Method, device, equipment, medium and product for identifying radar emitter individuals

Abstract

The application provides a radar emitter individual identification method, device, equipment, medium and product. The method comprises the following steps: obtaining an original signal of a radar emitter individual to be identified; K signal sets are extracted from the original signal, and time sequence change characteristics of each signal set are obtained respectively, each signal set comprises a plurality of sub-signals, the time sequence change characteristics comprise time sequence change characteristics in a cycle of each of the plurality of sub-signals and time sequence change characteristics between cycles of the plurality of sub-signals, and K is a positive integer; and a classification result of the radar emitter individual is determined based on the K time sequence change characteristics. The scheme can improve the identification accuracy of the radar emitter individual.

Description

Technical Field

[0001] This application relates to the field of signal analysis technology, and in particular to a method, apparatus, device, medium and product for identifying individual radar radiation sources. Background Technology

[0002] Radar radiation source identification refers to the process of measuring the characteristics of received electromagnetic signals and determining the individual radiation source that generated the signal based on existing prior information.

[0003] Currently, radar radiation source individual identification schemes extract features from radar signals from a single perspective, such as the time domain, frequency domain, or time-frequency domain, and classify them based on the extracted features to obtain radar radiation source individual identification results.

[0004] However, radar signals are often the result of multiple superimposed processes. This inherent multi-period property makes the changes in radar signals extremely complex. Therefore, the accuracy of using the above-mentioned scheme for individual identification of radar radiation sources is relatively low. Summary of the Invention

[0005] Based on the above-mentioned technological status, this application proposes a method, device, equipment, medium, and product for identifying individual radar radiation sources, which can improve the accuracy of radar radiation source identification results.

[0006] To achieve the above-mentioned technical objectives, this application proposes the following technical solution: The first aspect of this application proposes a method for identifying individual radar radiation sources, comprising: acquiring the original signal of the individual radar radiation source to be identified; extracting K signal sets from the original signal, and acquiring the temporal variation characteristics of each signal set, wherein each signal set includes multiple sub-signals, and the temporal variation characteristics include the intra-period temporal variation characteristics of each of the multiple sub-signals and the inter-period temporal variation characteristics between the multiple sub-signals, wherein K is a positive integer; and determining the classification result of the individual radar radiation source based on the K temporal variation characteristics.

[0007] In some implementations, K signal sets are extracted from the original signal, and the temporal variation characteristics of each signal set are obtained. The temporal variation characteristics of each signal set are fused to obtain the fused signal characteristics corresponding to the original signal. Based on the fused signal characteristics, the classification result of the individual radar radiation source is determined. This includes: inputting the original signal into a radar radiation source individual classification model so that the radar radiation source individual classification model extracts K signal sets from the original signal, obtains the temporal variation characteristics of each signal set, and determines the classification result of the individual radar radiation source based on the K temporal variation characteristics.

[0008] In some implementations, the original signal is a time-domain signal, and the step of extracting a set of K signals from the original signal includes: converting the original signal to the frequency domain to obtain a frequency-domain signal corresponding to the original signal; and extracting the set of K signals from the original signal based on the K frequency components with the largest frequency intensity in the frequency-domain signal.

[0009] In some implementations, determining the classification result of the individual radar radiation source based on K temporal variation features includes: fusing the K temporal variation features to obtain the fused signal features corresponding to the original signal; and determining the classification result of the individual radar radiation source based on the fused signal features.

[0010] In some implementations, the fused signal feature corresponding to the original signal is obtained by fusing the K time-series variation features, including: fusing the K time-series variation features to obtain an initial fused signal feature, and determining the initial fused signal feature as the fused signal feature corresponding to the original signal; or, fusing the K time-series variation features to obtain an initial fused signal feature, and determining the residual result between the initial fused signal feature and the original signal as the intermediate fused signal feature corresponding to the original signal, and if the number of iterations has not reached a preset number, using the intermediate fused signal feature as the new original signal for the next iteration, and repeatedly extracting K signal sets from the original signal, obtaining the time-series variation features of each signal set, fusing the time-series variation features of each signal set to obtain an initial fused signal feature, and determining the residual result between the initial fused signal feature and the original signal as the intermediate fused signal feature corresponding to the original signal, until the number of iterations reaches a preset number, and determining the intermediate fused signal feature obtained in the last iteration as the fused signal feature corresponding to the original signal.

[0011] In some implementations, fusing the K time-series variation features to obtain initial fused signal features includes: determining the weights of the K time-series variation features based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain signal corresponding to the original signal, wherein the magnitude of each frequency component is positively correlated with the weight of the time-series variation feature corresponding to each frequency component; and performing a weighted summation based on the K time-series variation features and their respective weights to obtain the initial fused signal features.

[0012] In some implementations, the training process of the radar radiation source individual classification model includes: extracting K sample signal sets from the acquired original sample signals; acquiring the time-series variation sample features of each sample signal set, wherein each sample signal set includes multiple sub-sample signals, and the time-series variation sample features include the time-series variation features within the period of each of the multiple sub-sample signals and the time-series variation features between the multiple sub-sample signals during the period; and training a preset radar radiation source individual classification network based on the K time-series variation sample features to obtain the radar radiation source individual classification model.

[0013] In some implementations, the original sample signal is a time-domain sample signal, and the step of extracting a set of K sample signals from the acquired original sample signal includes: converting the original sample signal to the frequency domain to obtain a frequency-domain sample signal corresponding to the original sample signal; and extracting the set of K sample signals from the original sample signal based on the K frequency components with the largest frequency intensity in the frequency-domain sample signal.

[0014] In some implementations, the step of training a preset radar source individual classification network based on K time-series variation features to obtain a radar source individual classification model includes: fusing the K time-series variation sample features to obtain fused sample signal features corresponding to the original sample signals; and training the preset radar source individual classification network based on the fused sample signal features to obtain a radar source individual classification model.

[0015] In some implementations, fusing the K time-varying sample features to obtain the fused sample signal features corresponding to the original signal includes: fusing the K time-varying sample features to obtain initial sample signal fusion features, and determining the initial sample signal fusion features as the fused sample signal features corresponding to the original sample signal; or, fusing the K time-varying sample features to obtain initial sample signal fusion features, and determining the residual result between the initial sample signal fusion features and the original sample signal as the intermediate sample signal fusion features corresponding to the original sample signal, and, if the number of iterations has not reached a preset number, then... The intermediate sample signal fusion feature is used as the new original sample signal in the next iteration. K sample signal sets are extracted from the original sample signal, and the time-series variation sample features of each sample signal set are obtained. The initial sample signal fusion feature is obtained by fusing the time-series variation sample features of each sample signal set. The residual result of the initial sample signal fusion feature and the original sample signal is determined as the intermediate sample signal fusion feature corresponding to the original sample signal. The iteration count is repeated until the preset number of iterations is reached. The intermediate sample signal fusion feature obtained in the last iteration is determined as the fusion sample signal feature corresponding to the original sample signal.

