Electromagnetic data recognition method and device, electronic equipment and storage medium

Abstract

The application provides an electromagnetic data recognition method and device, electronic equipment and storage medium, and belongs to the technical field of artificial intelligence, and comprises the following steps: inputting an electromagnetic signal to be recognized into an electromagnetic data analysis model; the electromagnetic data analysis model comprises a series connection of an electromagnetic signal coding network and a feature recognition network; the electromagnetic signal coding network comprises a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer extracts features of the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer enhances the features of the input shallow complex features to generate electromagnetic features; and the feature recognition network generates an electromagnetic data recognition result based on the input electromagnetic features. The electromagnetic data analysis model comprising the electromagnetic signal coding network and the feature recognition network is used innovatively, automatic learning and extraction of complex features of the electromagnetic signal are realized, deep features of the electromagnetic data can be effectively captured, and therefore the efficiency and accuracy of electromagnetic recognition are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to an electromagnetic data recognition method and device, electronic equipment and a storage medium. BACKGROUND

[0002] Electromagnetic recognition refers to analyzing acquired electromagnetic data to identify specific attributes of the electromagnetic data, such as communication individuals, modulation patterns or technical systems of the electromagnetic data.

[0003] At present, signal analysis tools are usually used in combination with expert rule libraries to perform electromagnetic recognition. The collected electromagnetic signals are processed by the signal analysis tools, and the transformed domain features of the electromagnetic signals are artificially designed and extracted, and then the extracted feature parameters are compared with the expert rule library, so as to realize electromagnetic recognition.

[0004] However, this method relies on artificially designed features and pre-established expert knowledge, and it is difficult to effectively capture the deep essential features of the electromagnetic data, thereby resulting in low electromagnetic recognition efficiency and low recognition accuracy. SUMMARY

[0005] The electromagnetic data recognition method, device, electronic equipment and storage medium provided by the present application solve the defects of low electromagnetic recognition efficiency and low recognition accuracy in the prior art.

[0006] The present application provides an electromagnetic data recognition method, comprising the following steps:

[0007] inputting an electromagnetic signal to be recognized into an electromagnetic data analysis model, and acquiring an electromagnetic data recognition result output by the electromagnetic data analysis model;

[0008] The electromagnetic data analysis model comprises a series connection of an electromagnetic signal coding network and a feature recognition network; the electromagnetic signal coding network comprises a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer extracts features of the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer enhances the features of the input shallow complex features to generate electromagnetic features; and the feature recognition network generates the electromagnetic data recognition result based on the input electromagnetic features.

[0009] According to the electromagnetic data recognition method provided by the present application, before the electromagnetic signal to be recognized is input into the electromagnetic data analysis model, the method further comprises:

[0010] performing Fourier transform on the electromagnetic signal to acquire an electromagnetic digital signal, wherein the electromagnetic digital signal is a complex signal containing in-phase components and quadrature components.

[0011] According to the electromagnetic data recognition method provided by the application, the complex convolution coding layer comprises a plurality of complex convolution blocks connected in series, each of the complex convolution blocks comprises a complex convolution layer, a complex batch normalization layer and a complex activation function layer.

[0012] The complex convolution layer performs complex convolution operation on the in-phase component and the quadrature component of the complex signal to generate a complex feature; the complex batch normalization layer performs normalization processing on the complex feature to generate a normalized complex feature; and the complex activation function layer performs nonlinear activation on the normalized complex feature to generate the shallow complex feature.

[0013] According to the electromagnetic data recognition method provided by the application, the complex feature enhancement layer comprises a complex multi-head attention mechanism module.

[0014] For each attention head of the complex multi-head attention mechanism module, the following processing is performed: generating a query matrix, a key matrix and a value matrix based on the shallow complex feature; and performing complex attention calculation based on the query matrix, the key matrix and the value matrix to generate a complex attention calculation result.

[0015] The complex attention calculation results generated by all the attention heads of the complex multi-head attention mechanism module are spliced and linearly transformed to generate the electromagnetic feature.

[0016] According to the electromagnetic data recognition method provided by the application, the electromagnetic signal coding network further comprises a multi-task selection layer, and the multi-task selection layer comprises:

[0017] A plurality of expert modules, each of which is used for processing a specific type of electromagnetic recognition task;

[0018] A routing module is configured to generate a probability distribution based on the electromagnetic feature, and select at least one expert module to process the electromagnetic feature based on the probability distribution.

[0019] According to the electromagnetic data recognition method provided by the application, the electromagnetic data analysis model further comprises a pixel recombination module, and the pixel recombination module processes the electromagnetic feature in the following manner:

[0020] The electromagnetic feature is segmented according to a preset window size to obtain a plurality of feature subsequences;

[0021] A plurality of feature elements that are continuous in time within each of the feature subsequences are combined to generate a local aggregated feature;

[0022] All the local aggregated features are randomly spliced;

[0023] The global aggregated features obtained after splicing are subjected to nonlinear transformation to generate recombined electromagnetic features.

[0024] According to the electromagnetic data recognition method provided by the application, the electromagnetic data analysis model further comprises a feature alignment mapping layer;

[0025] The feature alignment mapping layer generates aligned electromagnetic features matching the feature space dimension of the feature recognition network based on the recombined electromagnetic features;

[0026] The feature recognition network generates the electromagnetic data recognition result based on the input aligned electromagnetic features and text token sequence, wherein the text token sequence is obtained by performing word segmentation on the user input text query instruction;

[0027] The electromagnetic data recognition result comprises a text reply to the text query instruction, and the text reply comprises attribute recognition information of electromagnetic digital signals and / or knowledge question and answer information.

