent interference source localization. By constructing the spatial topology relationship matrix and the feature mean matrix to encode the spatial distribution structure of different electromagnetic monitoring nodes and their corresponding frequency features, it can provide discriminant information in the spatial dimension for subsequent interference source localization.

[0080] S103. Input the first spatio-temporal feature data into the dual-branch neural network for processing to obtain the first interference source coordinate prediction result;

[0081] Specifically, in this embodiment, the prediction result of the interference source coordinates is obtained by inputting the spatio-temporal feature data into the dual-branch neural network for processing. The dual-branch neural network can process different parts of the spatio-temporal feature data through different branches, extract the corresponding features, and then fuse and output the extracted features to obtain the two-dimensional interference source coordinate prediction result.

[0082] In some alternative embodiments, the dual-branch neural network includes a time feature extraction sub-network, a spatial feature extraction sub-network, a fusion layer, and an output layer. Inputting the first spatio-temporal feature data into the dual-branch neural network for processing to obtain the first interference source coordinate prediction result includes:

[0083] D1. Input the time feature matrix into the time feature extraction sub-network for processing to obtain the time feature vector;

[0084] D2. Input the spatial topology relationship matrix and the feature mean matrix into the spatial feature extraction sub-network for processing to obtain the spatial feature vector;

[0085] D3. Input the time feature vector and the spatial feature vector into the fusion layer for processing to obtain the fusion feature vector;

[0086] D4. Input the fusion feature vector into the output layer for processing to output the first interference source coordinate prediction result.

[0087] Specifically, the dual-branch neural network includes a time feature extraction subnetwork, a spatial feature extraction subnetwork, a fusion layer and an output layer. The time feature extraction subnetwork is used to process the time feature matrix of the electromagnetic monitoring node and extract the signal evolution characteristics of each node within the time window; the spatial feature extraction subnetwork is used to combine the spatial topological relationship between the electromagnetic monitoring nodes and the frequency characteristics of the content in the entire time window for analysis to extract the global spatial features; the fusion layer is used to splice and fuse the time feature vector with the spatial feature vector, and further compress the dimension of the vector to obtain a fused feature vector; the output layer is used to map the fused feature vector to the target prediction value, that is, the two-dimensional spatial position coordinates of the interference source in the monitoring area.

[0088] In this embodiment, the input of the time feature extraction subnetwork is a two-dimensional time feature matrix of dimension T×F, where T represents the number of sampling time periods and F represents the number of different types of frequency feature data extracted in each time period. The matrix can be regarded as a one-dimensional sequence containing T time steps, each step containing F features. It is processed by a one-dimensional convolutional neural network (1D CNN), and a one-dimensional convolution operation is performed on the time axis through several convolution kernels to extract its signal evolution characteristics on the time series.

[0089] The specific design of the temporal feature extraction subnetwork is as follows: when the dimension of the temporal feature matrix is 60×3, that is, T=60, F=3, it can be regarded as a one-dimensional sequence of length 60, and each element in the sequence is a three-dimensional vector. The convolution kernel size of the one-dimensional convolution layer can be designed to be 5, the number is 32, and the step size is 1. The convolution obtains a convolution result of dimension 32×56, and the convolution result is sent to the one-dimensional maximum pooling layer of size 2 for downsampling to obtain a downsampling result of dimension 32×28. The downsampling result is nonlinearly transformed by the ReLU activation function to obtain the activation output, and then the activation output is flattened by the Flatten layer to obtain a one-dimensional vector of dimension 896, and then continuously passes through the fully connected layer FC-256 and the fully connected layer FC-128 to obtain a temporal feature vector of dimension 128.

[0090] The input of the spatial feature extraction subnetwork includes the spatial topological relationship matrix and the feature mean matrix. The dimension of the spatial topological relationship matrix is S×S, where S represents the number of electromagnetic monitoring nodes, and the dimension of the feature mean matrix is S×F, where F represents the number of different frequency feature data types. For ease of processing, the spatial topological relationship matrix and the feature mean matrix can be flattened into one-dimensional vectors and concatenated before being processed through a fully connected layer.

[0091] The specific design of the spatial feature extraction sub-network can be as follows. When the dimension of the spatial topology relation matrix is 4×4 and the dimension of the feature mean matrix is 4×3, that is, S = 4 and F = 3, the two can be flattened into two one-dimensional vectors with dimensions of 16 and 12 respectively through the Flatten layer. After splicing, a joint input vector with a dimension of 28 is obtained. The joint input vector is sent to the fully connected layer FC-64 for processing to obtain a spatial feature vector with a dimension of 64.

