Radiation source individual identification method and device, electronic device and storage medium

Abstract

The invention relates to a radiation source individual identification method and device, an electronic device and a storage medium, and the method comprises the steps: obtaining a to-be-identified electromagnetic signal; preprocessing the to-be-identified electromagnetic signal to obtain target input data; inputting the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features; the radiation source information encoder is obtained through generative adversarial training aiming at feature decoupling; the feature decoupling is to decouple inherent radiation source information features associated with a radiation source in an originally collected electromagnetic signal and signal information features associated with the electromagnetic signal; and performing radiation source individual identification on the target radiation source information features to obtain a radiation source individual identification result. According to the method, decoupling of radiation source information features and signal information features in electromagnetic signals can be realized, so that the accuracy and robustness of radiation source individual identification are improved.

Description

Technical Field

[0001] The present application relates to the field of electromagnetic signal processing, and particularly to a method, apparatus, electronic device, and storage medium for identifying individual radiation sources. Background Art

[0002] Identifying individual radiation sources is a technology that extracts characteristic information representing the identity of the signal transmitting device itself to identify different device individuals. In traditional methods of identifying individual radiation sources, it is usually human processing to extract the characteristics of radiation sources in a transient process (for example, the signal instability state caused by the radiation source turning on, turning off, and switching working states), such as instantaneous frequency, instantaneous phase, instantaneous amplitude, signal envelope, transient signal spectrum characteristics, etc.; or the radiation source signal characteristics in a stable working state, such as reference frequency, modulation constellation diagram, signal preamble, etc. Then, the individual recognition task is completed through feature engineering and recognition models. With the increasing complexity and precision of electronic devices and the continuous evolution of modulation technologies, the electromagnetic environment has become increasingly complex, making electromagnetic signals diverse. Based on this, the above traditional methods gradually cannot meet the accuracy requirements of the radiation source individual recognition task.

[0003] Therefore, artificial intelligence technology, especially neural network technology, has begun to be applied to the identification of individual radiation sources. Currently, algorithms for using neural network technology to perform radiation source individual recognition tasks usually consist of a feature extraction module and a classifier module. Using an end-to-end training framework, the received electromagnetic signal is input, and the individual recognition result is output. Such algorithms rely on training data and can achieve better results for test data similar to the training data, but have poor classification and recognition effects for unknown radiation source data. In addition, due to the high degree of fitting of the model to the training data, even for radiation sources similar to the training data, the test results will be greatly affected due to changes in the content of the transmitted signal, modulation method, etc. Therefore, in the current related technologies, there are still problems of low recognition accuracy and robustness in the identification of individual radiation sources that need to be solved. Summary of the Invention

[0004] In this embodiment, a method, apparatus, electronic device, and storage medium for identifying individual radiation sources are provided to solve the problems of low recognition accuracy and robustness in the identification of individual radiation sources in related technologies.

[0005] In a first aspect, in this embodiment, a method for identifying individual radiation sources is provided, including:

[0006] Obtain an electromagnetic signal to be identified;

[0007] Preprocess the electromagnetic signal to be identified to obtain target input data;

[0008] Input the target input data into the trained radiation source information encoder for feature extraction to obtain the target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; the feature decoupling is to decouple the radiation source information features inherent in the original collected electromagnetic signals associated with the radiation source itself from the signal information features associated with the electromagnetic signals themselves;

[0009] Perform radiation source individual recognition on the target radiation source information features to obtain a radiation source individual recognition result.

[0010] In some embodiments, the radiation source information encoder performs the generative adversarial training as a component of a pre-constructed radiation source individual recognition system; the radiation source individual recognition system further includes: a data pre-processor, a signal information encoder, a signal generator, a radiation source identifier, a signal discriminator, and a decoupling discriminator; where during the generative adversarial training process:

[0011] The data pre-processor is used to perform data pre-processing tasks on the received signals;

[0012] The signal information encoder is used to perform signal information feature extraction tasks;

[0013] The radiation source information encoder is used to perform radiation source information feature extraction tasks;

[0014] The signal generator is used to perform signal reconstruction tasks according to the signal information features and the radiation source information features;

[0015] The signal discriminator is used to determine whether the reconstructed signal generated by the signal generator is real;

[0016] The radiation source identifier is used to perform radiation source individual recognition tasks based on radiation source information features;

[0017] The decoupling discriminator is used to determine the category to which the radiation source information features belong.

[0018] In some embodiments, pre-processing the electromagnetic signal to be recognized to obtain target input data includes:

[0019] Based on the data pre-processor, perform matrix processing on the electromagnetic signal to be recognized to obtain the target input data.

[0020] In some embodiments, the training process of the generative adversarial training includes:

[0021] Based on the unlabeled data in the sample dataset, perform adversarial training on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator;

[0022] After completing the training based on the unlabeled data, perform supervised training on the feature extraction ability of the radiation source information encoder based on the labeled data in the sample dataset, so that the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions.

[0023] In some embodiments, based on the unlabeled data in the sample dataset, performing adversarial training on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator includes:

[0024] Based on the unlabeled data, perform iterative training on each component of the radiation source individual recognition system according to the first information decoupling loss, the first information reconstruction loss, and the first generative adversarial loss until a preset first training termination condition is reached; where

[0025] The first information decoupling loss is determined based on the output data of the radiation source information encoder;

[0026] The first information reconstruction loss is determined based on the output data of the signal generator;

[0027] The first generative adversarial loss is determined based on the output data of the radiation source information encoder.