[0016] In some implementations, fusing the time-varying sample features of the K sample signal sets to obtain initial sample signal fusion features includes: determining the weights of the K time-varying sample features based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain sample signals corresponding to the original sample signals, wherein the magnitude of each frequency component is positively correlated with the weight of the time-varying sample feature corresponding to each frequency component; and performing a weighted summation based on the K time-varying sample features and their respective weights to obtain the initial sample signal fusion features.

[0017] In some implementations, training a preset radar source individual classification network based on the fused sample signal features to obtain a radar source individual classification model includes: classifying the fused sample signal features using the preset radar source individual classification network to obtain a predicted classification result for the radar source individual; determining a classification loss based on the difference between the predicted classification result and its corresponding true classification result; and converging the preset radar source individual classification network based on the classification loss to obtain the radar source individual classification model.

[0018] A second aspect of this application provides a device for identifying individual radar radiation sources, comprising: an acquisition unit for acquiring the original signal of the individual radar radiation source to be identified; a feature extraction unit for extracting K signal sets from the original signal and acquiring the temporal variation features of each signal set, wherein each signal set includes multiple sub-signals, and the temporal variation features include the intra-period temporal variation features of each of the multiple sub-signals and the inter-period temporal variation features between the multiple sub-signals, wherein K is a positive integer; and a classification unit for determining the classification result of the individual radar radiation source based on the K temporal variation features.

[0019] A third aspect of this application provides an electronic device, including a memory and a processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the method for identifying individual radar radiation sources as described in the first aspect and any of the implementations of the first aspect by running a program in the memory.

[0020] The fourth aspect of this application provides a storage medium storing a computer program, which, when executed by a processor, implements the method for identifying individual radar radiation sources as described in the first aspect and any of its implementations.

[0021] The fifth aspect of this application provides a computer program product, including computer program instructions, which, when executed by a processor, cause the processor to implement the method for identifying individual radar radiation sources as described in the first aspect and any of the implementations of the first aspect.

[0022] The radar source identification method, apparatus, equipment, medium, and product proposed in this application extract K signal sets from the original signal and obtain the temporal variation characteristics of each signal set. Each signal set includes multiple sub-signals. The temporal variation characteristics include the intra-period temporal variation characteristics of each sub-signal and the inter-period temporal variation characteristics between the multiple sub-signals. Thus, the radar source individual is classified based on the intra-period temporal variation characteristics and inter-period temporal variation characteristics of the original signal, thereby improving the accuracy of radar source individual classification. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating the individual identification method for radar radiation sources in related technologies.

[0025] Figure 2 This is a flowchart illustrating the method for identifying individual radar radiation sources provided in an embodiment of this application.

[0026] Figure 3 This is a schematic diagram illustrating the time-series variation characteristics within a period and during a week, as provided in the embodiments of this application.

[0027] Figure 4 This is a flowchart illustrating the training process of the radar radiation source individual classification model provided in the embodiments of this application.

[0028] Figure 5 This is a schematic diagram illustrating the overall implementation process of the radar radiation source identification method provided in the embodiments of this application.

[0029] Figure 6 This is a schematic diagram of the structure of a radar radiation source identification device provided in an embodiment of this application.

[0030] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0031] The technical solutions proposed in this application are applicable to radar radiation source individual identification application scenarios. By adopting the technical solutions in this application, the accuracy of radar radiation source individual identification can be improved.

[0032] Radar technology, as a primary passive detection method, has wide applications in national defense, civilian sectors, aerospace, and other fields. Individual radar radiation source identification technology can accurately distinguish between different radar radiation sources, thereby enabling the identification and location of targets.

[0033] In recent years, with the rapid development of deep learning technology in the field of computer vision, radar radiation source individual recognition technology based on deep learning has gradually emerged. It extracts features from individual radar radiation sources and uses these extracted features to identify the individual radar radiation sources. The process is described in detail below with reference to the accompanying diagram: Figure 1 This is a flowchart illustrating the individual radar radiation source identification method in related technologies. For example... Figure 1 As shown, the radar radiation source individual identification method includes four main steps: (1) Radar signal preprocessing By performing signal preprocessing operations such as pulse detection and segmentation on the received radar signals, they are made easier to extract features in the subsequent process.

[0034] (2) Feature extraction By extracting artificially designed radar "fingerprint" features from preprocessed radar signals, such as single-cycle features like pulse rising edge and high-order cumulant features extracted from a single perspective like the time domain, frequency domain, or time-frequency domain based on signal processing methods, as well as high-level latent representation vectors extracted based on unsupervised representation learning neural networks.

[0035] (3) Model classification The features extracted in step (2) are fed into the backend classifier for identification, thereby obtaining the individual identification results of the radar radiation source. The backend classifier can be a traditional machine learning algorithm such as support vector machine, or a deep learning algorithm such as deep residual network ResNet.

[0036] (4) Output recognition results The classification results of the output model are used to complete the entire process of individual radar radiation source identification.

[0037] It can be seen that current radar source identification methods are based on a single viewpoint and single period. However, in reality, radar signals are often the superposition of multiple processes, possessing inherent multi-period properties. This makes radar signal variations extremely complex, making it difficult to uncover reliable temporal relationships and subtle features through single-period modeling and identification. Therefore, radar source identification schemes based on single period and single viewpoint suffer from severe feature loss and poor model robustness, resulting in low accuracy in radar source identification.

[0038] In view of the above-mentioned technical status, this application proposes a new method for identifying individual radar radiation sources. By using this method to identify individual radar radiation sources, the accuracy of individual radar radiation source identification can be improved.

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] This application first proposes a method for identifying individual radar radiation sources, see [link to relevant documentation]. Figure 2 As shown, the method includes steps S201 to S203: S201. Obtain the raw signal of the radar radiation source individual to be identified.

[0041] Step S201 specifically includes: acquiring the high-frequency signal of the individual radar radiation source intercepted by the radar reconnaissance equipment, converting the high-frequency signal to an intermediate-frequency signal, and obtaining the original signal of the individual radar radiation source to be identified. The intermediate-frequency signal refers to a signal with a frequency range between 300 kHz and 30 MHz.

[0042] S202. Extract K signal sets from the original signal, and obtain the time-series variation characteristics of each signal set. Each signal set includes multiple sub-signals. The time-series variation characteristics include the time-series variation characteristics within the period of each sub-signal and the time-series variation characteristics between the multiple sub-signals during the period. K is a positive integer. For example, the value of K is an integer between 3 and 5.

[0043] In this embodiment, the original signal is a mixed waveform signal composed of electromagnetic waves of various different frequencies, or a mixed signal with multiple superimposed cycles. By extracting signals from the original signal from K different frequency perspectives, the original signal can be decomposed, thereby obtaining the signal sets corresponding to each of the K frequency perspectives.

[0044] The original signal of an individual radar radiation source is usually a time-domain signal. The specific implementation of extracting a set of K signals from the original signal can include the following steps a1) and a2): a1) Convert the original signal to the frequency domain to obtain the frequency domain signal corresponding to the original signal.

[0045] For example, suppose the original signal is a time-domain signal x(n) containing N sampling points. By performing a Fast Fourier Transform on x(n) using the following formula (1), the original signal can be transformed from the time domain to the frequency domain, resulting in the frequency domain signal x(k). Formula (1) is as follows: (1) In equation (1), k = 0, 1, ..., N-1.

