CN114417938A - Electromagnetic target classification method using knowledge vector embedding - Google Patents
Electromagnetic target classification method using knowledge vector embedding Download PDFInfo
- Publication number
- CN114417938A CN114417938A CN202210101982.8A CN202210101982A CN114417938A CN 114417938 A CN114417938 A CN 114417938A CN 202210101982 A CN202210101982 A CN 202210101982A CN 114417938 A CN114417938 A CN 114417938A
- Authority
- CN
- China
- Prior art keywords
- neural network
- graph
- electromagnetic
- electromagnetic target
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Complex Calculations (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an electromagnetic target classification method by embedding knowledge vectors, which utilizes the data of known electromagnetic target information to establish a graph structure of an electromagnetic target, and carries out embedded vector representation on graph nodes corresponding to each electromagnetic target category based on a graph neural network; acquiring an electromagnetic target signal, performing short-time Fourier transform on electromagnetic target data to acquire time-frequency data of the electromagnetic target data, and preprocessing the time-frequency data to be used as a sample for training a convolutional neural network; constructing a convolutional neural network, training the convolutional neural network based on the result of the embedded vector representation of the graph nodes corresponding to the electromagnetic target class, and finally obtaining a reference vector for subsequently classifying and identifying the acquired electromagnetic target signals; and classifying and identifying the acquired electromagnetic target signals by using the obtained reference vectors. The method has strong applicability, combines the class relation knowledge into the network training, and solves the defect that the traditional classification network is only suitable for identifying the classes appearing in the training set.
Description
Technical Field
The invention relates to the field of communication and artificial intelligence, in particular to an electromagnetic target classification method embedded by using knowledge vectors.
Background
In the knowledge graph, the knowledge of the graph structure mainly comprises node knowledge and relation knowledge. The nodes can represent a certain known category, the relationship is the degree of relation between each category, the degree of closeness of the node and the node relationship can be reflected visually, and after the graph structure is embedded into a vector form, the vector representation of each node can reflect the knowledge of the relationship between the nodes. In recent years, due to the development of the graph neural network, the attributes and the interrelation of the nodes in the graph structure can be effectively utilized, modeling is carried out by depending on the interrelation, and the nodes are embedded into vectors to provide a computable data format for the subsequent processing of a computer. At present, when a pattern neural network is used for sample embedding and then electromagnetic targets are classified, the following two problems mainly exist: firstly, the graph neural network architecture is mainly used for node classification, and the execution efficiency of a multi-sample embedded electromagnetic target classification task is low; secondly, the graph neural network needs the prior relation knowledge of each node and other nodes, and in an actual situation, the relation information is difficult to reflect on a sample level. The convolutional neural network can classify different inputs according to labels, and the traditional convolutional neural network has the following defects when being used for graph structure classification: (1) labels in the traditional method cannot reflect the relationship between samples and the closeness degree of the relationship; (2) the output of a conventional convolutional neural network cannot be represented as an embedded representation with knowledge of the graph structure.
Disclosure of Invention
Aiming at the problem of embedding a priori knowledge sample with a graph structure and classifying an electromagnetic target, the invention combines the characteristics of two neural networks with different structures, vector embedding is carried out on the sample with the priori knowledge structure, and then the vector embedding is used for a technology of classifying neural network training, so that the fusion of the priori knowledge such as the class relation of the electromagnetic target and the like and the electromagnetic data characteristics in a vector embedding space is realized, and an electromagnetic target classifier driven by data and knowledge together is constructed. According to the closeness degree of the relationship between the classes to which the samples belong, an embedding result capable of reflecting the relationship between the samples can be obtained. The distance of the embedded vector can be used to reflect the degree of relation of features between samples. The invention converts the hard decision of the traditional classification neural network represented by the convolutional neural network into the soft decision for calculating the similarity between vectors, thereby solving the problem that the current sample embedding result can not reflect the relationship between the classes of the sample and the hard decision during the classification decision.
The method is generally divided into two steps, namely embedding the graph nodes corresponding to the classes containing the mutual relations based on the graph neural network, and training the convolutional neural network based on the embedding result of the graph neural network, so that the embedding result of the samples under the classes can reflect the relation of the classes of the samples.