[0028] In some embodiments, based on the labeled data in the sample dataset, performing supervised training on the feature extraction ability of the radiation source information encoder so that the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions includes:

[0029] Based on the labeled data, perform iterative training on each component of the radiation source individual recognition system according to the second information decoupling loss, the second information reconstruction loss, the second generative adversarial loss, and the radiation source classification loss until a preset second training termination condition is reached;

[0030] The second information decoupling loss is determined based on the output data of the radiation source information encoder;

[0031] The second information reconstruction loss is determined based on the output data of the signal generator;

[0032] The second generative adversarial loss is determined based on the output data of the radiation source information encoder;

[0033] The radiation source classification loss is determined based on the output data of the radiation source identifier.

[0034] In some of these embodiments, the method further includes:

[0035] Collecting sample electromagnetic signals using a number of radio stations as sample radiation sources;

[0036] Stitching the in-phase component and the quadrature component in the sample electromagnetic signal to obtain a stitched signal; segmenting the stitched signal to obtain a segmented signal; performing convolution processing on the segmented signal to obtain a sample electromagnetic signal matrix;

[0037] Using the sample electromagnetic signal matrix as the input for the generative adversarial training.

[0038] Second, in this embodiment, a radiation source individual identifier is provided, including: an acquisition module, a preprocessing module, a feature extraction module, and an identification module; where:

[0039] The acquisition module is configured to acquire the electromagnetic signal to be identified;

[0040] The preprocessing module is configured to preprocess the electromagnetic signal to be identified to obtain target input data;

[0041] The feature extraction module is configured to input the target input data into a trained radiation source information encoder for feature extraction to obtain target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; the feature decoupling is to decouple the radiation source information features inherent in the originally collected electromagnetic signal related to the radiation source itself from the signal information features related to the electromagnetic signal itself;

[0042] The identification module is configured to perform radiation source individual identification on the target radiation source information features to obtain a radiation source individual identification result.

[0043] Third, in this embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the radiation source individual identification method described in the first aspect above is implemented.

[0044] Fourth, in this embodiment, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the radiation source individual identification method described in the first aspect above is implemented.

[0045] Compared with related technologies, a method, an apparatus, an electronic device, and a storage medium for identifying an individual radiation source are provided in this embodiment. In the method for identifying an individual radiation source, first, an electromagnetic signal to be identified is obtained; then, the electromagnetic signal to be identified is preprocessed to obtain target input data; next, the target input data is input into a trained radiation source information encoder for feature extraction to obtain target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; feature decoupling is to decouple the radiation source information features inherent in the originally collected electromagnetic signal related to the radiation source itself from the signal information features related to the electromagnetic signal itself; finally, the target radiation source information features are used for identifying an individual radiation source to obtain an individual radiation source identification result. It can achieve the decoupling of the radiation source information features and the signal information features in the electromagnetic signal, and realize the identification of an individual radiation source based on the radiation source information features of the radiation source itself, thereby improving the accuracy and robustness of the individual radiation source identification.

[0046] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects, and advantages of this application more concise and understandable. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The drawings described herein are used to provide a further understanding of this application and form a part of this application. The illustrative embodiments and descriptions thereof of this application are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0048] Figure 1 is a hardware structure block diagram of a terminal of the method for identifying an individual radiation source according to an embodiment of this application;

[0049] Figure 2 is a flowchart of the method for identifying an individual radiation source according to an embodiment of this application;

[0050] Figure 3 is a schematic diagram of data flow for feature extraction according to an embodiment of this application;

[0051] Figure 4 is a schematic diagram of a data preprocessing process according to an embodiment of this application;

[0052] Figure 5 is a schematic diagram of data flow for generative adversarial training according to some embodiments of this application;

[0053] Figure 6 is a flowchart of the method for identifying an individual radiation source according to some embodiments of this application;

[0054] Figure 7 is a structure block diagram of the apparatus for identifying an individual radiation source according to an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] To understand the purpose, technical solution and advantages of the present application more clearly, the present application will be described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0056] Unless otherwise defined, the technical terms or scientific terms involved in the present application shall have the general meanings understood by those with ordinary skills in the technical field to which the present application belongs. In the present application, words such as "a", "an", "one kind", "the", "these" and the like do not indicate a limitation in quantity, and they can be singular or plural. The terms "including", "comprising", "having" and any variants thereof involved in the present application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device including a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent in these processes, methods, products or devices. The terms "connected", "coupled" and the like involved in the present application are not limited to physical or mechanical connections, but may include electrical connections, whether directly or indirectly. The "multiple" involved in the present application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. Usually, the character " / " indicates that the objects associated before and after are in an "or" relationship. The terms "first", "second", "third" and the like involved in the present application only distinguish similar objects and do not represent a specific sorting of the objects.

[0057] The method embodiment provided in this embodiment can be executed on a terminal, a computer or a similar computing device. For example, running on a terminal, Figure 1 is the hardware structure block diagram of the terminal of the radiation source individual identification method in this embodiment. As Figure 1 shown, the terminal may include one or more ( Figure 1 only one is shown in the figure) processors 102 and a memory 104 for storing data. Among them, the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that, Figure 1 the structure shown is only schematic and does not limit the structure of the above terminal. For example, the terminal may further include more or fewer components than those shown in Figure 1 the figure, or have a different configuration from that shown in Figure 1 the figure.

[0058] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the radiation source individual identification method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the terminal through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.