[0028] According to the electromagnetic data recognition method provided by the application, the attribute recognition information of the electromagnetic digital signals comprises any one of the following:

[0029] Communication individual identification information, used for identifying the transmission source individual of the electromagnetic signal;

[0030] Modulation pattern identification information, used for identifying the modulation mode of the electromagnetic signal;

[0031] Technical system identification information, used for identifying the communication technical system adopted by the electromagnetic signal.

[0032] According to the electromagnetic data recognition method provided by the application, the electromagnetic data analysis model is trained based on the following method:

[0033] An electromagnetic training sample set is collected, the electromagnetic training sample set comprises multiple groups of training samples, each group of training samples comprises an electromagnetic digital signal sample, attribute labels of the electromagnetic digital signal sample, and constructed electromagnetic text question and answer pairs; the electromagnetic text question and answer pairs comprise a text question and a standard text reply;

[0034] The electromagnetic signal coding network is trained based on the electromagnetic digital signal sample and the attribute labels until the electromagnetic attribute recognition accuracy reaches a first preset threshold;

[0035] The parameters of the feature recognition network are frozen, and a fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer is iteratively performed until a second preset stop condition is met;

[0036] The fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer comprises:

[0037] selecting a batch of electromagnetic text question and answer pairs from the electromagnetic training sample set;

[0038] inputting the electromagnetic digital signal and the text question into the electromagnetic data analysis model to obtain a predicted text reply;

[0039] calculating a loss between the predicted text reply and the standard text reply;

[0040] updating parameters of the electromagnetic signal coding network and the feature alignment mapping layer based on the loss.

[0041] The application further provides an electromagnetic data recognition device, comprising the following modules:

[0042] an electromagnetic signal input unit configured to input an electromagnetic signal to be recognized into an electromagnetic data analysis model;

[0043] a recognition result output unit configured to obtain an electromagnetic data recognition result output by the electromagnetic data analysis model;

[0044] The electromagnetic data analysis model comprises an electromagnetic signal coding network and a feature recognition network connected in series; the electromagnetic signal coding network comprises a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer performs feature extraction on the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer performs feature enhancement on the input shallow complex features to generate the electromagnetic features; and the feature recognition network generates the electromagnetic data recognition result based on the input electromagnetic features.

[0045] The application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the electromagnetic data recognition method according to any one of the above when executing the program.

[0046] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable on a processor to implement the electromagnetic data recognition method according to any one of the above.

[0047] The application further provides a computer program product comprising a computer program, wherein the computer program is executable on a processor to implement the electromagnetic data recognition method according to any one of the above.

[0048] The electromagnetic data recognition method, device, electronic device, and storage medium provided by the application use the electromagnetic data analysis model comprising an electromagnetic signal coding network and a feature recognition network for the first time, realize automatic learning and extraction of complex features of electromagnetic signals, can effectively capture deep features of electromagnetic data, and thus improve the efficiency and accuracy of electromagnetic recognition. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is one of the flowcharts of the electromagnetic data identification method provided by the present invention.

[0051] Figure 2 This is a schematic diagram of the electromagnetic signal encoding network provided by the present invention.

[0052] Figure 3 This is a flowchart illustrating the complex multi-head attention mechanism module provided by the present invention.

[0053] Figure 4 This is a flowchart illustrating the multi-task selection layer provided by the present invention.

[0054] Figure 5 This is a schematic diagram of the process by which the pixel recombination module provided by the present invention processes electromagnetic features.

[0055] Figure 6 This is a schematic diagram of the electromagnetic data analysis model provided by the present invention.

[0056] Figure 7 This is a schematic diagram of the training process of the electromagnetic data analysis model provided by the present invention.

[0057] Figure 8 This is a schematic diagram of the question-and-answer interface of the electromagnetic data analysis model provided by the present invention.

[0058] Figure 9 This is a schematic diagram of the electromagnetic data identification device provided by the present invention.

[0059] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0061] It should be noted that in the description of the present application, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or device. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0062] The terms "first", "second", and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally a class, and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" means at least one of the connected objects, and the character " / ", generally indicates that the objects before and after are in a "or" relationship.

[0063] The following will be described in conjunction with Figures 1-10 The electromagnetic data recognition method, device, electronic equipment and storage medium provided by the present application are described.

[0064] Figure 1 is one of the flowcharts of the electromagnetic data recognition method provided by the present application, as Figure 1 shown, the execution subject of the electromagnetic data recognition method provided by the present application can be a server, a cloud computing platform, or a computer capable of executing the method of the present application and other types of devices. In the absence of special description, the server is taken as an example in the subsequent embodiments.

[0065] As an optional embodiment, the electromagnetic data recognition method mainly includes but is not limited to the following steps:

[0066] Step 110, inputting the electromagnetic signal to be identified into an electromagnetic data analysis model.

[0067] The electromagnetic data analysis model mainly includes a series of electromagnetic signal coding network and feature recognition network. The electromagnetic signal coding network mainly includes a complex convolution coding layer and a complex feature enhancement layer. The complex convolution coding layer extracts features from the input electromagnetic signal to generate shallow complex features. The complex feature enhancement layer enhances the input shallow complex features to generate electromagnetic features. The feature recognition network generates electromagnetic data recognition results based on the input electromagnetic features.

[0068] The electromagnetic signal to be identified can be collected by professional collection equipment such as electronic reconnaissance equipment or software defined radio (SDR) devices, and the collected electromagnetic signal is usually sample point data containing in-phase components and quadrature components, and is in the form of one-dimensional time sequence.

[0069] The electromagnetic data analysis model can be an electromagnetic-text multi-modal large model. The electromagnetic signal to be identified can be input into the electromagnetic data analysis model, which can be realized by various ways such as file uploading and data stream interface.