[0046] a2) Based on the K frequency components with the largest frequency domain intensity in the frequency domain signal, extract K signal sets from the original signal.

[0047] After converting the original signal into a frequency domain signal x(k), the K signal periods with the most significant periodicity in the original signal can be determined based on the frequency domain signal. Then, a set of K signals can be extracted from the original signal based on these K signal periods. Specifically, the K frequency components f1, f2, ..., f1 with the highest frequency intensity can be selected from x(k) based on the following formula (2). k Formula (2) is as follows: (2) These K frequency components correspond to the K most significant signal periods, denoted as p1, p2, ..., pk. Based on these K signal periods, the signal sets corresponding to each of the K signal periods can be extracted from the original signal x(n).

[0048] In this embodiment, the K frequency components extracted have the highest frequency intensity, which means that the period of the sub-signal corresponding to the K frequency components is also more significant. The features within and during the period are more significant and prominent. By performing feature analysis on the signal set corresponding to the K frequency components, more subtle temporal change features in the original signal can be extracted, thereby improving the accuracy of radar radiation source identification.

[0049] After obtaining K signal sets, this embodiment can also use a two-dimensional neural network to extract features from each of the K signal sets, thereby obtaining the temporal variation features from K frequency perspectives. The two-dimensional neural network can be a basic residual unit of ResNet, or it can be a network such as Inception or ConvNeXt.

[0050] The temporal variation characteristics corresponding to each signal set include the intra-period temporal variation characteristics of each of the multiple sub-signals in the signal set and the inter-period temporal variation characteristics between the multiple sub-signals. The intra-period temporal variation characteristics correspond to the short-term process of a sub-signal within each period, while the inter-period variation reflects the long-term trend of the changes of multiple sub-signals over consecutive periods. Assuming a sub-signal duration of 5 seconds, the temporal variation of a certain signal feature from 0 seconds to 5 seconds is the intra-period temporal variation characteristic. Analyzing this feature using the rising edge of a pulse means analyzing the temporal variation of the rising edge of a sub-signal within each period from 0 seconds to 5 seconds. It should be understood that the analyzable features here are not limited to the rising edge of a pulse; they can also include higher-order cumulant features, etc.

[0051] Assuming two sub-signals A and B are each 5 seconds, the change of a certain characteristic between these two sub-signals from 0 seconds to 5 seconds represents the timing variation characteristic during the cycle. Continuing with the example of the rising edge of a pulse, the analysis focuses on the timing variation of the rising edge of the pulse between the two sub-signals.

[0052] Figure 3 This is a schematic diagram illustrating the time-series variation characteristics within a period and during a week, as provided in the embodiments of this application. For example... Figure 3 As shown, the intra-period temporal variation characteristics refer to the temporal variation characteristics between period 1 and period 2 of sub-signal 1. The inter-period temporal variation characteristics refer to the temporal variation characteristics of a certain feature between period 1 of sub-signal 1 and period 2 of sub-signal 2, and the temporal variation characteristics between period 2 of sub-signal 1 and sub-signal 2. S203. Based on K temporal variation characteristics, determine the classification results of individual radar radiation sources.

[0053] After obtaining the K time-series variation features corresponding to the K signal sets, individual radar radiation sources can be identified based on these K time-series variation features. Specifically, step S203 may include steps b1) and b2): b1) Fuse the K time-series variation features to obtain the fused signal features corresponding to the original signal.

[0054] When identifying individual radar radiation sources based on K temporal variation features, it is necessary to fuse the K temporal variation features to identify individual radar radiation sources based on the fused signal features. There are several ways to implement step b1), which can be as follows: In some implementations, step b1) includes: fusing K time-series variation features to obtain initial fused signal features corresponding to the original signal, and determining the initial fused signal features as the fused signal features corresponding to the original signal.

[0055] The fusion of K time-series variation features can be achieved using a feature weighting method. Specifically, the initial fused signal features are obtained by fusing the K time-series variation features, including: determining the weights of the K time-series variation features based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain signal corresponding to the original signal, wherein the magnitude of each frequency component is positively correlated with the weight of the time-series variation feature corresponding to each frequency component; and then performing a weighted summation based on the K time-series variation features and their respective weights to obtain the initial fused signal features.

[0056] In this embodiment, the magnitude of each frequency component is positively correlated with the weight of the time-series variation feature corresponding to each frequency component. This can be understood as follows: the larger the frequency of a frequency component, the greater the weight of the time-series variation feature corresponding to that frequency component; conversely, the smaller the frequency of a frequency component, the smaller the weight of the time-series variation feature corresponding to that frequency component.

[0057] The sum of the weights corresponding to the K time-series change features is 1.

[0058] In some implementations, the weights of the K time-series variation features are determined based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain signal corresponding to the original signal. This includes determining the weights of the time-series variation features corresponding to the K frequency components based on the proportion of the frequency intensity of each of the K frequency components in the sum of the frequency intensities of the K frequency components.

[0059] For example, if the K frequency components are denoted as frequency component 1, frequency component 2, and frequency component 3, and their respective frequency intensities are 200Hz, 500Hz, and 300Hz, then the weight of the time-series variation feature corresponding to frequency component 1 is 200Hz / (200Hz+500Hz+300Hz)=0.2, the weight of the time-series variation feature corresponding to frequency component 2 is 500Hz / (200Hz+500Hz+300Hz)=0.5, and the weight of the time-series variation feature corresponding to frequency component 3 is 300Hz / (200Hz+500Hz+300Hz)=0.3.

[0060] The greater the frequency intensity, the more obvious the temporal variation characteristics of the signal set under that frequency component. Therefore, setting a greater weight for the temporal variation characteristics under that frequency component can ensure that the temporal variation characteristics of the weighted initial fused signal are more representative.

[0061] In some implementations, step b1) includes: fusing the temporal variation features of K signal sets to obtain initial fused signal features, determining the residual result between the initial fused signal features and the original signal as the intermediate fused signal features corresponding to the original signal, and if the number of iterations has not reached a preset number, using the intermediate signal features as the new original signal for the next iteration, and repeatedly extracting K signal sets from the original signal, obtaining the temporal variation features of each signal set, fusing the temporal variation features of each signal set to obtain initial fused signal features, and determining the residual result between the initial fused signal features and the original signal as the intermediate fused signal features corresponding to the original signal, until the number of iterations reaches a preset number, and determining the intermediate fused signal features obtained in the last iteration as the fused signal features corresponding to the original signal.

[0062] Here, the preset number of times is denoted as m, and m can be a positive integer. That is, m can be 1 or a positive integer greater than 1. When m is 1, step b1) includes: fusing the time-series variation features of the K signal sets to obtain initial fused signal features, determining the residual result between the initial fused signal features and the original signals as the intermediate fused signal features corresponding to the original signals, and determining the intermediate fused signal features as the fused signal features corresponding to the original signals.