The invention discloses an electromagnetic target classification method by embedding knowledge vectors, which comprises the following specific steps: establishing a graph structure of the electromagnetic targets by using the data of the known electromagnetic target information, wherein the graph structure comprises graph nodes and relations, the graph nodes are used for representing the known electromagnetic target categories, and the relations are used for representing the association degree between each electromagnetic target category;
and S1, carrying out embedded vector representation on the graph nodes corresponding to each electromagnetic object type based on the graph neural network. The basic description process of the electromagnetic object class corresponding graph nodes comprises the following steps: the relation between the categories of the electromagnetic objects is represented by an adjacency matrix D, and the elements D of the ith row and the jth column of the adjacency matrix DijAnd representing the relation between the ith electromagnetic object class and the jth electromagnetic object class, representing the characteristics of all the electromagnetic object classes by using a matrix F, representing the characteristics of the ith electromagnetic object class by using the ith row of the matrix F, inputting D and F into a neural network of the graph, and training the neural network of the graph to obtain the embedded vector representation of all the electromagnetic object classes corresponding to graph nodes.
The step S1 specifically includes:
in the graph nerve, ReLU (-) is a linear rectification function; ii | represents the norm of l 2; softmax (. circle.) represents a logistic regression functionCounting; e is a unit matrix, N represents the dimension of the unit matrix E and is also the number of layers of the neural network of the graph; wiWeight of the ith hidden layer of the neural network, W0∈RM×N,W1∈RN×HM is the dimension of the feature vector of the graph node, H is the dimension of the embedded vector output by the graph neural network, and k is the iteration number when the graph neural network is trained;
wherein the content of the first and second substances,
which represents a normalized laplacian matrix of the laplacian matrix,
a matrix of the degree of representation,
representation matrix
Row i, column i,
representation matrix
Row i, column j,
representing an undirected graph adjacency matrix; l isiA tag in the ith electromagnetic object class; q is an embedded vector representation of the electromagnetic object class corresponding to the graph node.
The method for training the graph neural network to obtain the embedded vector representation of all electromagnetic target classes corresponding to the graph nodes comprises the following specific steps:
s11, initializing a weight matrix W of the graph neural networkiN, N represents the number of layers of the neural network.
S12, calculating the output Y of the neural network of the graph, wherein the calculation formula is as follows:
wherein f () represents a calculation function of the neural network of the graph;
s13, calculating the Loss function Loss of the neural network of the graph under the output condition:
wherein L islLabels representing the ith electromagnetic object class, F represents a row in the matrix F, i.e. the eigenvectors of a certain electromagnetic object class, YfRepresenting the output obtained after the characteristic vector f is input into the neural network of the graph; l represents a set of tags of the electromagnetic object class.
S14, updating the weight of the neural network of the graph by using a batch gradient descent method (BGD) according to the loss function; adding 1 to the value of the iteration times k;
s15, repeating the process from the step S12 to the step S14 until the iteration number k reaches a preset value;
s16, calculating the embedded vector representation Q of the electromagnetic object class corresponding to the graph node, wherein the calculation formula is as follows:
thereby obtaining an embedded vector representation of the electromagnetic object class corresponding to the graph nodes.
S2, preprocessing the electromagnetic target signal;
the method comprises the steps of collecting electromagnetic target signals, storing the electromagnetic target signals according to collection time, carrying out short-time Fourier transform on the electromagnetic target data to obtain time-frequency data of the electromagnetic target data, preprocessing the time-frequency data, and transforming the time-frequency data into a data format which can be processed by a convolutional neural network to be used as a sample for training the convolutional neural network.
S3, constructing a convolutional neural network, training the convolutional neural network based on the result of the embedded vector representation of the graph nodes corresponding to the electromagnetic target classes obtained by the graph neural network, and finally obtaining a reference vector for subsequently classifying and identifying the collected electromagnetic target signals; and mapping the embedded representation of the signal characteristics of the electromagnetic target class to the embedded vector of the corresponding graph node of the electromagnetic target class by utilizing a convolutional neural network.