[0070] The electromagnetic data analysis model includes an electromagnetic signal coding network and a feature recognition network in series, wherein the electromagnetic signal coding network is specially designed for processing electromagnetic signal characteristics. The electromagnetic signal coding network includes a complex convolution coding layer, which receives the input electromagnetic signal and uses its complex processing capability to perform preliminary feature coding and multi-scale feature extraction on the signal to generate shallow complex features. For example, the complex convolution coding layer can convert and down-sample the original electromagnetic signal x =( x 1 , x 2 , x n ) to obtain a shallow complex feature z =( z 1 , z 2 , z m ) that can preliminarily represent the signal.

[0071] The complex convolution coding layer is followed by a complex feature enhancement layer, which performs deeper information aggregation and context semantic extraction on the shallow complex features output by the complex convolution coding layer to generate electromagnetic features containing more rich information. For example, the complex feature enhancement layer can further process the shallow complex features to convert them into a high-level semantic complex feature sequence c =( c 1 , c 2 , c k ).

[0072] The electromagnetic features generated by the electromagnetic signal coding network are input into a feature recognition network, which can be a pre-trained large language model, for receiving the electromagnetic features and generating a final electromagnetic data recognition result based on the electromagnetic features. The electromagnetic data recognition result can be a classification judgment on a certain property of the electromagnetic signal, or a natural language text reply generated in combination with a user question, thereby realizing intelligent analysis and interactive question answering of the electromagnetic data.

[0073] In step 120, an electromagnetic data recognition result output by the electromagnetic data analysis model is obtained.

[0074] The electromagnetic data recognition result can be obtained in an interactive user interface. For example, an electromagnetic data analysis model question answering system can be constructed, in which a user submits an electromagnetic data file to be analyzed and inputs a text query instruction in a graphical user interface (GUI) of the electromagnetic data analysis model question answering system, and the electromagnetic data recognition result generated by the feature recognition network (such as a large language model) is directly displayed to the user in the form of natural language text in a conversation window of the interface. The output text content can not only be a direct analysis conclusion, but also include an explanation of the analysis process and an elaboration of related electromagnetic knowledge, thereby providing a rich interactive experience.

[0075] As another optional embodiment, the electromagnetic data recognition result is obtained through an application programming interface (API). In this way, other programs or systems can call the electromagnetic data analysis model of the present application, and the electromagnetic data analysis model returns the generated text reply in the form of structured data (such as JSON format), facilitating subsequent automatic processing and system integration.

[0076] The electromagnetic data recognition method provided by the present application uses the electromagnetic data analysis model including the electromagnetic signal coding network and the feature recognition network to realize automatic learning and extraction of complex features of the electromagnetic signal, effectively captures deep features of the electromagnetic data, and improves the efficiency and accuracy of electromagnetic recognition.

[0077] In another embodiment provided by the present application, before the electromagnetic signal to be recognized is input into the electromagnetic data analysis model, the method further includes: performing Fourier transform on the electromagnetic signal to obtain an electromagnetic digital signal, the electromagnetic digital signal being a complex signal including an in-phase component and a quadrature component.

[0078] Performing Fourier transform on the electromagnetic signal refers to a signal preprocessing step, which is used to convert the received original electromagnetic signal from time domain to frequency domain, and demodulate to baseband or intermediate frequency, thereby obtaining the in-phase component and the quadrature component.

[0079] The electromagnetic digital signal refers to a one-dimensional complex time sequence obtained by Fourier transform and digital sampling of the electromagnetic signal, which completely retains the amplitude and phase information of the original electromagnetic signal. For example, the electromagnetic digital signal is the electromagnetic signal sampling point data that can be directly input to the electromagnetic data analysis model, wherein each sampling point is a complex number composed of a real part (in-phase component) and an imaginary part (quadrature component).

[0080] The electromagnetic data recognition method provided by the application can completely retain the amplitude and phase key information in the original electromagnetic signal by performing Fourier transform on the electromagnetic signal before processing by the electromagnetic data analysis model, to obtain a complex electromagnetic digital signal containing an in-phase component and a quadrature component, thereby providing a standardized and information-complete input for subsequent complex network processing, thereby laying a foundation for the electromagnetic data analysis model to more accurately learn and extract electromagnetic features.

[0081] Figure 2 is a structural schematic diagram of the electromagnetic signal coding network provided by the application, as shown in Figure 2 As another optional embodiment provided by the application, the complex convolution coding layer includes a plurality of complex convolution blocks connected in series, and each complex convolution block includes a complex convolution layer, a complex batch normalization layer and a complex activation function layer.

[0082] Specifically, the complex convolution layer is responsible for performing complex convolution operation on the in-phase component and the quadrature component of the input complex signal to generate complex features. The complex convolution operation refers to extending the traditional convolution operation to the complex domain, wherein the input features and the convolution kernel weight are complex numbers, so that the inherent phase relationship between the in-phase component and the quadrature component in the electromagnetic signal can be better captured and maintained.

[0083] For example, when the input of the complex convolution layer is the electromagnetic digital signal X = a + ib , the complex convolution kernel weight is W = A + iB , and the convolution result can be calculated in the following manner:

[0084]

[0085] wherein, represents the complex convolution kernel weight, represents the electromagnetic digital signal, represents the convolution operation, represents the real part of the complex convolution kernel weight, represents the imaginary part of the complex convolution kernel weight, represents the real part of the electromagnetic digital signal, represents the imaginary part of the electromagnetic digital signal, i is the imaginary unit.

[0086] Further, the complex batch normalization layer normalizes the complex feature output by the complex convolution layer. This processing aims to adjust the distribution of the complex feature, for example, by jointly normalizing the statistics (such as mean and variance) of the real part and the imaginary part of the complex feature within a batch, so as to accelerate the model convergence and improve the stability of the training process.