[0063] When m is a positive integer greater than 1, let t be the current iteration number, where t is a positive integer between 1 and m. Then step b1 can include steps b11) and b12). b11) For the t-th iteration, if t is not equal to m, the intermediate fused signal features obtained in the (t-1)-th iteration are used as the original signal of the t-th iteration, and the following steps are repeated for the original signal of the t-th iteration: extracting K signal sets from the original signal of the t-th iteration, obtaining the temporal change features of each signal set, fusing the temporal change features of each signal set to obtain the initial fused signal features, and determining the residual result of the initial fused signal features and the original signal of the t-th iteration as the intermediate fused signal features corresponding to the original signal of the t-th iteration.

[0064] b12) Perform the t+1 operation on t and take t+1 as the new t, and determine whether the new t is equal to m. If the new t is not equal to m, continue to execute step b11. If the new t is equal to m, determine that the number of iterations has reached the preset number, and determine the intermediate fusion signal feature obtained in the last iteration, i.e. the m-th iteration, as the fusion signal feature corresponding to the original signal.

[0065] While residual learning can increase network model depth, it can also lead to network degradation. That is, as network depth increases, although deeper features can be extracted, feature loss becomes more severe. This embodiment addresses this by performing a residual concatenation between the initial fused signal features and the original signal, and then using the residual result as the intermediate fused signal feature corresponding to the original signal. This approach extracts deeper features while reducing feature loss, thereby improving the accuracy of subsequent radar source identification.

[0066] b2) Based on the fused signal characteristics, determine the classification results of individual radar radiation sources.

[0067] After obtaining the fused signal characteristics, individual radar radiation sources can be identified based on these characteristics, thus obtaining the classification results of individual radar radiation sources.

[0068] In some embodiments, steps S201 to S203 described above can also be implemented using a radar source individual classification model. Specifically, the original signal can be input into the radar source individual classification model, so that the model extracts K signal sets from the original signal, obtains the temporal variation characteristics of each signal set, and determines the classification result of the individual radar source based on the K temporal variation characteristics. In other words, by inputting the original signal into the radar source individual classification model, the classification result of the individual radar source can be obtained, thereby achieving end-to-end radar source individual identification.

[0069] Before applying the radar radiation source individual classification model for individual radar radiation source identification, it is necessary to train the radar radiation source individual classification model. The training process of the radar radiation source individual classification model will be described in detail below with reference to specific embodiments: Figure 4 This is a flowchart illustrating the training process of the radar radiation source individual classification model provided in this application embodiment. Figure 4 As shown, the training process of the radar radiation source individual classification model includes the following steps S401 to S402: S401. Extract K sample signal sets from the acquired original sample signals, and obtain the time-series variation sample features of each sample signal set. Each sample signal set includes multiple sub-sample signals. The time-series variation sample features include the time-series variation features within the period of each of the multiple sub-sample signals and the time-series variation features between the multiple sub-sample signals during the period.

[0070] The original sample signal is a time-domain sample signal.

[0071] Step S401 involves extracting a set of K sample signals from the acquired original sample signal, including: converting the original sample signal to the frequency domain to obtain the frequency domain sample signal corresponding to the original sample signal; and extracting a set of K sample signals from the original sample signal based on the K frequency components with the largest frequency intensity in the frequency domain sample signal.

[0072] The process of converting the original sample signal to the frequency domain to obtain the frequency domain sample signal corresponding to the original sample signal may include: performing a fast Fourier transform on the original sample signal to convert the original sample signal to the frequency domain to obtain the frequency domain sample signal corresponding to the original sample signal.

[0073] For example, assuming the original sample signal is a time-domain signal x'(n) containing N sampling points, the original sample signal can be transformed from the time domain to the frequency domain by performing a Fast Fourier Transform on x'(n) using the following formula (3), resulting in the frequency-domain sample signal x'(k). The specific formula (3) is as follows: (3) In equation (3), k = 0, 1, ..., N-1.

[0074] After converting the original sample signal into a frequency domain sample signal x'(k), the K signal periods with the most significant signal period in the original sample signal can be determined based on the frequency domain sample signal, and then a set of K sample signals can be extracted from the original sample signal based on these K signal periods.

[0075] Specifically, based on the following formula (4), K frequency components f'1, f'2, ..., f' with the largest frequency intensity can be selected from x'(k).k Formula (4) is as follows: (4) These K frequency components correspond to the K most significant signal periods, denoted as p'1, ​​p'2, ..., p'k. Based on these K signal periods, the sample signal sets corresponding to each of the K signal periods can be extracted from the original sample signal x'(n).

[0076] After obtaining the set of K sample signals, this embodiment can also use a two-dimensional neural network to extract features from each of the K sample signal sets, thereby obtaining the time-series variation sample features from the perspective of K frequencies. The two-dimensional neural network can be a basic residual unit of ResNet, or it can be a network such as Inception or ConvNeXt.

[0077] The temporal variation sample features corresponding to each sample signal set include the temporal variation sample features within the period of each of the multiple sub-sample signals in the sample signal set, and the periodic temporal variation sample features between the multiple sub-sample signals.

[0078] S402. Based on the features of K time-series changing samples, train the preset radar radiation source individual classification network to obtain the radar radiation source individual classification model.

[0079] After obtaining the K time-series variation sample features corresponding to the K sample signal sets, individual radar radiation sources can be identified based on the K time-series variation sample features. Specifically, step S402 includes steps c1) and c2). c1) The features of K time-varying samples are fused to obtain the fused sample signal features corresponding to the original sample signal.

[0080] When identifying individual radar radiation sources based on K time-series varied sample features, it is necessary to fuse the K time-series varied sample features to identify individual radar radiation sources based on the fused sample signal features. There are several ways to implement step c1), which can be as follows: In some implementations, step c1) includes: fusing the features of K time-varying samples to obtain initial sample signal fusion features, and determining the initial sample signal fusion features as the fusion sample signal features corresponding to the original sample signals.

[0081] The fusion of K time-varying sample features can be achieved using a feature weighting method. Specifically, the initial sample signal fusion features are obtained by fusing the K time-varying sample features, including: determining the weights of the K time-varying sample features based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain sample signal corresponding to the original sample signal, wherein the magnitude of each frequency component is positively correlated with the weight of the time-varying sample feature corresponding to each frequency component; and then performing a weighted summation based on the K time-varying sample features and their respective weights to obtain the initial sample signal fusion features.

[0082] In this embodiment, the magnitude of each frequency component is positively correlated with the weight of the time-series variation sample feature corresponding to each frequency component. This can be understood as follows: the larger the frequency corresponding to a frequency component, the larger the weight of the time-series variation sample feature corresponding to that frequency component; conversely, the smaller the frequency corresponding to a frequency component, the smaller the weight of the time-series variation sample feature corresponding to that frequency component.

[0083] In this case, the sum of the weights corresponding to the K time-series variation sample features is 1.

[0084] In some implementations, the weights of the K time-varying sample features are determined based on the magnitudes of the K frequency components with the largest frequency intensities in the frequency domain sample signal corresponding to the original sample signal. This includes determining the weights of the time-varying sample features corresponding to the K frequency components based on the proportion of the frequency intensity of each of the K frequency components in the sum of the frequency intensities of the K frequency components.