The step S3 specifically includes:
the embedded vector representation of the graph node corresponding to the class to which the sample belongs is obtained through the step S1, the embedded vector of the graph node corresponding to the class to which the sample belongs is used as a label of the convolutional neural network, the time-frequency data of the electromagnetic target preprocessed in the step S2 is used as a training sample, and the distance between the embedded vector of the graph node corresponding to the class to which the sample belongs and the embedded vector of the training sample is used as a loss function for training the convolutional neural network. After the convolutional neural network is subjected to repeated iterative optimization training, a reference vector for subsequently classifying and identifying the acquired electromagnetic target signals is obtained. The multiple iteration optimization training process for the convolutional neural network specifically comprises the following steps:
s31, constructing a convolutional neural network with two convolutional layers, two pooling layers and two full-connection layer structures, and initializing graph node weights of the network; the convolution layer, the pooling layer and the full-connection layer are connected in sequence;
s32, inputting the time-frequency data P of the electromagnetic target preprocessed in the step S2 into the convolutional neural network as a training sample, and obtaining a feature vector R of the training sample at the last layer of the convolutional neural network;
s33, calculating a loss function of the network: l | R-Q |, where Q is an embedded vector representation of the graph node corresponding to the class to which the input training sample belongs, and | represents the norm;
s34, optimizing the parameters of the convolutional neural network according to a random gradient descent method, and updating the weight of the graph nodes in the network;
s35, repeating the steps S32 to S34 until the iteration number reaches a preset value, and finishing the training of the convolutional neural network;
s36, after the training of the convolutional neural network is completed, the obtained output R is the final embedded vector representation result of the sample;
s37, calculating an average value Rc of embedded vector representation results of sample data of the same type, and taking the Rc as a reference vector for subsequently classifying and identifying the acquired electromagnetic target signals;
s4, classifying and identifying the acquired electromagnetic target signals by using the reference vector obtained in the step S3, wherein the method specifically comprises the following steps:
s41, carrying out short-time Fourier transform and preprocessing on the collected electromagnetic target signal to obtain a sample P 'to be recognized, and inputting the sample P' to be recognized into the convolutional neural network trained in the step S3.
And S42, calculating to obtain an output vector R 'corresponding to the sample P' to be identified through a convolutional neural network.
S43, calculating the vector similarity of the output vector R 'and the average value of the embedded vector representation results of the sample data of each category, and representing the vector similarity of the output vector R' and the ith category as epsiloniN, n is the total number of categories, and the vector similarity is obtained by calculating the Pearson correlation coefficient of two vectors.
S44, comparing the obtained vector similarity with threshold eta respectively, if epsilon existsjIf j is larger than or equal to eta, j belongs to n, judging that the sample P' to be recognized belongs to the jth electromagnetic target category; if for any epsilonjAll have epsilonjIf the sample P ' to be identified belongs to the electromagnetic target class, j belongs to n, and if the sample P ' to be identified belongs to n, the sample P ' to be identified belongs to a new electromagnetic target class.
The invention has the beneficial effects that:
the method has strong applicability and wide application range, combines the class relation knowledge into the network training, and solves the defect that the traditional classification network is only suitable for identifying the classes appearing in the training set. And the embedding result of the sample is more effective by utilizing the relation information of the class to which the sample belongs. The invention trains the convolutional neural network based on the embedding result of the graph neural network, and effectively reflects the relationship information of the class to which the sample belongs on the embedding result. Facilitating subsequent analysis of the sample and downstream applications.
Drawings
FIG. 1 is a flow chart of an implementation of the method of the present invention;
fig. 2 is a relationship structure diagram of a category corresponding node in this embodiment;
FIG. 3 is a diagram of the result of the class embedding vector in this embodiment after dimension reduction;
fig. 4 is a diagram of the result obtained after dimension reduction of the sample embedding vector in this embodiment.
Detailed Description
For a better understanding of the present disclosure, an example is given here.
FIG. 1 is a flow chart of an implementation of the method of the present invention; fig. 2 is a relationship structure diagram of a category corresponding node in this embodiment; FIG. 3 is a diagram of the result of the class embedding vector in this embodiment after dimension reduction; fig. 4 is a diagram of the result obtained after dimension reduction of the sample embedding vector in this embodiment.