[0059] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by a communication provider of the terminal. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0060] In this embodiment, a radiation source individual identification method is provided. Figure 2 is a flowchart of the radiation source individual identification method of this embodiment, as Figure 2 shown, this process includes the following steps:

[0061] Step S210, obtain the electromagnetic signal to be identified.

[0062] The electromagnetic signal to be identified can specifically be an electromagnetic signal currently received and emitted by a radiation source with an unknown device identity. It is necessary to extract features from the electromagnetic signal to be identified and perform radiation source individual identification to determine the device identity of the corresponding radiation source. For example, it can be used in scenarios such as radar model identification, wireless device authentication, and vehicle identity authentication.

[0063] Step S220, preprocess the electromagnetic signal to be identified to obtain target input data.

[0064] The electromagnetic signal to be recognized collected can be adjusted in data form to form target input data adapted to feature extraction, meeting the input requirements for radiation source individual recognition. In some of these embodiments, if the electromagnetic signal to be recognized collected is an electromagnetic signal modulated by orthogonal carriers (IQ modulation), then the in-phase component (I component) and the quadrature component (Q component) in the IQ-modulated electromagnetic signal can be spliced, and then segmented into signal segments. Then, the signal segments are respectively flattened into a preset size through convolution and then composed into a signal matrix as the target input data.

[0065] Step S230: Input the target input data into the trained radiation source information encoder for feature extraction to obtain the target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; feature decoupling is to decouple the radiation source information features inherent in the original collected electromagnetic signal related to the radiation source itself from the signal information features related to the electromagnetic signal itself.

[0066] The radiation source information encoder can specifically be obtained through generative adversarial training. During the generative adversarial training process, feature decoupling is mainly performed on the radiation source information features and the signal information features in the original collected electromagnetic signal, so that the trained radiation source information encoder can accurately extract the radiation source information features from the target input data. Among them, the radiation source information features are the information features related to the radiation source in the electromagnetic signal, which can express the characteristics of the radiation source itself and focus on the inherent information of the radiation source individual. In this embodiment, the radiation source information features are features automatically learned using deep learning. Therefore, the radiation source information features extracted here do not require consistency with the physical characteristics of the radiation source or being directly understood and explained. The signal information features can be the information features related to the electromagnetic signal, such as the signal features related to parameter changes such as signal modulation mode and transmission frequency.

[0067] Generative adversarial training can be performed based on a pre-constructed generative adversarial network structure to decouple the radiation source information features and the signal information features in the received signal. Among them, during the training process, samples of electromagnetic signals can be collected to construct a sample data set. Then, using a preset iterative optimization method, iterative training is performed on each part of the generative adversarial network structure (including the above-mentioned radiation source information encoder), and the model parameters involved in each part are updated. Finally, the trained radiation source information encoder is used in the radiation source information feature extraction stage during the actual radiation source individual recognition process.

[0068] The above radiation source information encoder can be a feature extraction structure implemented based on a convolutional neural network. Exemplarily, a residual network (ResNet) can be used as the backbone network of the radiation source information encoder to adapt to multi-scale data input, followed by a feature splicing module and a feature fusion module. It can be understood that those skilled in the art can also select other structures as the radiation source information encoder according to the requirements of the actual application scenario, and this embodiment does not limit this.

[0069] In this step, a generative adversarial mechanism is used to decouple the features of the radiation source itself from the signal features of the electromagnetic signal, so as to successfully extract the radiation source information features from the electromagnetic signal for radiation source individual task recognition.

[0070] Step S240: Perform radiation source individual recognition on the target radiation source information features to obtain a radiation source individual recognition result.

[0071] The target radiation source information features can be recognized by using a pre-constructed recognition model to determine the radiation source individual recognition result. In particular, the target radiation source information features can be input into a radiation source recognizer that is synchronously trained with the radiation source information encoder for recognition to obtain the above radiation source individual recognition result. The radiation source recognizer can be implemented by using two stacked multi-layer perceptrons. It can be understood that those skilled in the art can also use other structures with feature processing and recognition capabilities to perform radiation source individual recognition.

[0072] Compared with the related technology, this embodiment enhances the attention to the inherent features of the radiation source itself. By using a generative adversarial mechanism, it decouples the inherent features of the radiation source individual and the signal features in the received signal, and uses the inherent features for the recognition task. It has universality for different forms of electromagnetic signals, excludes the interference caused by signal content, modulation method, etc. to the recognition task. For example, it is applicable to electromagnetic signals with different frequencies and different modulation methods. It can effectively extract high-quality feature representations, thereby improving the accuracy and robustness of radiation source individual recognition.

[0073] Therefore, through the above steps S210 to S240 in this embodiment, first, the electromagnetic signal to be recognized is acquired; then, the electromagnetic signal to be recognized is preprocessed to obtain the target input data; next, the target input data is input into the trained radiation source information encoder for feature extraction to obtain the target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; feature decoupling is to decouple the radiation source information features inherent in the originally collected electromagnetic signal related to the radiation source itself and the signal information features related to the electromagnetic signal itself; finally, the target radiation source information features are used for individual radiation source recognition to obtain the individual radiation source recognition result. It can achieve the decoupling of the radiation source information features and the signal information features in the electromagnetic signal, and realize individual radiation source recognition based on the radiation source information features of the radiation source itself, thereby improving the accuracy and robustness of individual radiation source recognition.