[0087] The complex activation function layer performs nonlinear activation on the normalized complex feature to increase the nonlinear expression capability of the entire electromagnetic signal coding network, so that it can learn and fit more complex patterns in the electromagnetic signal. For example, the nonlinear activation can use a complex Gaussian error linear unit (complex GELU) as the activation function.

[0088] As an optional embodiment, the complex convolution layer is composed of five complex convolution blocks connected in series. The kernel size, stride and other parameters of each layer can be set to different values to realize multi-scale feature extraction and down-sampling of the input electromagnetic signal layer by layer, and finally output a shallow complex feature containing rich information for subsequent complex feature enhancement layer.

[0089] The electromagnetic data recognition method provided by the application can construct a stable deep complex network structure by combining the complex convolution layer, the complex batch normalization layer and the complex activation function layer into a serializable complex convolution block, while ensuring the stability of the training process, enhancing the nonlinear expression capability of the model for complex electromagnetic patterns, and thus realizing more deep and effective feature extraction of the electromagnetic signal.

[0090] Figure 3 is a flowchart of the complex multi-head attention mechanism module provided by the application, as shown in Figure 3 As another optional embodiment provided by the application, the complex feature enhancement layer includes a complex multi-head attention mechanism module, including but not limited to the following steps:

[0091] Step 310: For each attention head of the complex multi-head attention mechanism module, the following processing is performed: generating a query matrix, a key matrix and a value matrix based on the shallow complex feature; performing complex attention calculation based on the query matrix, the key matrix and the value matrix to generate a complex attention calculation result.

[0092] Specifically, for the input shallow complex feature Z = M + iN wherein M and N are the real part and the imaginary part of the complex feature sequence, respectively, which can be obtained by using a learnable real-valued weight matrix WQ 、 W K and W V Linearly project the shallow complex feature Z to get the complex-valued query matrix Q = ZW Q , the key matrix K = ZW K and the value matrix V = ZW V . Then the complex attention calculation is performed, the core idea of which is to calculate the complex matrix product of the transpose of Q 、 K , K T and V three, the complete expansion of the calculation process is as follows:

[0093]

[0094] wherein, represents the real part of the complex attention calculation result, represents the imaginary part of the complex attention calculation result, i is the imaginary unit.

[0095] Step 320, based on the complex attention calculation results generated by all attention heads of the complex multi-head attention mechanism module, the complex attention calculation results are spliced and linearly transformed to generate the electromagnetic features.

[0096] In order to efficiently implement the above complex attention calculation, the embodiment decomposes it into the combined operation of multiple parallel real number domain multi-head attention (MultiHead, MH) modules. Each MH module MH( Q , K , V ) performs a standard scaled dot-product attention calculation, that is, calculates the dot product of the query matrix and the key matrix to obtain the similarity score, normalizes it by the Softmax function to obtain the attention weight, and finally performs weighted summation on the value matrix.

[0097] Based on this, the above complex attention calculation can be realized by the following formula:

[0098]

[0099] wherein, Y is the input complex feature, M and N are the real part and the imaginary part thereof respectively, indicates that the query matrix, the key matrix and the value matrix thereof all come from the real part M , The query matrix comes from the real part M The key matrix and the value matrix come from the imaginary part N , The query matrix comes from the imaginary part N The key matrix comes from the real part M The value matrix comes from the imaginary part N , The query matrix and the key matrix come from the imaginary part N The value matrix comes from the real part M , The query matrix and the key matrix come from the real part M The value matrix comes from the imaginary part N , The query matrix and the key matrix come from the imaginary part N The value matrix comes from the real part M , The query matrix comes from the imaginary part N The key matrix and the value matrix come from the real part M , The query matrix, the key matrix and the value matrix all come from the imaginary part N .

[0100] It should be noted that each attention head of the complex multi-head attention mechanism focuses on a different subspace of the input features. The application concatenates the complex features output by all attention heads in the channel dimension, and then inputs the high-dimensional features after concatenation to a final linear transformation layer for feature fusion and dimension reduction, thereby generating electromagnetic features with high-level deep semantic as the final output of the layer. The entire complex feature enhancement layer, together with components such as residual connection (Add) and layer normalization (Norm), constitutes a complete complex Transformer Encoder layer, and can be stacked N times to enhance the feature extraction capability of the model.

[0101] The electromagnetic data recognition method provided by the application can effectively capture long-distance dependencies in the signal sequence while maintaining the complex characteristics of the electromagnetic signal through the complex multi-head attention mechanism, thereby providing more rich and accurate feature representation for subsequent recognition tasks.

[0102] Figure 4 is a flowchart of the multi-task selection layer provided by the application, as shown in Figure 4 As another optional embodiment provided by the application, the electromagnetic signal coding network further includes a multi-task selection layer, and the multi-task selection layer includes: a plurality of expert modules, each expert module being used for processing a specific type of electromagnetic recognition task; a routing module, used for generating a probability distribution based on the electromagnetic features, and selecting at least one expert module based on the probability distribution to process the electromagnetic features.

[0103] An electromagnetic recognition task refers to a task of classifying or analyzing a specific attribute of an electromagnetic signal. For example, an electromagnetic recognition task can be communication individual recognition, modulation pattern recognition (such as 2ASK, 8PSK, etc.), and technology system recognition (such as Bluetooth, WIFI, etc.) of an electromagnetic signal.

[0104] A probability distribution refers to a weight or score output by the routing module for each expert module, which represents the matching degree of each expert module in processing the current input electromagnetic feature. For example, for four expert modules, expert 1, expert 2, expert 3, and expert 4, the routing module can output a probability distribution such as [0.1, 0.8, 0.05, 0.05], at this time, the routing module will select the expert 2 with the highest probability value to participate in the calculation according to the probability distribution, and other expert modules are not activated, so as to realize flexible allocation of computing resources according to the input characteristics.