The invention discloses an electromagnetic target classification method by embedding knowledge vectors, which comprises the following specific steps: establishing a graph structure of the electromagnetic targets by using the data of the known electromagnetic target information, wherein the graph structure comprises graph nodes and relations, the graph nodes are used for representing the known electromagnetic target categories, and the relations are used for representing the association degree between each electromagnetic target category;
and S1, carrying out embedded vector representation on the graph nodes corresponding to each electromagnetic object type based on the graph neural network. The basic description process of the electromagnetic object class corresponding graph nodes comprises the following steps: the relation between the categories of the electromagnetic objects is represented by an adjacency matrix D, and the elements D of the ith row and the jth column of the adjacency matrix DijRepresenting the relation between the ith and jth electromagnetic object classes, the characteristics of all electromagnetic object classes being usedAnd F, the matrix F represents, the ith row of the matrix F represents the characteristics of the ith electromagnetic object class, D and F are input into a graph neural network, and the graph neural network is trained to obtain the embedded vector representation of all electromagnetic object classes corresponding to graph nodes.
In the graph nerve, ReLU (-) is a linear rectification function; | represents a 12-norm; softmax (·) represents a logistic regression function; e is a unit matrix, N represents the dimension of the unit matrix E and is also the number of layers of the neural network of the graph; wiWeight of the ith hidden layer of the neural network, W0∈RM×N,W1∈RN×HM is the dimension of the feature vector of the graph node, H is the dimension of the embedded vector output by the graph neural network, and k is the iteration number when the graph neural network is trained;
wherein the content of the first and second substances,
which represents a normalized laplacian matrix of the laplacian matrix,
a matrix of the degree of representation,
representation matrix
Row i, column i,
representation matrix
Row i, column j,
representing an undirected graph adjacency matrix; l isiA tag in the ith electromagnetic object class; q is an embedded vector representation of the electromagnetic object class corresponding to the graph node.
The method for training the graph neural network to obtain the embedded vector representation of all electromagnetic target classes corresponding to the graph nodes comprises the following specific steps:
s11, initializing a weight matrix W of the graph neural networkiN, N represents the number of layers of the neural network.
S12, calculating the output Y of the neural network of the graph, wherein the calculation formula is as follows:
wherein f () represents a calculation function of the neural network of the graph;
s13, calculating the Loss function Loss of the neural network of the graph under the output condition:
wherein L islLabels representing the ith electromagnetic object class, F represents a row in the matrix F, i.e. the eigenvectors of a certain electromagnetic object class, YfRepresenting the output obtained after the characteristic vector f is input into the neural network of the graph; l represents a set of tags of the electromagnetic object class;
after the feature vectors representing all electromagnetic target classes are input into the neural network of the graph, the sum of all outputs obtained by the neural network of the graph is calculated. A feature vector of a certain electromagnetic object class, i.e. a vector formed by a number of specific parameters of a certain electromagnetic object class.
S14, updating the weight of the neural network of the graph by using a batch gradient descent method (BGD) according to the loss function; adding 1 to the value of the iteration times k;
s15, repeating the process from the step S12 to the step S14 until the iteration number k reaches a preset value;
s16, calculating the embedded vector representation Q of the electromagnetic object class corresponding to the graph node, wherein the calculation formula is as follows:
s2, preprocessing the electromagnetic target signal;
the method comprises the steps of collecting an electromagnetic target signal, storing the electromagnetic target signal according to collection time, carrying out short-time Fourier transform on the electromagnetic target data to obtain time-frequency data of the electromagnetic target data, preprocessing the time-frequency data, converting the time-frequency data into a data format which can be processed by a convolutional neural network, and using the data format as a sample for training the convolutional neural network, wherein the data format which can be processed by the convolutional neural network comprises picture or matrix data, so that the convolutional operation of the convolutional neural network is facilitated.
S3, constructing a convolutional neural network, training the convolutional neural network based on the result of the embedded vector representation of the graph nodes corresponding to the electromagnetic target classes obtained by the graph neural network, and finally obtaining a reference vector for subsequently classifying and identifying the collected electromagnetic target signals; and mapping the embedded representation of the signal characteristics of the electromagnetic target class to the embedded vector of the corresponding graph node of the electromagnetic target class by utilizing a convolutional neural network.