[0092] The specific design of the fusion layer can be: merge the time feature vector with a dimension of 128 and the spatial feature vector with a dimension of 64 in the way of vector splicing to obtain a fusion input vector with a dimension of 192, and then continuously pass through the fully connected layer FC-128 and the fully connected layer FC-64 to obtain a fusion feature vector with a dimension of 64.

[0093] The specific design of the output layer can be: receive the 64-dimensional fusion feature vector output by the fusion layer, and map it to a two-dimensional spatial position coordinate vector (x, y) through the fully connected layer FC-2, which respectively represents the predicted values of the abscissa and ordinate of the interference source in the monitoring area, to obtain the predicted result of the interference source coordinates.

[0094] In some alternative embodiments, the double-branch neural network is trained through the following steps:

[0095] E1. Obtain historical monitoring data, construct a second three-dimensional data matrix according to the historical monitoring data, and determine the true position coordinates of the interference source corresponding to the second three-dimensional data matrix;

[0096] E2. Preprocess the second three-dimensional matrix to obtain second spatio-temporal feature data;

[0097] E3. Input the second spatio-temporal feature data into the double-branch neural network for processing to obtain a second predicted result of the interference source coordinates;

[0098] E4. Determine the loss value of the training according to the second predicted result of the interference source coordinates and the true position coordinates of the interference source;

[0099] E5. Update the parameters of the double-branch neural network according to the loss value.

[0100] Specifically, in this embodiment, in order to enable the dual-branch neural network to effectively map the input data to the coordinates of the interference source, it is necessary to supervise and train the network with known labeled data. Specifically, a training sample set is constructed from historical electromagnetic monitoring data, and data such as the number of electromagnetic monitoring nodes, time window, and original electromagnetic signals in it are extracted to construct a second three-dimensional data matrix with the same data structure as the first three-dimensional data matrix as the basic data source in the model training stage. At the same time, it is necessary to confirm the true position coordinates of the corresponding interference source from the historical data as the labels for training. Subsequently, preprocessing operations such as data cleaning, data augmentation, and multi-dimensional feature extraction are performed on the second three-dimensional data matrix to obtain the second spatio-temporal feature data, which is fed into the dual-branch neural network for processing. The loss value is calculated based on the output result and the label, and then the parameters of the dual-branch neural network are updated through the backpropagation algorithm according to the loss value until the model converges.

[0101] The specific training parameters and training strategies can be designed as follows: The batch size and the number of training epochs can be set to 64 and 100 respectively to adapt to the training scale of this model. The Huber loss function is used as the loss function, which is close to the mean squared error when the predicted value is close to the target value, has strong smoothness, and is close to the absolute error when the error is large, maintaining a certain sensitivity, enhancing the robustness to outliers, and helping to prevent overfitting. The Adam algorithm is selected as the optimizer for backpropagation, and the initial learning rate is set to 0.001. At the same time, the branch freezing training strategy, dynamic learning rate adjustment strategy, and early stopping strategy are combined to improve the training efficiency.

[0102] S104. Determine the geographical coordinates of the interference source according to the first interference source coordinate prediction result, including:

[0103] In some alternative embodiments, determining the geographical coordinates of the interference source according to the first interference source coordinate prediction result includes:

[0104] F1. Calibrate the geographical coordinates of each electromagnetic monitoring node;

[0105] F2. Convert the first interference source coordinate prediction result into the geographical coordinates of the interference source according to the geographical coordinates of the electromagnetic monitoring node and the preset mapping relationship.

[0106] Specifically, in this embodiment, the coordinates of the interference source prediction result are coordinates in the relative coordinate system. It is necessary to establish a mapping relationship between the relative coordinate system coordinates and the geographical coordinate system coordinates to determine the geographical coordinates of the interference source. First, the geographical coordinates of the electromagnetic monitoring nodes are calibrated. For example, the longitude and latitude of each node are obtained through the GPS system or the absolute coordinate system of the node relative to a certain reference point is calibrated on-site by surveying tools. Then, the interference source coordinate prediction result is converted into the geographical coordinates of the interference source through a pre-determined mapping relationship model to obtain the final positioning result.