[0074] Wherein, in one embodiment, the radiation source information encoder is subjected to generative adversarial training as a component of a pre-constructed individual radiation source recognition system; the individual radiation source recognition system further includes: a data preprocessor, a signal information encoder, a signal generator, a radiation source identifier, a signal discriminator, and a decoupling discriminator; wherein during the generative adversarial training process:

[0075] The data preprocessor is used to perform data preprocessing tasks on the received signal; the signal information encoder is used to perform signal information feature extraction tasks; the radiation source information encoder is used to perform radiation source information feature extraction tasks; the signal generator is used to perform signal reconstruction tasks according to the signal information features and the radiation source information features; the signal discriminator is used to judge whether the reconstructed signal generated by the signal generator is real; the radiation source identifier is used to perform individual radiation source recognition tasks based on the radiation source information features; the decoupling discriminator is used to judge the category to which the radiation source information features belong.

[0076] In this embodiment, based on the structure of the generative adversarial network, a radiation source individual recognition system is implemented. The radiation source individual recognition system may include a data pre-processor, a signal information encoder, a radiation source information encoder, a signal generator, a radiation source recognizer, a signal discriminator, and a decoupling discriminator. Among them, the data pre-processor is used to pre-process the signal data input to the radiation source individual recognition system so that the data form and size of the signal data can meet the input requirements of the radiation source individual recognition system. The signal information encoder, as a feature extraction structure, is used to extract signal-related information features of electromagnetic signals, such as signal features related to parameter changes such as signal modulation mode and transmission frequency. The radiation source information encoder is used to obtain radiation source information features related to the inherent information of the radiation source itself in the electromagnetic signal. The signal generator is used to fuse the signal information features output by the signal information encoder and the radiation source information features output by the radiation source information encoder, and perform signal reconstruction to obtain a reconstructed signal. The radiation source recognizer is used to perform the individual recognition task of the radiation source by using the radiation source information features output by the radiation source information encoder. The signal discriminator is used to judge whether the reconstructed signal generated by the signal generator is real enough. The decoupling discriminator is used to judge the category to which the features extracted by the radiation source information encoder belong, for example, to judge whether the features extracted by the radiation source information encoder are radiation source information features or signal information features.

[0077] The above signal information encoder and radiation source information encoder may adopt the same structure and have a similar feature extraction process. Exemplarily, both the signal information encoder and the radiation source information encoder may include a backbone network, a feature splicing module, and a feature fusion module. The backbone network may be ResNet. Figure 3 A schematic diagram of the data flow of feature extraction for this embodiment. The Figure 3 feature extraction process can be adapted to the feature extraction of the signal information encoder and the feature extraction of the radiation source information encoder. As Figure 3 shown, the signal processed by the data pre-processor is input to the backbone network in matrix form ( Figure 3 matrix 1, matrix 2, matrix 3, matrix 4 in the figure). After being processed by the backbone network, corresponding features (feature 1, feature 2, feature 3, feature 4) are output for each input matrix. Among them, matrices 1 to 4 are different-scale processes of the same segment of signal to obtain multi-scale information. For example, the signal is divided into a segments and then spliced together to form matrix 1, and the signal is divided into b segments and then spliced together to form matrix 2, and so on. Different segmentations and splicings of the signal can make the content covered in the receptive field different during the convolution calculation, so as to capture the connection between different time series segments of the signal.

[0078] Subsequently, each feature output by the backbone network is input into the feature splicing module for feature splicing. The direct splicing method can be adopted to obtain the spliced features. Then, the features output by the splicing feature module are passed through the fully connected layer, which is a feature fusion module, to fuse multi-scale feature information and obtain the finally output signal information features or radiation source information features.

[0079] Exemplarily, the above signal generator can specifically adopt a deconvolution neural network with a residual connection structure. The above radiation source identifier can adopt two stacked multi-layer perceptrons. The above signal discriminator and decoupling discriminator can be a structure of a convolutional network plus a fully connected layer.

[0080] It can be understood that those skilled in the art can also select other network structures as the above signal generator, signal discriminator, signal discriminator, decoupling discriminator, and radiation source identifier according to the requirements of the actual application scenario. This embodiment does not make specific limitations in this regard.

[0081] In one embodiment, preprocessing the electromagnetic signal to be recognized to obtain target input data may specifically include: based on a data preprocessor, performing matrix processing on the electromagnetic signal to be recognized to obtain target input data. The signal matrix can be formed as the target input data by performing splicing, segmentation, and convolution processing on the electromagnetic signal to be recognized to adapt to the input of the subsequent signal information encoder and radiation source information encoder.

[0082] Additionally, in one embodiment, the training process of the above generative adversarial training may include:

[0083] Based on the unlabeled data in the sample dataset, perform adversarial training on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator; after completing the training based on the unlabeled data, based on the labeled data in the sample dataset, perform supervised training on the feature extraction ability of the radiation source information encoder to make the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions.

[0084] In this embodiment, the generative adversarial training is divided into two parts. The first part is the pre-training based on unlabeled data. The training of the first part mainly utilizes the decoupling and recombination of data features and uses the generative adversarial network to implement the training of the above signal information encoder, radiation source information encoder, and signal generator. The second part of the training is the supervised fine-tuning using labeled data based on the training of the first part.

[0085] Among them, for the training of the first part, the signal information encoder and the radiation source information encoder are mainly used to extract features from the input signal, decouple the signal information features and the radiation source information features, and use the decoupling discriminator to judge the category to which the features extracted by the radiation source information encoder belong (as radiation source information features or signal information features), so as to constrain the generation of the radiation source information features to meet the requirements of feature decoupling, and retain the original information of the signal, so that it can be used by the signal generator to reconstruct the original signal. Whether the signal reconstructed by the signal generator is real enough is judged by the signal discriminator (it can be judged by comparing the similarity with the real signal). In this way, the performance of the signal generator, the radiation source information encoder, and the signal information encoder is improved through the adversarial generation form. After training, the feature extraction capabilities of the signal information encoder and the radiation source information encoder are finally improved, and the decoupled signal information features and radiation source information features are obtained.

[0086] For the training of the second part, it is necessary to make the radiation source information features after decoupling the original signal meet the requirements of radiation source individual recognition. For example, the accuracy of radiation source individual recognition based on the extracted radiation source information features reaches the preset accuracy requirement. Among them, during the training process of the second part, a radiation source identifier can be used to perform the radiation source individual recognition task. Therefore, a classification loss can be added to the loss constructed in the first part of the training to form the total loss of the second part, and the extraction of the radiation source information features is further constrained, so that the radiation source information features extracted by the radiation source information encoder can ultimately be used for the radiation source individual recognition task.

[0087] In the existing algorithms for radiation source individual recognition tasks using neural network technology, the processes of collecting, annotating, and processing relevant data all require the participation of operators, thus limiting the scale of the dataset, especially the scale of labeled data, and further limiting the recognition performance of the model. In addition, due to the high degree of fitting of the model to the training data, even for radiation sources similar to the training data, the test results will be greatly affected due to changes in the content of the transmitted signal, modulation method, etc.

[0088] In the training process of the first part of this embodiment, unlabeled data is fully utilized, solving the problems of difficult data annotation and limited model performance in related technologies, and enhancing the generalization ability of the model. Therefore, compared with related technologies, the dependence on labeled data is reduced, and the limitation on the performance of radiation source individual recognition is removed. In the training of the second part, based on the supervised fine-tuning of labeled data, a small amount of labeled data is used to adaptively adjust the model parameters, further enhancing the recognition accuracy of the model.

[0089] Among them, in one embodiment, based on the unlabeled data in the sample dataset, adversarial training is performed on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator. Specifically, it may include:

[0090] Based on the unlabeled data, according to the first information decoupling loss, the first information reconstruction loss, and the first generative adversarial loss, each component of the radiation source individual recognition system is iteratively trained until a preset first training termination condition is reached; among them, the first information decoupling loss is determined based on the output data of the radiation source information encoder; the first information reconstruction loss is determined based on the output data of the signal generator; the first generative adversarial loss is determined based on the output data of the radiation source information encoder.

[0091] Among them, in each iteration round, the signal data pair after being processed by the data preprocessor in the unlabeled data is input into the signal information encoder for signal information feature extraction to obtain a first output feature pair, and is input into the radiation source information encoder for radiation source information feature extraction to obtain a second output feature pair; after cross-combining the first output feature pair and the second output feature, it is input into the signal generator for signal reconstruction to obtain a first output signal pair; the first output signal pair is input into the signal information encoder for signal information feature extraction to obtain a third output feature pair, and is input into the radiation source information encoder for radiation source information feature extraction to obtain a fourth output feature pair; the third output feature pair and the fourth output feature pair are input into the signal generator for signal reconstruction to obtain a second output signal pair. Finally, based on the radiation source information features in the first output feature pair and the second output feature pair, the above-mentioned first information decoupling loss and generative adversarial loss are determined, and based on the second output signal pair and the signal pair processed by the data preprocessor, the signal reconstruction loss is determined.

[0092] Among them, the training process of the first part is specifically as follows: First, considering that the first part does not involve the radiation source individual recognition task, the parameters of the radiation source recognizer are frozen to reduce the calculation amount and speed up the model inference speed. Then, training signal data pairs are randomly selected from the training set of the sample dataset and processed into matrix form by the data preprocessor, for example, , are respectively input into the signal information encoder and the radiation source information encoder for feature extraction to obtain signal information features , (the first output feature pair) and radiation source information features , (the second output feature pair); cross-combine the two pairs of feature pairs, combine , , combine , Combine and input into the signal generator to obtain the reconstructed signal and (the first output signal pair), and then and Input signal information encoder and radiation source information encoder to obtain the signal information feature , (the third output feature pair) and radiation source information feature , (the fourth output feature pair); Input , and , into the input signal generator to obtain the reconstructed signal and (the second output signal pair).

[0093] Complete backpropagation and parameter update by calculating each loss and gradient, thereby completing the training of the first part. Among them, in this embodiment, the loss calculation of the first part of the training includes the first information decoupling loss, the first information reconstruction loss, and the first generative adversarial loss. Among them:

[0094] The calculation method of the first information decoupling loss is:

[0095]

[0096] ;

[0097] The calculation method of the signal reconstruction loss is:

[0098] ;

[0099] The calculation method of the generative adversarial loss is:

[0100]

[0101] ;

[0102] Finally, the total loss of the first part is:

[0103] .

[0104] The above "#" indicates the connection of formula segments.

[0105] In the training of the first part, a signal information encoder is used to extract signal information features, and a radiation source information encoder is used to extract radiation source information features. By using a cross-combination method and combining with a generative adversarial network, the effectiveness of the features can be guaranteed, and relevant loss functions are used to constrain the above feature extraction during the training process, so as to ultimately achieve the purpose of feature decoupling. The above is the output of the decoupling discriminator, and is the output of the signal discriminator. The specific forms of the decoupling discriminator and the signal discriminator are not limited.