[0085] For example, if the K frequency components are denoted as frequency component 1, frequency component 2, and frequency component 3, and their respective frequency intensities are 400Hz, 600Hz, and 1000Hz, then the weight of the time-series variation feature corresponding to frequency component 1 is 400Hz / (400Hz+600Hz+1000Hz) = 0.2, the weight of the time-series variation feature corresponding to frequency component 2 is 600Hz / (400Hz+600Hz+1000Hz) = 0.3, and the weight of the time-series variation feature corresponding to frequency component 3 is 1000Hz / (400Hz+600Hz+1000Hz) = 0.5.

[0086] In some implementations, K time-varying sample features are fused to obtain initial sample signal fusion features. The residual result between the initial sample signal fusion features and the original sample signals is determined as the intermediate sample signal fusion features corresponding to the original sample signals. If the number of iterations has not reached a preset number, the intermediate sample signal fusion features are used as the new original sample signals for the next iteration. The process of extracting K sample signal sets from the original sample signals, obtaining the time-varying sample features of each sample signal set, fusing the time-varying sample features of each sample signal set to obtain the initial sample signal fusion features, and determining the residual result between the initial sample signal fusion features and the original sample signals as the intermediate sample signal fusion features corresponding to the original sample signals is repeated until the number of iterations reaches a preset number. Finally, the intermediate sample signal fusion features obtained in the last iteration are determined as the fusion sample signal features corresponding to the original sample signals.

[0087] Here, the preset number of times is denoted as m, and m can be a positive integer. That is, m can be 1 or a positive integer greater than 1. When m is 1, step c1) includes: fusing the time-series variation sample features of the K sample signal sets to obtain the initial sample signal fusion feature, determining the residual result between the initial sample signal fusion feature and the original sample signal as the intermediate sample signal fusion feature corresponding to the original sample signal, and determining the intermediate sample signal fusion feature as the fused sample signal feature corresponding to the original sample signal.

[0088] When m is a positive integer greater than 1, let t be the current iteration number. Then step c1) can include steps c11) and c12). c11) For the t-th iteration, if t is not equal to m, the intermediate sample signal fusion feature obtained in the (t-1)-th iteration is used as the original sample signal of the t-th iteration, and the following steps are repeated for the original sample signal of the t-th iteration: extracting K sample signal sets from the original sample signal of the t-th iteration, obtaining the time-series change sample features of each sample signal set, fusing the time-series change sample features of each sample signal set to obtain the initial sample signal fusion feature, and determining the residual result of the initial sample signal fusion feature and the original sample signal of the t-th iteration as the intermediate sample signal fusion feature corresponding to the original sample signal of the t-th iteration.

[0089] c12) Perform the t+1 operation on t and take t+1 as the new t, and determine whether the new t is equal to m. If the new t is not equal to m, continue to execute step b11. If the new t is equal to m, determine that the number of iterations has reached the preset number, and determine the fusion feature of the intermediate sample signal obtained in the last iteration, i.e. the m-th iteration, as the fusion sample signal feature corresponding to the original sample signal.

[0090] While residual learning can increase network model depth, it can also lead to network degradation. That is, as network depth increases, although deeper features can be extracted, feature loss becomes more severe. This embodiment addresses this by performing a residual connection between the initial sample signal fusion feature and the original sample signal, and then using the residual result as the intermediate sample signal fusion feature corresponding to the original sample signal. Based on this intermediate sample signal fusion feature, the fused sample feature is determined. This approach can extract deeper features while reducing feature loss, thereby training a radar radiation source individual classification model with higher identification accuracy, ultimately improving the accuracy of radar radiation source individual identification.

[0091] c2) The radar radiation source individual classification network is trained based on the fused sample signal features to obtain the radar radiation source individual classification model.

[0092] The preset radar radiation source individual classification network can use a linear classifier.

[0093] Step c2) specifically includes: classifying the fused sample signal features using a pre-defined radar source individual classification network to obtain the predicted classification result of the radar source individual; determining the classification loss based on the difference between the predicted classification result and its corresponding true classification result; and converging the pre-defined radar source individual classification network based on the classification loss to obtain the radar source individual classification model.

[0094] The classification loss can be determined using the cross-entropy loss function. When the pre-defined radar source individual classification network converges based on the classification loss, backpropagation can be performed to update the parameters of the network. Training continues based on the updated network parameters. Training continues as long as the classification loss is greater than the pre-defined loss. When the classification loss is less than or equal to the pre-defined loss, the convergence condition is met, and training ends. The converged radar source individual classification network is then used as the radar source individual classification model.

[0095] It should be understood that convergence conditions may also include reaching a preset number of training iterations or a preset training accuracy, etc., but this embodiment does not limit this.

[0096] The overall implementation process of the radar radiation source identification method is described below with reference to the accompanying drawings.

[0097] Figure 5 This is a schematic diagram illustrating the overall implementation process of the radar radiation source identification method provided in the embodiments of this application. (See attached diagram.) Figure 5 As shown, the overall implementation process of the radar radiation source identification method includes the following steps: 1) Time-domain to frequency-domain conversion Please see Figure 5 The waveform shown in the electromagnetic signal waveform diagram is the time domain signal. By performing frequency domain conversion on it, we can obtain the spectrum diagram, and the waveform shown in the spectrum diagram is the frequency domain signal.

[0098] 2) Extract K signal sets As can be seen from the spectrum, there are three frequency components with the most significant and prominent frequency intensity. Therefore, based on these three frequency components, three signal sets can be extracted from the original signal, namely signal set 1, signal set 2, and signal set 3.

[0099] Figure 5 The waveforms shown within each black rectangle represent a signal set. Each waveform within a black rectangle represents a sub-signal. As can be seen, each signal set contains multiple sub-signals. Signal set 1 contains two waveforms from top to bottom; these two waveforms are the sub-signals of signal set 1. Similarly, signal set 2 contains three waveforms, which are the sub-signals of signal set 2. Signal set 3 contains six waveforms, which are the sub-signals of signal set 3.

[0100] 3) Feature extraction By using a two-dimensional neural network module to extract features from the three signal sets, we can obtain the time-series variation features from three frequency perspectives, namely time-series variation feature 1, time-series variation feature 2 and time-series variation feature 3 in the figure.

[0101] 4) Feature fusion Feature fusion refers to the fusion of time-series variation features from three frequency perspectives into an initial fused signal feature.

[0102] 5) Fusion features and original signal residual connection To reduce feature loss, this embodiment can also perform residual connection on the initial fused features and the original signal after fusion to reduce this phenomenon.

[0103] 6) Multiple iterations After obtaining the intermediate fused signal features after residual connection, classification can be performed directly based on these intermediate fused signal features, or feature extraction and fusion steps can be performed on these intermediate fused signal features for multiple iterations until the preset number of iterations is reached. Then, the intermediate fused signal features from the last iteration are sent to the backend classifier for classification.

[0104] The radar radiation source identification method in this embodiment analyzes the temporal variation characteristics of the intercepted raw signal from within and during the period, thereby enabling highly robust and accurate identification under complex electromagnetic environment conditions such as low signal-to-noise ratio and incomplete signal acquisition.

[0105] The beneficial effects of this application are explained below with reference to experimental data: The CPU used in the experiment was an Intel Xeon series CPU, and the GPU was a Tesla T4. The detailed configuration is shown in Table 1.