[0105] The electromagnetic data recognition method provided by the application can realize unified modeling of multiple electromagnetic recognition tasks such as communication individual, modulation pattern, etc. in a single model framework, and significantly improve the reasoning efficiency and task expansion capability of the electromagnetic data analysis model.

[0106] Figure 5 is a flowchart of processing electromagnetic features by the pixel recombination module provided by the application, as shown in Figure 5 As another optional embodiment provided by the application, the electromagnetic data analysis model further includes a pixel recombination module, which processes electromagnetic features in the following manner, including but not limited to the following steps:

[0107] Step 510: segmenting the electromagnetic features according to a preset window size to obtain a plurality of feature subsequences.

[0108] Specifically, the continuous one-dimensional electromagnetic feature sequence received from the complex feature enhancement layer is cut in the time dimension. For example, if the input electromagnetic feature is a time sequence with a length of 1024, and the preset window size is 64, the sequence will be evenly and non-overlappingly cut into 1024 / 64=16 continuous feature subsequences, and the length of each subsequence is 64.

[0109] Step 520: combining a plurality of feature elements that are continuous in time in each feature subsequence to generate a local aggregated feature.

[0110] Specifically, each longer feature sub-sequence is compressed into one or a few feature vectors that can represent its core information. The combination operation can be a pooling operation, such as performing mean pooling or max pooling on all feature elements within each feature sub-sequence. For example, after performing the pooling operation on the aforementioned 16 sub-sequences each of length 64, 16 corresponding local aggregated feature vectors are obtained.

[0111] Step 530, randomly splice all local aggregated features.

[0112] Specifically, this is the core step of implementing shuffle. It breaks the inherent order of individual feature segments in the original time sequence. For example, after obtaining 16 local aggregated features in the previous step, they still logically maintain the chronological order (numbered 1 to 16). This step generates a random index sequence, such as [5, 12, 2,..., 9], and then rearranges and splices the 16 local aggregated feature vectors according to the new random order to form a global aggregated feature that is related to the original input length or dimension but has been shuffled.

[0113] Step 540, perform nonlinear transformation on the global aggregated feature obtained after splicing to generate the shuffled electromagnetic feature.

[0114] Specifically, the global aggregated feature can be input into a nonlinear transformation module, which can typically be a multilayer perceptron (MLP). Through the fully connected layers and activation functions in the MLP, feature mapping and integration are performed, and the final output is the shuffled electromagnetic feature with stronger position invariance and generalization ability.

[0115] The electromagnetic data recognition method provided by the present application can break the fixed position dependence of features in the time sequence by randomly splicing and shuffling the local segments of electromagnetic features using the pixel shuffle module, so that the electromagnetic data analysis model pays more attention to the internal semantics of the features themselves, thereby significantly enhancing the generalization ability and recognition robustness of the electromagnetic data analysis model.

[0116] Figure 6 is a structural schematic diagram of the electromagnetic data analysis model provided by the present application, as Figure 6As shown, as another optional embodiment provided by the present application, the electromagnetic data analysis model further comprises a feature alignment mapping layer; the feature alignment mapping layer generates aligned electromagnetic features matching the feature space dimension of the feature recognition network based on the reorganized electromagnetic features; the feature recognition network generates an electromagnetic data recognition result based on the input aligned electromagnetic features and the text token sequence; the text token sequence is obtained by performing word segmentation processing on the user input text query instruction; the electromagnetic data recognition result comprises a text reply to the text query instruction, and the text reply comprises attribute recognition information of the electromagnetic digital signal and / or knowledge question and answer information.

[0117] Specifically, the feature alignment mapping layer is composed of a linear transformation layer and a GELU activation function, which projects and transforms the feature vector representing electromagnetic information output by the electromagnetic signal encoding network into a feature space with the same dimension as the feature recognition network (i.e., a large language model). The word segmenter can convert the user input text query instruction into a text token sequence that can be understood by the large language model. The large language model takes the aligned electromagnetic features and the text token sequence as input for cross-modal semantic understanding and reasoning, and generates a natural language text as a text reply to the text query instruction.

[0118] For example, the user inputs a segment of electromagnetic signal and inputs a question in the text box: "What is the modulation pattern of this electromagnetic signal". At this time, the feature alignment mapping layer converts the electromagnetic features extracted from the electromagnetic signal into aligned electromagnetic features, and the word segmenter converts the question into a text token sequence. After receiving the two inputs, the large language model outputs attribute recognition information, such as the text reply: "After electromagnetic feature extraction and pattern recognition, it is determined that the signal adopts 16-QAM modulation type." If the user continues to input a knowledge question and answer type question: "Explain what 16QAM modulation is", the large language model can further generate detailed explanatory text, thereby realizing intelligent question and answer of electromagnetic knowledge.

[0119] The electromagnetic data recognition method provided by the present application can align the electromagnetic features and the text across modalities through the introduction of the feature alignment mapping layer, and can perform joint understanding and generation on the fused information by using the large language model, thereby breaking through the limitation of the traditional recognition model that only outputs fixed category labels, and enhancing the human-computer interaction capability of the electromagnetic recognition system.

[0120] In another embodiment provided by the present application, the attribute recognition information of the electromagnetic digital signal comprises any one of the following: communication individual identification information, used for identifying the transmission source individual of the electromagnetic signal; modulation pattern identification information, used for identifying the modulation mode of the electromagnetic signal; technical system identification information, used for identifying the communication technical system adopted by the electromagnetic signal.

[0121] The emission source individual refers to a physical device entity that generates and emits electromagnetic signals.