[0107] It can be recognized that in the embodiments of the present invention, by constructing a three-dimensional data matrix including time dimension, space dimension, and feature dimension to uniformly model the monitoring data of multiple electromagnetic monitoring nodes, the subsequent data processing process can simultaneously consider the temporal evolution characteristics and spatial distribution characteristics of electromagnetic signals, providing structured multi-dimensional input data for the subsequent interference source localization task; by performing data preprocessing on the three-dimensional data matrix, three types of feature information, namely the time feature matrix, the spatial topology relationship matrix, and the feature mean matrix, are extracted to ensure that the neural network model can fully utilize the potential features of electromagnetic signals in the time and space dimensions, improving the interference source localization ability of the neural network model; by constructing a dual-branch neural network to process the time feature and the spatial feature respectively and output the prediction result of the interference source coordinates after fusion, the ability of the neural network model to capture temporal change features and global spatial features is further improved, which helps the neural network model to maintain high interference source localization accuracy and robustness in a complex electromagnetic environment.

[0108] Referring to Figure 2 , the embodiments of the present invention provide an electromagnetic signal interference source localization system based on a neural network, including:

[0109] A three-dimensional data matrix construction module, configured to collect original electromagnetic signals through a plurality of electromagnetic monitoring nodes within a preset time window, and construct a first three-dimensional data matrix according to the number of electromagnetic monitoring nodes, the time window, and the original electromagnetic signals;

[0110] A data preprocessing module, configured to preprocess the first three-dimensional data matrix to obtain first spatio-temporal feature data;

[0111] An interference source coordinate prediction module, configured to input the first spatio-temporal feature data into a dual-branch neural network for processing to obtain a first interference source coordinate prediction result;

[0112] An interference source localization module, configured to determine the geographical coordinates of the interference source according to the first interference source coordinate prediction result.

[0113] The content in the above method embodiments is applicable to the system embodiments of the present invention. The functions specifically implemented by the system embodiments of the present invention are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0114] Referring to Figure 3 , the embodiments of the present invention provide a device, including:

[0115] At least one processor;

[0116] At least one memory, configured to store at least one program;

[0117] When at least one of the above-mentioned programs is executed by at least one of the above-mentioned processors, the at least one processor implements the above-mentioned method for locating an electromagnetic signal interference source based on a neural network.

[0118] An embodiment of the present invention further provides a computer-readable storage medium, in which a program executable by a processor is stored. The program executable by the processor is used to execute the above-mentioned method for locating an electromagnetic signal interference source based on a neural network when executed by the processor.

[0119] A computer-readable storage medium according to an embodiment of the present invention can execute a method for locating an electromagnetic signal interference source based on a neural network provided by an embodiment of the method of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0120] An embodiment of the present invention also discloses a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of the device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the device executes Figure 1 the method for locating an electromagnetic signal interference source based on a neural network as shown.

[0121] In some alternative embodiments, the functions / operations mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two consecutive blocks shown may actually be executed substantially simultaneously or the above-mentioned blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowchart of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical processes presented herein. Alternative embodiments are foreseeable, in which the order of various operations is changed and the sub-operations described as part of a larger operation are executed independently.

[0122] In addition, although the present invention has been described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the above-described functions and / or features may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It should also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. Rather, given the attributes, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skills of an engineer. Thus, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation. It should also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0123] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the above methods in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0124] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0125] More specific examples (nonexhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which the above programs can be printed, because the above programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other appropriate processing as necessary, and then stored in a computer memory.

[0126] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0127] In the above description of this specification, the descriptions referring to the terms "one embodiment / example", "another embodiment / example", or "certain embodiments / examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0128] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0129] The above has specifically described the preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for locating electromagnetic signal interference sources based on neural networks, characterized in that, Including: Collecting original electromagnetic signals through a number of electromagnetic monitoring nodes within a preset time window, and constructing a first three-dimensional data matrix according to the number of the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals; Preprocessing the first three-dimensional data matrix to obtain first spatio-temporal feature data; Inputting the first spatio-temporal feature data into a dual-branch neural network for processing to obtain a first interference source coordinate prediction result; Determining the geographical coordinates of the interference source according to the first interference source coordinate prediction result.