[0106] Additionally, in one embodiment, based on the labeled data in the sample dataset, the feature extraction ability of the radiation source information encoder is supervised and trained so that the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions. Specifically, it may include:

[0107] Based on the labeled data, according to the second information decoupling loss, the second information reconstruction loss, the second generative adversarial loss, and the radiation source classification loss, each component of the radiation source individual recognition system is iteratively trained until the preset second training termination condition is reached; the second information decoupling loss is determined based on the output data of the radiation source information encoder; the second information reconstruction loss is determined based on the output data of the signal generator; the second generative adversarial loss is determined based on the output data of the radiation source information encoder; the radiation source classification loss is determined based on the output data of the radiation source identifier.

[0108] Among them, the training of the second part uses the labeled data for supervised fine-tuning. In each iteration round:

[0109] Randomly extract training signal data pairs from the training set and input them into the aforementioned radiation source individual recognition system. After being processed by the data preprocessor, ; Respectively pass through the signal information encoder and the radiation source information encoder to obtain the signal information features , and the radiation source information features , ; Cross-combine the two pairs of features, and input , and , into the signal generator to obtain the reconstructed signals and , and then input and into the signal information encoder and the radiation source information encoder to obtain the signal information features , And the radiation source information features 、 ; Combine 、 and 、 input into a signal generator to obtain a reconstructed signal and ; Input the radiation source information features 、 、 、 into a radiation source identifier to obtain classification judgments p1, p2 and 、 .

[0110] The calculation of the loss in the second part of the training includes the second information decoupling loss, the second information reconstruction loss, the second generative adversarial loss, and the radiation source classification loss. Among them, the calculation process of the second information decoupling loss is as follows:

[0111]

[0112] ;

[0113] The calculation process of the second signal reconstruction loss is as follows:

[0114] ;

[0115] The calculation process of the second generative adversarial loss is as follows:

[0116]

[0117] ;

[0118] The calculation process of the radiation source classification loss is as follows:

[0119] ;

[0120] Finally, the total loss of the second part of the training is:

[0121] .

[0122] Additionally, in one embodiment, the above radiation source individual identification method may further include:

[0123] Collect sample electromagnetic signals using several radio stations as sample radiation sources; splice the in-phase component and the quadrature component in the sample electromagnetic signals to obtain a spliced signal; segment the spliced signal to obtain a segmented signal; perform convolution processing on the segmented signal to obtain a sample electromagnetic signal matrix; use the sample electromagnetic signal matrix as the input for generative adversarial training.

[0124] When collecting sample electromagnetic signals, the collection conditions can be set to direct radiation and diffraction, and various settings can be made for parameters such as frequency band and modulation mode during collection. The collected signal can be an IQ modulated electromagnetic signal. The sample data set can be divided into a training set, a validation set, and a test set, for example, in a ratio of 3:1:1. If a total of 15,000 data are collected, the data in the training set, validation set, and test set are 9,000, 3,000, and 3,000, respectively.

[0125] In the data preprocessor P, the acquired IQ modulated electromagnetic signal is preprocessed: for an IQ signal with a length of L, the signal , The components are spliced ​​to obtain a spliced ​​signal :

[0126] ;

[0127] in Quantity and The size of the components is (1, L). The signal after splicing The size of the signal is (2, L), then Divide into [K1, K2, K3, K4] segments, each segment size is (2, N i ), satisfying the relationship:

[0128] ;

[0129] The signal segments are flattened into sizes (1, N i ), the signal fragments are reassembled into the input matrix As the input of the radiation source information encoder and the signal information encoder, this data preprocessing can mine the correlation between the I / Q components of the same time sequence segment and between different time sequence segments, and realize multi-scale fusion of data.

[0130] Figure 4 FIG. 1 is a schematic diagram of a data preprocessing process of this embodiment. Figure 4 As shown, for the input signal a, the I component and the Q component are concatenated to obtain a concatenated signal a with a size of (2, L). Then, the concatenated signal a is segmented to obtain several signal segments, such as Figure 4 K1 to K4 in the i ), i is one of 1, 2, 3, and 4. Then, the size of each signal segment is flattened, and the size of each signal is (1, N i ). The signal fragments are combined into matrices to obtain matrix 1, matrix 2, matrix 3, and matrix 4.