[0122] The modulation mode of the electromagnetic signal refers to a specific technical means of loading a modulation signal onto a high-frequency carrier. For example, the modulation mode can be 2-level Amplitude-Shift Keying (2ASK), 8-Phase Shift Keying (8PSK), or 16-Quadrature Amplitude Modulation (16-QAM).

[0123] The communication technology system refers to a complete set of communication protocols, standards or system specifications followed by electromagnetic signals. For example, the communication technology system can be Bluetooth, WIFI, CDMA2000, and 5G New Radio (5GNR).

[0124] The electromagnetic data recognition method provided by the present application can specify the specific application direction and the actual problem solved by the present application by specificizing the electromagnetic data recognition result into key attribute information such as communication individual, modulation mode and technology system, so that it can directly serve the core business needs of the electromagnetic sensing knowledge recognition field.

[0125] Figure 7 is the training process schematic diagram of the electromagnetic data analysis model provided by the present application, as shown in Figure 7 As another optional embodiment provided by the present application, the electromagnetic data analysis model is trained based on the following mode:

[0126] Step 710, collect the electromagnetic training sample set, the electromagnetic training sample set includes multiple groups of training samples, each group of training samples includes an electromagnetic digital signal sample, an attribute label of the electromagnetic digital signal sample, and an electromagnetic text question and answer pair constructed; the electromagnetic text question and answer pair includes a text question and a standard text reply.

[0127] For example, the electromagnetic training sample can include a specific electromagnetic digital signal sample, an attribute label (such as a modulation mode label 8PSK) corresponding to the sample, and an electromagnetic text question and answer pair constructed based on the label, wherein the text question can be "What is the modulation mode of this electromagnetic signal?", and the standard text reply is "The electromagnetic signal adopts 8PSK modulation mode".

[0128] Step 720, train the electromagnetic signal coding network based on the electromagnetic digital signal sample and the attribute label until the electromagnetic attribute recognition accuracy reaches a first preset threshold.

[0129] Specifically, the electromagnetic signal coding network receives electromagnetic digital signal samples as input and outputs prediction results of the properties of the electromagnetic digital signal samples. By comparing the prediction results with the true property labels, calculating the loss using a cross-entropy loss function, and continuously optimizing the parameters of the electromagnetic signal coding network using a backpropagation algorithm until a preset stopping condition is met.

[0130] For example, for the training of the technical system identification task, the stopping condition can be set as the identification accuracy of the electromagnetic signal coding network on the validation set reaching a first preset threshold of 96%. Once the threshold is reached, it is considered that the coding network has been fully trained, and this pre-training phase is ended.

[0131] Step 730, freeze the parameters of the feature recognition network, and iteratively perform the fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer until a second preset stopping condition is met.

[0132] The fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer includes: selecting a batch of electromagnetic text question and answer pairs from the electromagnetic training sample set; inputting the electromagnetic digital signal and the text question into the electromagnetic data parsing model to obtain a predicted text answer; calculating the loss between the predicted text answer and the standard text answer; updating the parameters of the electromagnetic signal coding network and the feature alignment mapping layer based on the loss.

[0133] Considering that the feature recognition network (i.e., a large language model) itself already has general knowledge understanding and text generation capabilities, if its parameters are also updated during the fine-tuning process for the electromagnetic field, it may destroy the general knowledge learned during the pre-training phase, leading to a catastrophic decline in its performance in general question and answer scenarios. Therefore, the parameters of the feature recognition network are frozen during the fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer.

[0134] For example, when calculating the loss between the predicted text answer and the standard text answer, a cross-entropy loss function can be used to update the parameters of the electromagnetic signal coding network and the feature alignment mapping layer. The second preset stopping condition can be that the training reaches a preset number of iterations, or that the loss on the validation set no longer significantly decreases.

[0135] It should be noted that the evaluation indicators of the electromagnetic data analysis model can include recognition accuracy, bilingual evaluation understudy (BLEU), recall-oriented summary evaluation understudy (ROUGE), and metric for evaluation of translation with explicit ordering (METEOR). Among them, the recognition accuracy is used for the electromagnetic data classification task, the proportion of the number of correctly predicted samples to the total number of samples is calculated to reflect the basic judgment ability of the model. BLEU is used to measure the similarity of the text reply generated by the model and the standard text reply in the vocabulary segment to evaluate the sentence coherence of the generated content. ROUGE is used to focus on evaluating the recall rate of the generated content to the core information in the standard text reply to judge the integrity of the content. METEOR is used to integrate word matching and semantic matching, and gives a score closer to semantic consistency by considering synonyms, semantic similarity and word sequence difference.

[0136] The electromagnetic data recognition method provided by the present application can efficiently align the professional knowledge in the electromagnetic field to the semantic space of the large language model by adopting the two-stage training strategy of pre-training and alignment fine-tuning, and freezing the parameters of the feature recognition network in the alignment fine-tuning stage, while retaining the powerful general knowledge and text generation ability of the large language model, avoiding the catastrophic forgetting problem of the model in the training process, thereby improving the training efficiency and final performance of the electromagnetic data analysis model.

[0137] Figure 8 is the question and answer interface schematic diagram of the electromagnetic data analysis model provided by the present application, as Figure 8As shown, the present application provides a man-machine interaction platform for users, displays a "system" reply area on the left side of the interface and a "user" question area on the right side, and clearly presents the multi-round question and answer process between the user and the system. The system first feeds back "Data uploaded successfully", prompting the user that the electromagnetic data has been successfully uploaded, and then identifies and explains the modulation mode based on the input electromagnetic signal characteristics, for example, determines that the input signal adopts "16QAM modulation type", and describes the characteristics of 16QAM modulation and its typical application scenarios in digital communication in detail. The user can input text query instructions through the interface, such as "What is the modulation style of this data" "Explain what 16QAM modulation is", etc., and the system replies in natural language based on the cross-modal multi-task reasoning of electromagnetic data and text, the content covers attribute identification and knowledge question and answer of electromagnetic signals, and the intelligent interaction capability of the large language model and electromagnetic signal feature alignment mapping of the present application is exhibited. The interface is provided with a file selection box and a mode selection area, the user can conveniently upload the electromagnetic data file, select "electromagnetic data + text mode" or "pure text mode" as the interaction mode, input the question in the input bar, and click "Send question" to realize the question and answer service. The interface design realizes the complete working process from the electromagnetic signal raw data upload of the electromagnetic data analysis model to the cross-modal semantic understanding and knowledge question and answer output, and embodies the practicability and interactivity of the multi-modal electromagnetic analysis technology of the present application.