2. The electromagnetic signal interference source localization method based on neural network according to claim 1, wherein The three-dimensional data matrix includes time dimension information, space dimension information, and feature dimension information. The constructing of the three-dimensional data matrix according to the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals includes: Performing time alignment on the original electromagnetic signals, dividing a number of sampling time periods according to the time window, and performing feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data; Determining the time dimension information according to the sampling time periods, determining the space dimension information according to the number of the electromagnetic monitoring nodes, and determining the frequency dimension information according to the frequency feature data, thereby obtaining the three-dimensional data matrix.

3. The electromagnetic signal interference source localization method based on a neural network according to claim 2, wherein, The frequency feature data includes main frequency point amplitude, total band energy, and spectrum entropy. The performing of feature extraction on the original electromagnetic signals corresponding to each sampling time period to obtain a number of frequency feature data includes: Performing short-time Fourier transform on the original electromagnetic signals corresponding to each sampling time period to obtain the spectrum corresponding to each sampling time period; Determining the main frequency point amplitude, the total band energy, and the spectrum entropy according to the spectrum.

4. The electromagnetic signal interference source localization method based on a neural network according to claim 2, characterized in that, The first spatio-temporal feature data includes a time feature matrix, a spatial topological relationship matrix, and a feature mean matrix. The preprocessing of the first three-dimensional data matrix to obtain first spatio-temporal feature data includes: Constructing the time feature matrix according to the time dimension information and the feature dimension information of the first three-dimensional data matrix; Constructing the spatial topological relationship matrix according to the distances between the electromagnetic monitoring nodes; Constructing the feature mean matrix according to the space dimension information and the feature dimension information of the first three-dimensional data matrix.

5. A method for locating an electromagnetic signal interference source based on a neural network according to claim 4, characterized in that The dual-branch neural network includes a time feature extraction sub-network, a space feature extraction sub-network, a fusion layer, and an output layer. The inputting of the first spatio-temporal feature data into the dual-branch neural network for processing to obtain a first interference source coordinate prediction result includes: Inputting the time feature matrix into the time feature extraction sub-network for processing to obtain a time feature vector; Inputting the spatial topological relationship matrix and the feature mean matrix into the space feature extraction sub-network for processing to obtain a space feature vector; Inputting the time feature vector and the space feature vector into the fusion layer for processing to obtain a fusion feature vector; Inputting the fusion feature vector into the output layer for processing to output the first interference source coordinate prediction result.

6. The electromagnetic signal interference source location method based on a neural network according to claim 1, wherein The dual-branch neural network is trained through the following steps: Obtain historical monitoring data, construct a second three-dimensional data matrix according to the historical monitoring data, and determine the true position coordinates of the interference source corresponding to the second three-dimensional data matrix; Preprocess the second three-dimensional matrix to obtain second spatio-temporal feature data; Input the second spatio-temporal feature data into the dual-branch neural network for processing to obtain a second interference source coordinate prediction result; Determine the loss value of training according to the second interference source coordinate prediction result and the true position coordinates of the interference source; Update the parameters of the dual-branch neural network according to the loss value.

7. A method for locating an electromagnetic signal interference source based on a neural network according to claim 2, characterized in that The determining the geographical coordinates of the interference source according to the first interference source coordinate prediction result includes: Calibrate the geographical coordinates of each electromagnetic monitoring node; Convert the first interference source coordinate prediction result into the geographical coordinates of the interference source according to the geographical coordinates of the electromagnetic monitoring node and a preset mapping relationship.

8. An electromagnetic signal interference source localization system based on a neural network, characterized in that, It includes: A three-dimensional data matrix construction module, configured to collect original electromagnetic signals through a plurality of electromagnetic monitoring nodes within a preset time window, and construct a first three-dimensional data matrix according to the number of the electromagnetic monitoring nodes, the time window, and the original electromagnetic signals; A data preprocessing module, configured to preprocess the first three-dimensional data matrix to obtain first spatio-temporal feature data; An interference source coordinate prediction module, configured to input the first spatio-temporal feature data into a dual-branch neural network for processing to obtain a first interference source coordinate prediction result; An interference source positioning module, configured to determine the geographical coordinates of the interference source according to the first interference source coordinate prediction result.

9. A device, characterized in that, It includes: At least one processor; At least one memory, configured to store at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a method for locating an electromagnetic signal interference source based on a neural network according to any one of claims 1-7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that, The program executable by the processor, when executed by the processor, is used to execute a method for locating an electromagnetic signal interference source based on a neural network according to any one of claims 1-7.

Images