[0131] Figure 5 Schematic diagram of data flow for generative adversarial training of some of the embodiments, as Figure 5 shown. Input signal 1 and input signal 2 are input as a pair of signals into a data pre-processor. After data pre-processing, processed signal 1 ( Figure 5 solid arrow in the data pre-processor pointing to the signal information encoder and the radiation source information encoder in Figure 5 ) and processed signal 2 ( dashed arrow in the data pre-processor pointing to the signal information encoder and the radiation source information encoder in ) are obtained. Processed signal 1 and processed signal 2 are respectively input into the signal information encoder and the radiation source information encoder for feature extraction processing. The signal 1 signal information output by the signal information encoder (solid arrow from the upper-layer signal information encoder to the upper-layer signal generator), signal 2 signal information (dashed arrow from the upper-layer signal information encoder to the upper-layer signal generator), together with the signal 1 radiation source information output by the radiation source information encoder (solid arrow from the upper-layer radiation source information encoder to the upper-layer signal generator and the upper-layer decoupled discriminator), signal 2 radiation source information (dashed arrow from the upper-layer radiation source information encoder to the upper-layer signal generator and the upper-layer decoupled discriminator) are cross-combined and then input into the signal generator for signal reconstruction, obtaining signal 1' (dashed arrow from the upper-layer signal generator to the lower-layer signal information encoder and the lower-layer radiation source information encoder), signal 2' (solid arrow from the upper-layer signal generator to the lower-layer signal information encoder and the lower-layer radiation source information encoder). The signal 1 radiation source information and signal 2 radiation source information output by the radiation source information encoder are input into the decoupled discriminator for judging the category to which the features belong. Signal 1' and signal 2' are input into the signal discriminator to verify the authenticity of the signals. Additionally, signal 1' and signal 2' are respectively input into the signal information encoder and the radiation source information encoder for feature extraction, and the obtained signal 1' radiation source information (dashed arrow from the lower-layer radiation source information encoder to the lower-layer signal generator and the lower-layer decoupled discriminator), signal 2' radiation source information (solid arrow from the lower-layer radiation source information encoder to the lower-layer signal generator and the lower-layer decoupled discriminator) are input into the decoupled discriminator for judging the category to which the features belong. Then, the signal 1' signal information (dashed arrow from the lower-layer signal information encoder to the lower-layer signal generator), signal 2' signal information (solid arrow from the lower-layer signal information encoder to the lower-layer signal generator), signal 1' radiation source information, and signal 2' radiation source information are cross-combined and then input into the signal generator for signal reconstruction, obtaining reconstructed signal and reconstructed signal Input it into the signal discriminator for authenticity verification. Additionally, based on the radiation source identifier, perform individual radiation source identification on the radiation source information of signal 1, signal 2, signal 1', and signal 2' output by the radiation source information encoder, obtain the radiation source identification result, and verify the accuracy of the identification result based on the radiation source identification result and the corresponding radiation source identification label.

[0132] Figure 6 It is a flowchart of the method for individual radiation source identification in some embodiments. As Figure 6 shown, the method for individual radiation source identification includes the following steps:

[0133] Step S601, collect samples to construct an electromagnetic signal data set; among them, use multiple ultra-short wave radio stations as radiation sources to collect electromagnetic signals to construct an electromagnetic signal data set.

[0134] Step S602, construct an individual radiation source identification system; the composition of this individual radiation source identification system can refer to the above embodiments.

[0135] Step S603, based on the generative adversarial mechanism, train the individual radiation source identification system based on feature decoupling. The specific training process can refer to the above embodiments and will not be elaborated here.

[0136] Step S604, after training is completed, use a data pre-processor to pre-process the electromagnetic signal to be identified to obtain target input data.

[0137] Step S605, input the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features.

[0138] Step S606, perform individual radiation source identification on the target radiation source information features to obtain the individual radiation source identification result.

[0139] In this embodiment, an individual radiation source identification device is also provided. This device is used to implement the above embodiments and preferred implementation manners, and those that have been described will not be elaborated again. The following terms "module", "unit", "sub-unit", etc. can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0140] Figure 7 It is a structural block diagram of the individual radiation source identification device 70 in this embodiment. As Figure 7 shown, the individual radiation source identification device 70 includes: an acquisition module 72, a pre-processing module 74, a feature extraction module 76, and an identification module 78; where:

[0141] An acquisition module 72 for acquiring the electromagnetic signal to be recognized;

[0142] A preprocessing module 74 for preprocessing the electromagnetic signal to be recognized to obtain target input data; A feature extraction module 76 for inputting the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features; The radiation source information encoder is obtained through generative adversarial training for feature decoupling; Feature decoupling is to decouple the radiation source information features inherent in the original collected electromagnetic signal associated with the radiation source itself from the signal information features associated with the electromagnetic signal itself; An identification module 78 for performing radiation source individual identification on the target radiation source information features to obtain a radiation source individual identification result.

[0143] It should be noted that the above-mentioned various modules can be functional modules or program modules, and can be implemented either by software or by hardware. For the modules implemented by hardware, the above-mentioned various modules can be located in the same processor; or the above-mentioned various modules can also be located in different processors in any combined form.

[0144] Specific examples in this embodiment can refer to the examples described in the above-mentioned embodiment and optional implementation manners, and will not be elaborated herein.

[0145] In this embodiment, an electronic device is further provided, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0146] Optionally, the above-mentioned electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the above-mentioned processor, and the input / output device is connected to the above-mentioned processor.

[0147] Optionally, in this embodiment, the above-mentioned processor may be configured to execute the following steps through a computer program:

[0148] S1. Acquire the electromagnetic signal to be recognized;

[0149] S2. Preprocess the electromagnetic signal to be recognized to obtain target input data;

[0150] S3. Input the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features; The radiation source information encoder is obtained through generative adversarial training for feature decoupling; Feature decoupling is to decouple the radiation source information features inherent in the original collected electromagnetic signal associated with the radiation source itself from the signal information features associated with the electromagnetic signal itself;

[0151] S4. Perform individual identification of the target radiation source information features to obtain the individual identification result of the radiation source.

[0152] Correspondingly, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation manners, and will not be elaborated herein.

[0153] In addition, in combination with the radiation source individual identification method provided in the above embodiments, a storage medium can also be provided in this embodiment to implement it. A computer program is stored on the storage medium; when the computer program is executed by a processor, any one of the radiation source individual identification methods in the above embodiments is implemented.

[0154] It should be understood that the specific embodiments described here are only used to explain this application, rather than to limit it. According to the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of this application.

[0155] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties.