[0138] Figure 9 is a structural schematic diagram of an electromagnetic data recognition device provided by the present application, as Figure 9 shown, mainly includes but is not limited to:

[0139] The electromagnetic signal input unit 910 is used for inputting the electromagnetic signal to be recognized into the electromagnetic data analysis model.

[0140] The recognition result output unit 920 is used for acquiring the electromagnetic data recognition result output by the electromagnetic data analysis model.

[0141] The electromagnetic data analysis model includes a series connection of an electromagnetic signal coding network and a feature recognition network; the electromagnetic signal coding network includes a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer performs feature extraction on the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer performs feature enhancement on the input shallow complex features to generate the electromagnetic features; and the feature recognition network generates the electromagnetic data recognition result based on the input electromagnetic features.

[0142] It should be noted that the electromagnetic data recognition device provided by the present application can execute the electromagnetic data recognition method described in any of the above embodiments when it is actually operated, and the present embodiment will not be described here.

[0143] The electromagnetic data recognition device provided by the application initiates the use of an electromagnetic data analysis model containing an electromagnetic signal coding network and a feature recognition network, realizes automatic learning and extraction of complex features of electromagnetic signals, can effectively capture deep features of electromagnetic data, and thus improves the efficiency and accuracy of electromagnetic recognition.

[0144] Figure 10 is a structural schematic diagram of an electronic device provided by the application, as Figure 10 shown, the electronic device can include a processor (processor) 1010, a communications interface (Communications Interface) 1020, a memory (memory) 1030 and a communications bus 1040, wherein the processor 1010, the communications interface 1020, the memory 1030 complete mutual communication through the communications bus 1040. The processor 1010 can call the logic instructions in the memory 1030 to execute the electromagnetic data recognition method, which includes: inputting the electromagnetic signal to be recognized into the electromagnetic data analysis model, and obtaining the electromagnetic data recognition result output by the electromagnetic data analysis model; the electromagnetic data analysis model includes a series connection of an electromagnetic signal coding network and a feature recognition network; the electromagnetic signal coding network includes a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer extracts features of the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer enhances the features of the input shallow complex features to generate the electromagnetic features; the feature recognition network generates the electromagnetic data recognition result based on the input electromagnetic features.

[0145] In addition, the logic instructions in the memory 1030 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium, includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk and various program code storage media.

[0146] In another aspect, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions that, when executed by a computer, enable the computer to perform the electromagnetic data recognition method provided by any of the above embodiments, the method comprising: inputting an electromagnetic signal to be recognized into an electromagnetic data analysis model, and obtaining an electromagnetic data recognition result output by the electromagnetic data analysis model; the electromagnetic data analysis model comprising an electromagnetic signal coding network and a feature recognition network connected in series; the electromagnetic signal coding network comprising a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer performing feature extraction on the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer performing feature enhancement on the input shallow complex features to generate electromagnetic features; and the feature recognition network generating the electromagnetic data recognition result based on the input electromagnetic features.

[0147] In yet another aspect, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the electromagnetic data recognition method provided by any of the above embodiments, the method comprising: inputting an electromagnetic signal to be recognized into an electromagnetic data analysis model, and obtaining an electromagnetic data recognition result output by the electromagnetic data analysis model; the electromagnetic data analysis model comprising an electromagnetic signal coding network and a feature recognition network connected in series; the electromagnetic signal coding network comprising a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer performing feature extraction on the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer performing feature enhancement on the input shallow complex features to generate electromagnetic features; and the feature recognition network generating the electromagnetic data recognition result based on the input electromagnetic features.

[0148] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0149] Those skilled in the art can clearly understand the technical solutions of the various embodiments from the above description of the embodiments, and the various embodiments can be implemented by means of software with the necessary general hardware platforms, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, and the like, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0150] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of electromagnetic data recognition, characterized in that, The method comprises the steps of: inputting an electromagnetic signal to be identified into an electromagnetic data analysis model, and obtaining an electromagnetic data identification result output by the electromagnetic data analysis model; the electromagnetic data analysis model comprises an electromagnetic signal coding network and a feature recognition network connected in series; the electromagnetic signal coding network comprises a complex convolution coding layer and a complex feature enhancement layer; the complex convolution coding layer extracts features of the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer enhances the features of the input shallow complex features to generate electromagnetic features; the feature recognition network generates the electromagnetic data identification result based on the input electromagnetic features; the electromagnetic data analysis model further comprises a feature alignment mapping layer; the electromagnetic data analysis model is trained based on the following method: an electromagnetic training sample set is collected, the electromagnetic training sample set comprises multiple groups of training samples, each group of training samples comprises an electromagnetic digital signal sample, an attribute label of the electromagnetic digital signal sample, and an electromagnetic text question and answer pair; the electromagnetic text question and answer pair comprises a text question and a standard text answer; the electromagnetic signal coding network is trained based on the electromagnetic digital signal sample and the attribute label until the electromagnetic attribute identification accuracy reaches a first preset threshold; the parameters of the feature recognition network are frozen, and a fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer is iteratively performed until a second preset stop condition is met; the fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer comprises: selecting a batch of electromagnetic text question and answer pairs from the electromagnetic training sample set; inputting the electromagnetic digital signal and the text question into the electromagnetic data analysis model to obtain a predicted text answer; calculating the loss between the predicted text answer and the standard text answer; updating the parameters of the electromagnetic signal coding network and the feature alignment mapping layer based on the loss.

2. The electromagnetic data recognition method of claim 1, wherein, Before inputting the electromagnetic signal to be identified into the electromagnetic data analysis model, the method further comprises: performing Fourier transform on the electromagnetic signal to obtain an electromagnetic digital signal, the electromagnetic digital signal being a complex signal comprising in-phase components and quadrature components.

3. The electromagnetic data recognition method of claim 2, wherein, The complex convolution coding layer comprises multiple complex convolution blocks connected in series, and each complex convolution block comprises a complex convolution layer, a complex batch normalization layer, and a complex activation function layer; the complex convolution layer performs complex convolution operation on the in-phase components and quadrature components of the complex signal to generate complex features; the complex batch normalization layer normalizes the complex features to generate normalized complex features; and the complex activation function layer performs nonlinear activation on the normalized complex features to generate the shallow complex features.

4. The electromagnetic data recognition method of claim 1, wherein, The complex feature enhancement layer comprises a complex multi-head attention mechanism module; for each attention head of the complex multi-head attention mechanism module, the following processing is performed: generating a query matrix, a key matrix, and a value matrix based on the shallow complex features; and performing complex attention calculation based on the query matrix, the key matrix, and the value matrix to generate a complex attention calculation result. The complex attention calculation results generated by all attention heads of the complex multi-head attention mechanism module are spliced and linearly transformed to generate the electromagnetic feature.

5. The electromagnetic data recognition method of claim 1, wherein, The electromagnetic signal encoding network further includes a multi-task selection layer, which includes: a plurality of expert modules, each of which is used to process a specific type of electromagnetic identification task; the electromagnetic identification task refers to a task of attribute classification or analysis on the electromagnetic signal; a routing module configured to generate a probability distribution based on the electromagnetic feature and select at least one of the expert modules to process the electromagnetic feature based on the probability distribution.

6. The electromagnetic data recognition method of claim 1, wherein, The electromagnetic data analysis model further includes a pixel recombination module configured to process the electromagnetic feature in the following manner: segmenting the electromagnetic feature according to a preset window size to obtain a plurality of feature subsequences; combining a plurality of feature elements that are continuous in time within each of the feature subsequences to generate local aggregated features; randomly splicing all local aggregated features; nonlinearly transforming the globally aggregated features obtained after splicing to generate recombined electromagnetic features.

7. The electromagnetic data recognition method of claim 6, wherein, The feature alignment mapping layer generates aligned electromagnetic features matching the feature space dimension of the feature recognition network based on the recombined electromagnetic features; The feature recognition network generates the electromagnetic data recognition result based on the input aligned electromagnetic features and a text token sequence; the text token sequence is obtained by processing a text query instruction input by a user; The electromagnetic data recognition result includes a text reply to the text query instruction, and the text reply includes attribute identification information and / or knowledge question and answer information of the electromagnetic digital signal.

8. The electromagnetic data recognition method of claim 7, wherein, The attribute identification information of the electromagnetic digital signal includes any of the following: communication individual identification information, used to identify the transmission source individual of the electromagnetic signal; modulation pattern identification information, used to identify the modulation mode of the electromagnetic signal; technology system identification information, used to identify the communication technology system adopted by the electromagnetic signal.

9. An electromagnetic data recognition device, characterized by includes: an electromagnetic signal input unit configured to input an electromagnetic signal to be identified to an electromagnetic data analysis model; an identification result output unit configured to obtain an electromagnetic data recognition result output by the electromagnetic data analysis model; The electromagnetic data analysis model includes a series connection of an electromagnetic signal encoding network and a feature recognition network; the electromagnetic signal encoding network includes a complex convolution encoding layer and a complex feature enhancement layer; the complex convolution encoding layer extracts features from the input electromagnetic signal to generate shallow complex features; the complex feature enhancement layer enhances the features of the input shallow complex features to generate electromagnetic features; The feature recognition network generates the electromagnetic data recognition result based on the input electromagnetic features; The electromagnetic data analysis model further includes a feature alignment mapping layer; The electromagnetic data analysis model is trained in the following manner: The electromagnetic training sample set includes multiple groups of training samples, each group of training samples including an electromagnetic digital signal sample, an attribute label of the electromagnetic digital signal sample, and an electromagnetic text question and answer pair; the electromagnetic text question and answer pair includes a text question and a standard text answer; The electromagnetic signal coding network is trained based on the electromagnetic digital signal sample and the attribute label until the electromagnetic attribute recognition accuracy reaches a first preset threshold; The parameters of the feature recognition network are frozen, and a fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer is iteratively performed until a second preset stop condition is met; The fine-tuning process of the electromagnetic signal coding network and the feature alignment mapping layer includes: selecting a batch of the electromagnetic text question and answer pairs from the electromagnetic training sample set; inputting the electromagnetic digital signal and the text question into the electromagnetic data analysis model to obtain a predicted text answer; calculating the loss between the predicted text answer and the standard text answer; updating the parameters of the electromagnetic signal coding network and the feature alignment mapping layer based on the loss.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the electromagnetic data recognition method of any one of claims 1-8. 11.A non-transitory computer-readable storage medium having stored thereon a computer program. The computer program is executed by the processor to implement the electromagnetic data recognition method of any one of claims 1-8.

12. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the electromagnetic data recognition method of any one of claims 1-8.

Images