The embedded vector representation of the graph nodes corresponding to the class to which the sample belongs is obtained through the step S1, the graph node representation corresponding to the class can reflect the closeness degree of the relationship between the classes, the embedded vector of the graph node corresponding to the class to which the sample belongs is used as the label of the convolutional neural network, the time-frequency data of the electromagnetic target preprocessed in the step S2 is used as the training sample, and the distance between the embedded vector of the graph node corresponding to the class to which the sample belongs and the embedded vector of the training sample is used as the loss function for training the convolutional neural network. After the convolutional neural network is subjected to repeated iterative optimization training, a reference vector for subsequently classifying and identifying the acquired electromagnetic target signals is obtained. The multiple iteration optimization training process for the convolutional neural network specifically comprises the following steps:
s31, constructing a convolutional neural network with two convolutional layers, two pooling layers and two full-connection layer structures, and initializing graph node weights of the network; the convolution layer, the pooling layer and the full-connection layer are connected in sequence;
s32, inputting the time-frequency data P of the electromagnetic target preprocessed in the step S2 into the convolutional neural network as a training sample, and obtaining a feature vector R of the training sample at the last layer of the convolutional neural network;
s33, calculating a loss function of the network: l | R-Q |, where Q is an embedded vector representation of the graph node corresponding to the class to which the input training sample belongs, and | represents the norm;
s34, optimizing the parameters of the convolutional neural network according to a random gradient descent method, and updating the weight of the graph nodes in the network;
s35, repeating the steps S32 to S34 until the iteration number reaches a preset value, and finishing the training of the convolutional neural network;
s36, after the training of the convolutional neural network is completed, the obtained output R is the final embedded vector representation result of the sample;
s37, calculating an average value Rc of embedded vector representation results of sample data of the same type, and taking the Rc as a reference vector for subsequently classifying and identifying the acquired electromagnetic target signals;
s4, classifying and identifying the acquired electromagnetic target signals by using the reference vector obtained in the step S3, wherein the method specifically comprises the following steps:
s41, carrying out short-time Fourier transform and preprocessing on the collected electromagnetic target signal to obtain a sample P 'to be recognized, and inputting the sample P' to be recognized into the convolutional neural network trained in the step S3.
And S42, calculating to obtain an output vector R 'corresponding to the sample P' to be identified through a convolutional neural network.
S43, calculating the vector similarity of the output vector R' and the average value of the embedded vector representation result of the sample data of each category,the vector similarity of the output vector R' to the ith class is expressed as εiN, n is the total number of categories, and the vector similarity is obtained by calculating the Pearson correlation coefficient of two vectors.
S44, comparing the obtained vector similarity with threshold eta respectively, if epsilon existsjIf j is larger than or equal to eta, j belongs to n, judging that the sample P' to be recognized belongs to the jth electromagnetic target category; if for any epsilonjAll have epsilonjIf the sample P ' to be identified belongs to the electromagnetic target class, j belongs to n, and if the sample P ' to be identified belongs to n, the sample P ' to be identified belongs to a new electromagnetic target class.
By training the convolutional neural network based on the embedding results of the graph neural network, an embedded representation of the samples for each class can be finally obtained. Classification recognition based on the similarity of the embedding vectors can be performed by using the embedding representation. The sample embedding result can reflect the similarity degree between samples and the affinity degree of the class relationship of the samples. When the similarity degree is smaller than the set threshold value, the system can be judged as a new type.
Graph nodes corresponding to the classes containing the mutual relation information are embedded based on the graph neural network, and the embedding result of the nodes corresponding to each class can be obtained through training of the graph neural network. The distance between the embedded vectors can better reflect the degree of interconnection of the nodes. FIG. 2 is a verification result of embedding a node using a graph neural network.
For the class nodes, there are 15 nodes corresponding to the class in total, the size of the adjacency matrix a is 15 × 15, and the size of the node feature matrix X is 15 × 15. The convolutional layer of the selected graph neural network is two layers, the size of the embedding result Q is 15 x 128, and the dimension of the embedding vector of each node is 128 dimensions.
The relationship structure of the class corresponding node and the result of the embedded vector after dimension reduction are shown in fig. 3.
Training a convolutional neural network based on the graph neural network embedding result; after the embedding of the class to which the sample belongs is completed, the sample needs to be embedded by using the prior knowledge of the class embedding result. There are 100 samples under each class, there are 1500 samples in 15 classes altogether, carry on the short-time Fourier transform to the sample and obtain its time-frequency diagram. The embedded vector obtained by a convolutional neural network with two convolutional layers, two pooling layers and two full-connection layer structures is 128-dimensional. The distance between the embedded vectors of samples may reflect how close the relationship between samples is. After the embedded vector of samples is subjected to dimensionality reduction, the result shown in fig. 4 can be obtained.