[0156] Obviously, the accompanying drawings are only some examples or embodiments of this application. For those of ordinary skill in the art, this application can also be applied to other similar situations based on these drawings without creative work. In addition, it can be understood that although the work done during this development process may be complex and time-consuming, for those of ordinary skill in the art, certain design, manufacturing, or production changes based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient disclosure of this application.

[0157] The term "embodiment" in this application means that the specific features, structures, or characteristics described in combination with the embodiment can be included in at least one embodiment of this application. The phrase appears in various positions in the specification does not necessarily mean the same embodiment, nor does it mean being independent or alternative to other embodiments and mutually exclusive. Those of ordinary skill in the art can clearly or implicitly understand that the embodiments described in this application can be combined with other embodiments without conflict.

[0158] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of patent protection. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for identifying a radiation source individual, characterized in that, Including: Obtain the electromagnetic signal to be recognized; Preprocess the electromagnetic signal to be recognized to obtain target input data; Input the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; The feature decoupling is to decouple the radiation source information features inherent in the originally collected electromagnetic signal related to the radiation source itself and the signal information features related to the electromagnetic signal itself; Perform radiation source individual recognition on the target radiation source information features to obtain a radiation source individual recognition result.

2. The method for identifying an individual radiation source according to claim 1, wherein The radiation source information encoder is used as a component of a pre-constructed radiation source individual recognition system to perform the generative adversarial training; The radiation source individual recognition system further includes: a data preprocessor, a signal information encoder, a signal generator, a radiation source identifier, a signal discriminator, and a decoupling discriminator; where during the generative adversarial training process: The data preprocessor is used to perform data preprocessing tasks on the received signal; The signal information encoder is used to perform signal information feature extraction tasks; The radiation source information encoder is used to perform radiation source information feature extraction tasks; The signal generator is used to perform signal reconstruction tasks according to the signal information features and the radiation source information features; The signal discriminator is used to determine whether the reconstructed signal generated by the signal generator is real; The radiation source identifier is used to perform radiation source individual recognition tasks based on radiation source information features; The decoupling discriminator is used to determine the category to which the radiation source information features belong.

3. The radiation source individual identification method according to claim 2, wherein, Preprocessing the electromagnetic signal to be recognized to obtain target input data includes: Based on the data preprocessor, perform matrix processing on the electromagnetic signal to be recognized to obtain the target input data.

4. The method for identifying an individual radiation source according to claim 2, wherein The training process of the generative adversarial training includes: Based on the unlabeled data in the sample dataset, perform adversarial training on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator; After completing the training based on the unlabeled data, perform supervised training on the feature extraction ability of the radiation source information encoder based on the labeled data in the sample dataset, so that the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions.

5. The method for identifying an individual radiation source according to claim 4, characterized in that, Based on the unlabeled data in the sample dataset, performing adversarial training on the feature extraction ability of the signal information encoder, the feature extraction ability of the radiation source information encoder, and the signal reconstruction ability of the signal generator includes: Based on the unlabeled data, perform iterative training on each component of the radiation source individual recognition system according to the first information decoupling loss, the first information reconstruction loss, and the first generative adversarial loss until a preset first training termination condition is reached; where, The first information decoupling loss is determined based on the output data of the radiation source information encoder; The first information reconstruction loss is determined based on the output data of the signal generator; The first generative adversarial loss is determined based on the output data of the radiation source information encoder.

6. The radiation source individual identification method according to claim 4, characterized in that Based on the labeled data in the sample dataset, the feature extraction ability of the radiation source information encoder is supervised and trained to make the radiation source information features extracted by the radiation source information encoder meet the preset radiation source individual recognition conditions, including: Based on the labeled data, according to the second information decoupling loss, the second information reconstruction loss, the second generative adversarial loss, and the radiation source classification loss, each component of the radiation source individual recognition system is iteratively trained until a preset second training termination condition is reached; The second information decoupling loss is determined based on the output data of the radiation source information encoder; The second information reconstruction loss is determined based on the output data of the signal generator; The second generative adversarial loss is determined based on the output data of the radiation source information encoder; The radiation source classification loss is determined based on the output data of the radiation source identifier.

7. The method for identifying an individual radiation source according to any one of claims 1 to 6, characterized in that, The method further includes: Collecting sample electromagnetic signals using several radio stations as sample radiation sources; Splicing the in-phase component and the quadrature component in the sample electromagnetic signal to obtain a spliced signal; segmenting the spliced signal to obtain a segmented signal; performing convolution processing on the segmented signal to obtain a sample electromagnetic signal matrix; Using the sample electromagnetic signal matrix as the input of the generative adversarial training.

8. A radiation source individual identification device, characterized in that, Including: An acquisition module, a preprocessing module, a feature extraction module, and an identification module; wherein: The acquisition module is configured to acquire an electromagnetic signal to be identified; The preprocessing module is configured to preprocess the electromagnetic signal to be identified to obtain target input data; The feature extraction module is configured to input the target input data into the trained radiation source information encoder for feature extraction to obtain target radiation source information features; the radiation source information encoder is obtained through generative adversarial training for feature decoupling; the feature decoupling is to decouple the radiation source information features inherent in the originally collected electromagnetic signal related to the radiation source itself from the signal information features related to the electromagnetic signal itself; The identification module is configured to perform radiation source individual identification on the target radiation source information features to obtain a radiation source individual identification result.

9. An electronic device, comprising a memory and a processor, characterized in that, A computer program is stored in the memory, and the processor is configured to run the computer program to execute the radiation source individual identification method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the radiation source individual identification method according to any one of claims 1 to 7 are implemented.

Images