[0106] Table 1. Configuration of the experimental platform The experimental data consisted of data from 20 radar emitters, comprising 10 individuals from each of two different models. Each individual emitted 1000 intermediate frequency pulses, with a pulse wave (PW) ranging from a few microseconds to several hundred microseconds and an RF range from 1 GHz to 10 GHz. The sampling rate was 500 MHz, and the data was stored in int16 format. The ratio of training data to validation data to test data was 7:2:1.

[0107] To verify the beneficial effects of the embodiments of the radar radiation source individual identification method of this application, this application conducted experimental verification on data from two different radar models, and built a strong baseline in the existing scheme (based on CPC unsupervised representation learning to extract radar signal pulses, and then using deep residual network ResNet as the back-end classifier) ​​as a baseline for performance comparison. The experimental results are shown in Table 2.

[0108] Table 2 Experimental Results Record As can be seen from the experimental results in Table 2, the radar individual end-to-end identification scheme based on multi-period view adaptive fusion modeling proposed in this application significantly improves the individual identification accuracy on radar models 1 and 2 compared with the baseline based on single-period view, and this application performs well on data from both radar models.

[0109] Corresponding to the model training method described above, this application also proposes a device for identifying individual radar radiation sources, such as... Figure 6As shown, the device includes: an acquisition unit 601, used to acquire the original signal of the radar radiation source individual to be identified; a feature extraction unit 602, used to extract K signal sets from the original signal and acquire the temporal variation features of each signal set, each signal set including multiple sub-signals, the temporal variation features including the intra-period temporal variation features of each of the multiple sub-signals and the inter-period temporal variation features between the multiple sub-signals, where K is a positive integer; and a classification unit 603, used to determine the classification result of the radar radiation source individual based on the K temporal variation features.

[0110] In some embodiments, the feature extraction unit 602 extracts K signal sets from the original signal and obtains the temporal variation features of each signal set. By fusing the temporal variation features of each signal set, a fused signal feature corresponding to the original signal is obtained. Based on the fused signal feature, the classification result of the individual radar radiation source is determined, including: inputting the original signal into a radar radiation source individual classification model so that the radar radiation source individual classification model extracts K signal sets from the original signal and obtains the temporal variation features of each signal set. Based on the K temporal variation features, the classification result of the individual radar radiation source is determined.

[0111] In some embodiments, the original signal is a time-domain signal, and the feature extraction unit 602 extracts a set of K signals from the original signal, including: converting the original signal to the frequency domain to obtain a frequency domain signal corresponding to the original signal; and extracting the set of K signals from the original signal based on the K frequency components with the largest frequency intensity in the frequency domain signal.

[0112] In some embodiments, the classification unit 603 determines the classification result of the individual radar radiation source based on K temporal variation features, including: fusing the K temporal variation features to obtain the fused signal features corresponding to the original signal; and determining the classification result of the individual radar radiation source based on the fused signal features.

[0113] In some embodiments, the feature extraction unit 602 fuses the K time-series variation features to obtain fused signal features corresponding to the original signal, including: fusing the K time-series variation features to obtain initial fused signal features, and determining the initial fused signal features as the fused signal features corresponding to the original signal; or, fusing the K time-series variation features to obtain initial fused signal features, and determining the residual result of the initial fused signal features and the original signal as intermediate fused signal features corresponding to the original signal, and if the number of iterations has not reached a preset number, using the intermediate fused signal features as the new original signal in the next iteration, and repeatedly extracting K signal sets from the original signal, obtaining the time-series variation features of each signal set, fusing the time-series variation features of each signal set to obtain initial fused signal features, and determining the residual result of the initial fused signal features and the original signal as intermediate fused signal features corresponding to the original signal, until the number of iterations reaches a preset number, and determining the intermediate fused signal features obtained in the last iteration as the fused signal features corresponding to the original signal.

[0114] In some embodiments, the feature extraction unit 602 fuses the K time-series variation features to obtain initial fused signal features, including: determining the weights of the K time-series variation features based on the magnitudes of the K frequency components with the largest frequency intensity in the frequency domain signal corresponding to the original signal, wherein the magnitude of each frequency component is positively correlated with the weight of the time-series variation feature corresponding to each frequency component; and performing a weighted summation based on the K time-series variation features and their respective weights to obtain the initial fused signal features.

[0115] In some embodiments, the device further includes a training unit 604, configured to perform the following steps: extracting K sets of sample signals from the acquired original sample signals; acquiring time-series variation sample features of each set of sample signals, wherein each set of sample signals includes multiple sub-sample signals, and the time-series variation sample features include the time-series variation features within the period of each of the multiple sub-sample signals and the time-series variation features between the multiple sub-sample signals during the period; and training a preset radar radiation source individual classification network based on the K time-series variation sample features to obtain a radar radiation source individual classification model.

[0116] In some embodiments, the original sample signal is a time-domain sample signal, and the training unit 604 extracts a set of K sample signals from the acquired original sample signal, including: converting the original sample signal to the frequency domain to obtain a frequency-domain sample signal corresponding to the original sample signal; and extracting the set of K sample signals from the original sample signal based on the K frequency components with the largest frequency intensity in the frequency-domain sample signal.

[0117] In some embodiments, the training unit 604 trains a preset radar radiation source individual classification network based on K time-series variation features to obtain a radar radiation source individual classification model, including: fusing the K time-series variation sample features to obtain fused sample signal features corresponding to the original sample signal; and training the preset radar radiation source individual classification network based on the fused sample signal features to obtain a radar radiation source individual classification model.

[0118] In some embodiments, the training unit 604 fuses K time-varying sample features to obtain fused sample signal features corresponding to the original signal, including: fusing the K time-varying sample features to obtain initial sample signal fusion features, and determining the initial sample signal fusion features as the fused sample signal features corresponding to the original sample signal; or, fusing the K time-varying sample features to obtain initial sample signal fusion features, and determining the residual result of the initial sample signal fusion features and the original sample signal as the intermediate sample signal fusion features corresponding to the original sample signal, and when the number of iterations has not reached a preset number. The intermediate sample signal fusion feature is used as the new original sample signal in the next iteration. K sample signal sets are extracted from the original sample signal, and the time-series variation sample features of each sample signal set are obtained. The initial sample signal fusion feature is obtained by fusing the time-series variation sample features of each sample signal set. The residual result of the initial sample signal fusion feature and the original sample signal is determined as the intermediate sample signal fusion feature corresponding to the original sample signal. The iteration count is repeated until the preset number of iterations is reached. The intermediate sample signal fusion feature obtained in the last iteration is determined as the fusion sample signal feature corresponding to the original sample signal.

[0119] In some embodiments, the training unit 604 fuses the time-varying sample features of the K sample signal sets to obtain initial sample signal fusion features, including: determining the weights corresponding to the K time-varying sample features based on the magnitudes of the K frequency components with the largest frequency intensity in the frequency domain sample signal corresponding to the original sample signal, wherein the magnitude of each frequency component is positively correlated with the weight of the time-varying sample feature corresponding to each frequency component; and performing a weighted summation based on the K time-varying sample features and their respective weights to obtain the initial sample signal fusion features.

[0120] In some embodiments, the training unit 604 trains a preset radar source individual classification network based on the fused sample signal features to obtain a radar source individual classification model, including: classifying the fused sample signal features using the preset radar source individual classification network to obtain a predicted classification result for radar source individuals; determining a classification loss based on the difference between the predicted classification result and its corresponding true classification result; and converging the preset radar source individual classification network based on the classification loss to obtain the radar source individual classification model.

[0121] The radar source identification device provided in this embodiment belongs to the same concept as the radar source identification method provided in the above embodiments of this application. It can execute the radar source identification method provided in any of the above embodiments of this application and has the corresponding functional modules and beneficial effects for executing the radar source identification method. Technical details not described in detail in this embodiment can be found in the specific processing content of the radar source identification method provided in the above embodiments of this application, and will not be repeated here.

[0122] The functions of each unit in the radar radiation source identification device described above can be implemented by the same or different processors, and this application embodiment does not limit this.

[0123] It should be understood that the units in the radar source identification device described above can be implemented by a processor calling software. For example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to realize the functions of each unit in the device. The processor can be a general-purpose processor, such as a CPU or microprocessor, and the memory can be internal or external to the device. Alternatively, the units in the device can be implemented as hardware circuits. By designing the hardware circuits, some or all of the unit functions can be realized. The hardware circuit can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above units are realized by designing the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a PLD, such as an FPGA, which can include a large number of logic gates. The connection relationships between the logic gates are configured through a configuration file to realize the functions of some or all of the above units. All units of the above device can be implemented entirely by a processor calling software, entirely by hardware circuits, or partially by a processor calling software with the remaining parts implemented by hardware circuits.

[0124] In this application embodiment, a processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction reading and execution capabilities, such as a CPU, microprocessor, GPU, or DSP. In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. These logical relationships are fixed or reconfigurable. For example, the processor may be a hardware circuit implemented as an ASIC or PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the processor loading instructions to implement the functions of some or all of the above units. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as an NPU, TPU, or DPU.

[0125] As can be seen, each unit in the above device can be one or more processors (or processing circuits) configured to implement the above methods, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0126] Furthermore, the units in the above devices can be integrated in whole or in part, or they can be implemented independently. In one implementation, these units are integrated together and implemented in the form of a System-on-Chip (SoC). The SoC may include at least one processor for implementing any of the above methods or implementing the functions of the units in the device. The at least one processor may be of different types, such as CPU and FPGA, CPU and artificial intelligence processor, CPU and GPU, etc.

[0127] This application also proposes a control device, which includes a processor and an interface circuit. The processor in the control device is connected to a data input component through the interface circuit of the control device.

[0128] The data input component specifically refers to a functional component that can input or collect voice data, such as a microphone, etc.

[0129] The aforementioned interface circuit can be any interface circuit capable of implementing data communication functions, such as a USB interface circuit, a Type-C interface circuit, a serial port circuit, a PCIe circuit, etc.

[0130] The processor in this control device is also a circuit with signal processing capabilities, which executes the radar radiation source identification method described in the above embodiments. For specific implementation details of the processor, please refer to the processor implementation methods described above; this application does not impose strict limitations on these implementations.

[0131] Another embodiment of this application also provides an electronic device, see [link to relevant documentation] Figure 6 As shown, the device includes: Memory 200 and processor 210; The memory 200 is connected to the processor 210 and is used to store programs; The processor 210 is used to implement the radar radiation source identification method disclosed in any of the above embodiments by running the program stored in the memory 200.

[0132] Specifically, the aforementioned electronic device may also include: a bus, a communication interface 220, an input device 230, and an output device 240.

[0133] The processor 210, memory 200, communication interface 220, input device 230, and output device 240 are interconnected via a bus. Among them: A bus can include a pathway for transmitting information between various components of a computer system.

[0134] Processor 210 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits used to control the execution of the program of the present invention. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0135] Processor 210 may include a main processor, as well as a baseband chip, modem, etc.

[0136] The memory 200 stores a program that executes the technical solution of this invention, and may also store an operating system and other key business functions. Specifically, the program may include program code, which includes computer operation instructions. More specifically, the memory 200 may include read-only memory (ROM), other types of static storage devices capable of storing static information and instructions, random access memory (RAM), other types of dynamic storage devices capable of storing information and instructions, disk storage, flash memory, etc.

[0137] Input device 230 may include a device for receiving user input data and information, such as a keyboard, mouse, camera, scanner, light pen, voice input device, touch screen, pedometer, or gravity sensor.

[0138] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0139] The communication interface 220 may include a device that uses any transceiver to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0140] The processor 210 executes the program stored in the memory 200 and calls other devices, which can be used to implement each step of any of the radar radiation source identification methods provided in the above embodiments of this application.

[0141] This application also proposes a chip, which includes a processor and a data interface. The processor reads and runs a program stored in a memory through the data interface to execute the radar radiation source identification method described in any of the above embodiments. For the specific processing procedure and its beneficial effects, please refer to the above embodiments of the radar radiation source identification method.

[0142] In addition to the methods and devices described above, embodiments of this application may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the method for identifying individual radar radiation sources described in any of the embodiments described above.

[0143] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0144] Furthermore, embodiments of this application may also be storage media storing a computer program, which is executed by a processor to perform the steps in the radar radiation source identification method described in any of the above embodiments of this specification.

[0145] For the foregoing method embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0146] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are basically similar to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0147] The steps in the methods of the various embodiments of this application can be adjusted, merged, or deleted in order according to actual needs, and the technical features described in each embodiment can be replaced or combined.

[0148] The modules and sub-modules in the various embodiments of the present application's devices and terminals can be merged, divided, and deleted according to actual needs.

[0149] It should be understood that the disclosed terminals, devices, and methods can be implemented in other ways, given the several embodiments provided in this application. For example, the terminal embodiments described above are merely illustrative. For instance, the division of modules or sub-modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple sub-modules or modules may be combined or integrated into another module, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0150] The modules or submodules described as separate components may or may not be physically separate. The components that constitute a module or submodule may or may not be physical modules or submodules; that is, they may be located in one place or distributed across multiple network modules or submodules. Some or all of the modules or submodules can be selected to achieve the purpose of this embodiment's solution, depending on actual needs.

[0151] Furthermore, the functional modules or sub-modules in the various embodiments of this application can be integrated into one processing module, or each module or sub-module can exist physically separately, or two or more modules or sub-modules can be integrated into one module. The integrated modules or sub-modules described above can be implemented in hardware or in the form of software functional modules or sub-modules.

[0152] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0153] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software unit executed by a processor, or a combination of both. The software unit can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0154] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0155] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying individual radar radiation sources, characterized in that, include: Acquire the raw signal of the individual radar radiation source to be identified; K signal sets are extracted from the original signal, and the temporal variation characteristics of each signal set are obtained. Each signal set includes multiple sub-signals. The temporal variation characteristics include the temporal variation characteristics within the period of each of the multiple sub-signals and the temporal variation characteristics between the multiple sub-signals during the period. K is a positive integer. Sub-signals within the same signal set have the same frequency. Sub-signals within different signal sets have different frequencies. Based on K temporal variation features, the classification results for the individual radar radiation sources are determined.

2. The method according to claim 1, characterized in that, K signal sets are extracted from the original signal, and the temporal variation characteristics of each signal set are obtained. The temporal variation characteristics of each signal set are fused to obtain the fused signal characteristics corresponding to the original signal. Based on the fused signal characteristics, the classification result of the individual radar radiation source is determined, including: The original signal is input into the radar radiation source individual classification model, so that the radar radiation source individual classification model extracts K signal sets from the original signal, obtains the temporal variation characteristics of each signal set, and determines the classification result of the radar radiation source individual based on the K temporal variation characteristics.

3. The method according to claim 1 or 2, characterized in that, The original signal is a time-domain signal, and the extraction of K signal sets from the original signal includes: The original signal is converted to the frequency domain to obtain the frequency domain signal corresponding to the original signal; Based on the K frequency components with the largest frequency intensity in the frequency domain signal, the K signal sets are extracted from the original signal.

4. The method according to claim 1 or 2, characterized in that, The classification result for the individual radar radiation sources based on K temporal variation features includes: By fusing the K time-series variation features, the fused signal features corresponding to the original signal are obtained; Based on the fused signal characteristics, the classification result of the individual radar radiation sources is determined.

5. The method according to claim 4, characterized in that, By fusing the K time-series variation features, the fused signal features corresponding to the original signal are obtained, including: The K time-series variation features are fused to obtain initial fused signal features, and the initial fused signal features are determined as the fused signal features corresponding to the original signal. or, The K time-series variation features are fused to obtain an initial fused signal feature. The residual result between the initial fused signal feature and the original signal is determined as the intermediate fused signal feature corresponding to the original signal. If the number of iterations has not reached a preset number, the intermediate fused signal feature is used as the new original signal for the next iteration. The process of extracting K signal sets from the original signal, obtaining the time-series variation features of each signal set, fusing the time-series variation features of each signal set to obtain the initial fused signal feature, and determining the residual result between the initial fused signal feature and the original signal as the intermediate fused signal feature corresponding to the original signal is repeated until the number of iterations reaches a preset number. Finally, the intermediate fused signal feature obtained in the last iteration is determined as the fused signal feature corresponding to the original signal.

6. The method according to claim 5, characterized in that, The process of fusing the K time-series variation features to obtain initial fused signal features includes: Based on the magnitudes of the K frequency components with the largest frequency intensity in the frequency domain signal corresponding to the original signal, the weights of the K time-series variation features are determined, wherein the magnitude of each frequency component and the weight of the time-series variation feature corresponding to each frequency component are positively correlated. The initial fused signal features are obtained by weighted summation based on the K time-series variation features and their respective weights.

7. The method according to claim 2, characterized in that, The training process of the radar radiation source individual classification model includes: extracting a set of K sample signals from the acquired original sample signals; The temporal variation sample features of each sample signal set are obtained respectively. Each sample signal set includes multiple sub-sample signals. The temporal variation sample features include the intra-period temporal variation features of each of the multiple sub-sample signals and the inter-period temporal variation features between the multiple sub-sample signals. Based on the features of K time-varying samples, a pre-defined radar radiation source individual classification network is trained to obtain a radar radiation source individual classification model.

8. The method according to claim 7, characterized in that, The original sample signal is a time-domain sample signal, and the extraction of a set of K sample signals from the acquired original sample signal includes: The original sample signal is converted to the frequency domain to obtain the frequency domain sample signal corresponding to the original sample signal; Based on the K frequency components with the largest frequency intensity in the frequency domain sample signal, the set of K sample signals is extracted from the original sample signal.

9. The method according to claim 7, characterized in that, The process involves training a pre-defined radar source individual classification network based on K temporal variation features to obtain a radar source individual classification model, including: The features of K time-series varied samples are fused to obtain the fused sample signal features corresponding to the original sample signal; The radar radiation source individual classification network is trained based on the fused sample signal features to obtain the radar radiation source individual classification model.

10. The method according to claim 9, characterized in that, The process of fusing the features of K time-varying samples to obtain the fused sample signal features corresponding to the original signal includes: The features of the K time-series changing samples are fused to obtain the initial sample signal fusion features, and the initial sample signal fusion features are determined as the fusion sample signal features corresponding to the original sample signal. or, The K time-varying sample features are fused to obtain initial sample signal fusion features. The residual result between the initial sample signal fusion features and the original sample signal is determined as the intermediate sample signal fusion features corresponding to the original sample signal. If the number of iterations has not reached a preset number, the intermediate sample signal fusion features are used as the new original sample signal for the next iteration. The process of extracting K sample signal sets from the original sample signal, obtaining the time-varying sample features of each sample signal set, fusing the time-varying sample features of each sample signal set to obtain initial sample signal fusion features, and determining the residual result between the initial sample signal fusion features and the original sample signal as the intermediate sample signal fusion features corresponding to the original sample signal is repeated until the number of iterations reaches a preset number. Finally, the intermediate sample signal fusion features obtained in the last iteration are determined as the fused sample signal features corresponding to the original sample signal.

11. The method according to claim 10, characterized in that, The process of fusing the time-series variation sample features of the K sample signal sets to obtain initial sample signal fusion features includes: Based on the magnitudes of the K frequency components with the largest frequency intensity in the frequency domain sample signal corresponding to the original sample signal, the weights of the K time-varying sample features are determined, wherein the magnitude of each frequency component and the weight of the time-varying sample feature corresponding to each frequency component are positively correlated. The initial sample signal fusion features are obtained by weighted summation based on the features of the K time-varying samples and their respective weights.

12. The method according to claim 9, characterized in that, The step of training a preset radar source individual classification network based on the fused sample signal features to obtain a radar source individual classification model includes: The fused sample signal features are classified using the preset radar radiation source individual classification network to obtain the predicted classification results of radar radiation source individuals; The classification loss is determined based on the difference between the predicted classification result and its corresponding true classification result; The radar radiation source individual classification network is converged based on the classification loss to obtain the radar radiation source individual classification model.

13. A device for identifying an individual radar radiation source, characterized in that, include: The acquisition unit is used to acquire the raw signal of the radar radiation source individual to be identified; The feature extraction unit is used to extract K signal sets from the original signal and obtain the temporal variation features of each signal set. Each signal set includes multiple sub-signals. The temporal variation features include the temporal variation features within the period of each of the multiple sub-signals and the temporal variation features between the multiple sub-signals during the period. K is a positive integer. Sub-signals within the same signal set have the same frequency. Sub-signals within different signal sets have different frequencies. The classification unit is used to determine the classification result of the individual radar radiation source based on K temporal variation features.

14. An electronic device, characterized in that, Including memory and processor; The memory is connected to the processor and is used to store programs; The processor is configured to implement the method as described in any one of claims 1 to 12 by running a program in the memory.

15. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method as described in any one of claims 1 to 12.

16. A computer program product, characterized in that, It includes computer program instructions that, when executed by a processor, cause the processor to perform the method as described in any one of claims 1 to 12.

Images