After the sample embedding result is obtained, the method is beneficial to the subsequent analysis of the sample data and the downstream application based on the embedding vector.
The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (4)
1. The electromagnetic target classification method by using knowledge vector embedding is characterized in that a graph structure of an electromagnetic target is established by using data of known electromagnetic target information, the graph structure comprises graph nodes and relations, the graph nodes are used for representing known electromagnetic target classes, and the relations are used for representing the degree of association between each electromagnetic target class; the method comprises the following specific steps:
s1, carrying out embedded vector representation on graph nodes corresponding to each electromagnetic target type based on a graph neural network; the basic description process of the electromagnetic object class corresponding graph nodes comprises the following steps: the relation between the categories of the electromagnetic objects is represented by an adjacency matrix D, and the elements D of the ith row and the jth column of the adjacency matrix DijRepresenting the relation between the ith electromagnetic target class and the jth electromagnetic target class, representing the characteristics of all the electromagnetic target classes by using a matrix F, representing the characteristics of the ith electromagnetic target class by using the ith row of the matrix F, inputting D and F into a graph neural network, and training the graph neural network to obtain the embedded vector representation of all the electromagnetic target classes corresponding to graph nodes;
s2, preprocessing the electromagnetic target signal;
acquiring an electromagnetic target signal, storing the electromagnetic target signal according to acquisition time, performing short-time Fourier transform on the electromagnetic target data to acquire time-frequency data of the electromagnetic target signal, preprocessing the time-frequency data, and transforming the time-frequency data into a data format which can be processed by a convolutional neural network to be used as a sample for training the convolutional neural network;
s3, constructing a convolutional neural network, training the convolutional neural network based on the result of the embedded vector representation of the graph nodes corresponding to the electromagnetic target classes obtained by the graph neural network, and finally obtaining a reference vector for subsequently classifying and identifying the collected electromagnetic target signals; mapping of embedding expression of signal features of electromagnetic target classes to embedding vectors of nodes of electromagnetic target class corresponding graphs is achieved by utilizing a convolutional neural network;
and S4, classifying and identifying the acquired electromagnetic target signals by using the reference vector obtained in the step S3.
2. The method for electromagnetic object classification using knowledge vector embedding of claim 1,
the step S1 specifically includes:
in the graph nerve, ReLU (-) is a linear rectification function; ii | represents the norm of l 2; softmax (·) represents a logistic regression function; e is a unit matrix, N represents the dimension of the unit matrix E and is also the number of layers of the neural network of the graph; wiWeight of the ith hidden layer of the neural network, W0∈RM×N,W1∈RN×HM is the dimension of the feature vector of the graph node, H is the dimension of the embedded vector output by the graph neural network, and k is the iteration number when the graph neural network is trained;
wherein the content of the first and second substances,
express normalized pullThe phase of the,
a matrix of the degree of representation,
representation matrix
Row i, column i,
representation matrix
Row i, column j,
representing an undirected graph adjacency matrix; l isiA tag in the ith electromagnetic object class; q is an embedded vector representation of the electromagnetic object class corresponding to the graph nodes;
the method for training the graph neural network to obtain the embedded vector representation of all electromagnetic target classes corresponding to the graph nodes comprises the following specific steps:
s11, initializing a weight matrix W of the graph neural networkiN, N represents the number of layers of the neural network;
s12, calculating the output Y of the neural network of the graph, wherein the calculation formula is as follows:
wherein f () represents a calculation function of the neural network of the graph;
s13, calculating the Loss function Loss of the neural network of the graph under the output condition:
wherein L islLabels representing the ith electromagnetic object class, F represents a row in the matrix F, i.e. the eigenvectors of a certain electromagnetic object class, YfRepresenting the output obtained after the characteristic vector f is input into the neural network of the graph; l represents a set of tags of the electromagnetic object class;
s14, updating the weight of the neural network of the graph by using a batch gradient descent method (BGD) according to the loss function; adding 1 to the value of the iteration times k;
s15, repeating the process from the step S12 to the step S14 until the iteration number k reaches a preset value;
s16, calculating the embedded vector representation Q of the electromagnetic object class corresponding to the graph node, wherein the calculation formula is as follows:
thereby obtaining an embedded vector representation of the electromagnetic object class corresponding to the graph nodes.
3. The method for electromagnetic object classification using knowledge vector embedding of claim 1,
the step S3 specifically includes:
obtaining the embedded vector representation of the graph nodes corresponding to the class to which the sample belongs through the step S1, taking the embedded vector of the graph nodes corresponding to the class to which the sample belongs as a label of the convolutional neural network, taking the time-frequency data of the electromagnetic target preprocessed in the step S2 as a training sample, and taking the distance between the embedded vector of the graph nodes corresponding to the class to which the sample belongs and the embedded vector of the training sample as a loss function for training the convolutional neural network; after the convolutional neural network is subjected to repeated iterative optimization training, reference vectors for subsequently classifying and identifying the acquired electromagnetic target signals are obtained; the multiple iteration optimization training process for the convolutional neural network specifically comprises the following steps:
s31, constructing a convolutional neural network with two convolutional layers, two pooling layers and two full-connection layer structures, and initializing graph node weights of the network; the convolution layer, the pooling layer and the full-connection layer are connected in sequence;
s32, inputting the time-frequency data P of the electromagnetic target preprocessed in the step S2 into the convolutional neural network as a training sample, and obtaining a feature vector R of the training sample at the last layer of the convolutional neural network;
s33, calculating a loss function of the network: l | R-Q |, where Q is an embedded vector representation of the graph node corresponding to the class to which the input training sample belongs, and | represents the norm;
s34, optimizing the parameters of the convolutional neural network according to a random gradient descent method, and updating the weight of the graph nodes in the network;
s35, repeating the steps S32 to S34 until the iteration number reaches a preset value, and finishing the training of the convolutional neural network;
s36, after the training of the convolutional neural network is completed, the obtained output R is the final embedded vector representation result of the sample;
and S37, calculating the average value Rc of the embedded vector representation results of the sample data of the same type, and taking the Rc as a reference vector for classifying and identifying the acquired electromagnetic target signal.
4. The method for electromagnetic object classification using knowledge vector embedding of claim 1,
the step S4 includes the following steps:
s41, carrying out short-time Fourier transform and preprocessing on the acquired electromagnetic target signal to obtain a sample P 'to be recognized, and inputting the sample P' to be recognized into the convolutional neural network trained in the step S3;
s42, calculating to obtain an output vector R 'corresponding to the sample P' to be identified through a convolutional neural network;
s43, calculating the output vector R' and the embedded vector representation of the sample data of each categoryThe vector similarity of the mean of the results, the vector similarity of the output vector R' with the i-th class is denoted εiN, n is the total number of categories, and the vector similarity is obtained by calculating the Pearson correlation coefficient of two vectors;
s44, comparing the obtained vector similarity with threshold eta respectively, if epsilon existsjIf j is larger than or equal to eta, j belongs to n, judging that the sample P' to be recognized belongs to the jth electromagnetic target category; if for any epsilonjAll have epsilonjIf the sample P ' to be identified belongs to the electromagnetic target class, j belongs to n, and if the sample P ' to be identified belongs to n, the sample P ' to be identified belongs to a new electromagnetic target class.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210101982.8A CN114417938B (en) | 2022-01-27 | 2022-01-27 | Electromagnetic target classification method embedded by using knowledge vector |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210101982.8A CN114417938B (en) | 2022-01-27 | 2022-01-27 | Electromagnetic target classification method embedded by using knowledge vector |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114417938A true CN114417938A (en) | 2022-04-29 |
| CN114417938B CN114417938B (en) | 2023-06-27 |
Family
ID=81279571
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210101982.8A Active CN114417938B (en) | 2022-01-27 | 2022-01-27 | Electromagnetic target classification method embedded by using knowledge vector |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114417938B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304885A (en) * | 2023-05-11 | 2023-06-23 | 之江实验室 | Event identification method, device and equipment based on graph node embedding |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977232A (en) * | 2019-03-06 | 2019-07-05 | 中南大学 | A kind of figure neural network visual analysis method for leading figure based on power |
| CN110086737A (en) * | 2019-03-13 | 2019-08-02 | 西安电子科技大学 | A kind of recognition methods of the modulation mode of communication signal based on figure neural network |
| CN110210330A (en) * | 2019-05-13 | 2019-09-06 | 清华大学 | Electromagnetic signal recognition methods and device based on Tacit Knowledge structure figures convolutional network |
| CN110427877A (en) * | 2019-08-01 | 2019-11-08 | 大连海事大学 | A method of the human body three-dimensional posture estimation based on structural information |
| US20200387135A1 (en) * | 2019-06-06 | 2020-12-10 | Hitachi, Ltd. | System and method for maintenance recommendation in industrial networks |
-
2022
- 2022-01-27 CN CN202210101982.8A patent/CN114417938B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109977232A (en) * | 2019-03-06 | 2019-07-05 | 中南大学 | A kind of figure neural network visual analysis method for leading figure based on power |
| CN110086737A (en) * | 2019-03-13 | 2019-08-02 | 西安电子科技大学 | A kind of recognition methods of the modulation mode of communication signal based on figure neural network |
| CN110210330A (en) * | 2019-05-13 | 2019-09-06 | 清华大学 | Electromagnetic signal recognition methods and device based on Tacit Knowledge structure figures convolutional network |
| US20200387135A1 (en) * | 2019-06-06 | 2020-12-10 | Hitachi, Ltd. | System and method for maintenance recommendation in industrial networks |
| CN110427877A (en) * | 2019-08-01 | 2019-11-08 | 大连海事大学 | A method of the human body three-dimensional posture estimation based on structural information |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116304885A (en) * | 2023-05-11 | 2023-06-23 | 之江实验室 | Event identification method, device and equipment based on graph node embedding |
| CN116304885B (en) * | 2023-05-11 | 2023-08-22 | 之江实验室 | Event identification method, device and equipment based on graph node embedding |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114417938B (en) | 2023-06-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109214452B (en) | HRRP target identification method based on attention depth bidirectional cyclic neural network | |
| CN112015863B (en) | Multi-feature fusion Chinese text classification method based on graphic neural network | |
| Lopes et al. | Automatic histogram threshold using fuzzy measures | |
| CN110866530A (en) | Character image recognition method and device and electronic equipment | |
| CN111079847B (en) | Remote sensing image automatic labeling method based on deep learning | |
| CN112906720B (en) | Multi-label image identification method based on graph attention network | |
| CN113392931B (en) | Hyperspectral open set classification method based on self-supervision learning and multitask learning | |
| CN107169117B (en) | Hand-drawn human motion retrieval method based on automatic encoder and DTW | |
| CN112685504B (en) | Production process-oriented distributed migration chart learning method | |
| CN115410088B (en) | Hyperspectral image field self-adaption method based on virtual classifier | |
| CN111539444A (en) | Gaussian mixture model method for modified mode recognition and statistical modeling | |
| CN113269647A (en) | Graph-based transaction abnormity associated user detection method | |
| CN111428511A (en) | Event detection method and device | |
| CN113222072A (en) | Lung X-ray image classification method based on K-means clustering and GAN | |
| CN110991247B (en) | Electronic component identification method based on deep learning and NCA fusion | |
| CN112183237A (en) | Automatic white blood cell classification method based on color space adaptive threshold segmentation | |
| Defriani et al. | Recognition of Regional Traditional House in Indonesia Using Convolutional Neural Network (CNN) Method | |
| CN114417938B (en) | Electromagnetic target classification method embedded by using knowledge vector | |
| CN111325259A (en) | Remote sensing image classification method based on deep learning and binary coding | |
| CN108960005B (en) | Method and system for establishing and displaying object visual label in intelligent visual Internet of things | |
| CN117372144A (en) | Wind control strategy intelligent method and system applied to small sample scene | |
| CN117078007A (en) | Multi-scale wind control system integrating scale labels and method thereof | |
| CN115393631A (en) | Hyperspectral image classification method based on Bayesian layer graph convolution neural network | |
| CN112990279B (en) | Radar high-resolution range profile library outside target rejection method based on automatic encoder | |
| Yahia et al. | K-nearest neighbor and C4. 5 algorithms as data mining methods: advantages and difficulties |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |