Electromagnetic Signal Detection and Classification Method and Device

Abstract

The present invention provides a method and apparatus for electromagnetic signal detection and classification, including: acquiring electromagnetic signal data to be processed; using a pre-trained signal detection and classification model to process the electromagnetic signal data to be processed, and obtaining a detection result and a classification result; wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; the feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result; the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result. The present invention has high classification accuracy for electromagnetic signal data to be processed with complex interference factors and various signal types, and the entire detection and classification process is more intelligent.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic signal acquisition and intelligent recognition, and particularly to a method and device for detecting and classifying electromagnetic signals. Background Art

[0002] With the rapid development of communication technologies, wireless electromagnetic signals are diverse in type and large in number, and the wireless channel has strong time-varying characteristics with numerous interference factors, resulting in an increasingly complex electromagnetic environment. In practical applications, wireless electromagnetic signals exhibit complex characteristics such as weak energy and multiple categories, posing significant challenges to the detection and classification of electromagnetic signals. Against this background, intelligent communication technology has become one of the mainstream directions in the development of the wireless communication field. Its basic idea is to break through the fixed signal processing architecture in traditional communication systems and, according to the changes in the external electromagnetic environment where the communication terminal is located, dynamically reconstruct the signal parameters of the communication system in real time, effectively combining machine learning and wireless communication technologies to improve resource utilization and achieve the purpose of communication anti-interference. Among them, wireless communication signal detection and classification technology plays a key role and has important research value for applications such as cognitive radio, intelligent transportation systems, electromagnetic spectrum monitoring, and electronic warfare.

[0003] Electromagnetic signal detection and classification technology refers to the process of detecting the presence of received signals and analyzing the signal types when the receiving end in non-cooperative communication is unknown or only has a small amount of prior knowledge. In non-cooperative communication scenarios, the receiver cannot obtain prior information such as the signal type, code rate, and bandwidth of the sending end. Signal demodulation and subsequent processing rely on electromagnetic signal detection and classification technology. This technology complements each other throughout the communication process, can reduce protocol overhead, and achieve signal demodulation and subsequent analysis processes. With the increasingly complex wireless electromagnetic environment, the performance of traditional electromagnetic signal detection and classification methods has severely declined. Therefore, there is an urgent need for an electromagnetic signal detection and classification method with high anti-interference ability and high recognition accuracy in complex environments. Summary of the Invention

[0004] The present invention provides a method and device for detecting and classifying electromagnetic signals to solve the above problems.

[0005] The present invention provides a method for detecting and classifying electromagnetic signals, including:

[0006] Obtaining electromagnetic signal data to be processed;

[0007] Processing the electromagnetic signal data to be processed by using a pre-trained signal detection and classification model to obtain a detection result and a classification result;

[0008] Wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network;

[0009] The feature extraction backbone network extracts features from the electromagnetic signal data to be processed, and obtains a mixed component feature map and a low-frequency component feature map;

[0010] The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result;

[0011] The classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

[0012] According to an electromagnetic signal detection and classification method provided by the present invention, the feature extraction backbone network includes a multi-scale high-low frequency grouped convolution layer;

[0013] Correspondingly, the feature extraction backbone network extracts features from the electromagnetic signal data to be processed, and obtains a mixed component feature map and a low-frequency component feature map, including:

[0014] Obtain a mixed component feature map and a low-frequency component feature map from the electromagnetic signal data to be processed, and respectively use them as the mixed component feature map and the low-frequency component feature map at the previous moment;

[0015] Perform downsampling on the mixed component feature map at the previous moment to obtain a downsampled mixed component feature map;

[0016] Perform upsampling on the low-frequency component feature map at the previous moment to obtain an upsampled low-frequency component feature map;

[0017] Obtain the mixed component feature map at the current moment according to the mixed component feature map at the previous moment and the upsampled low-frequency component feature map;

[0018] Obtain the low-frequency component feature map at the current moment according to the low-frequency component feature map at the previous moment and the downsampled mixed component feature map.

[0019] According to an electromagnetic signal detection and classification method provided by the present invention, the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result, including;

[0020] Perform masking processing on the low-frequency component feature map to obtain a detection result;

[0021] Correspondingly, the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result, including:

[0022] Perform Hadamard product on the mixed component feature map and the detection result to obtain a classification result.

[0023] An electromagnetic signal detection and classification method provided by the present invention, the obtaining of electromagnetic signal data to be processed includes:

[0024] Obtain initial electromagnetic signal data;

[0025] Perform radio frequency front-end processing on the initial electromagnetic signal data to obtain signal data after radio frequency front-end processing, and the radio frequency front-end processing at least includes gain adjustment, spectrum shifting, and low-pass filtering;

[0026] Perform digital sampling on the signal data after radio frequency front-end processing, and segment and save the sampled signal data according to the sampling information to obtain electromagnetic signal data to be processed.

[0027] An electromagnetic signal detection and classification method provided by the present invention, before the obtaining of the initial electromagnetic signal data, the method further includes:

[0028] Use a preset signal generation method to generate initial electromagnetic signal data in real time;

[0029] Wherein, the signal generation method at least includes binary source data generation, signal modulation and pulse shaping, and addition of interference factors.

[0030] An electromagnetic signal detection and classification method provided by the present invention, the using of a preset signal generation method to generate initial electromagnetic signal data in real time includes:

[0031] Obtain information in byte form and perform serial-to-parallel conversion on it to obtain binary source data;

[0032] Modulate the binary source data using a preset digital modulation method to obtain modulated source data;

[0033] Perform pulse shaping on the modulated source data using a preset pulse shaping method to obtain pulse-shaped source data;

[0034] Add multiple interference factors to the pulse-shaped source data to obtain initial electromagnetic signal data; wherein, the interference factors at least include thermal noise, frequency offset, clock offset, and multipath fading.

[0035] An electromagnetic signal detection and classification method provided by the present invention, after using a pre-trained signal detection and classification model to process the electromagnetic signal data to be processed to obtain a detection result and a classification result, the method further includes:

[0036] Display the electromagnetic signal data to be processed and the detection result and the classification result in a pre-constructed visualization interface.

[0037] The present invention also provides an electromagnetic signal detection and classification device, including:

[0038] A data acquisition module for acquiring electromagnetic signal data to be processed;

[0039] A detection and classification module for processing the electromagnetic signal data to be processed by using a pre-trained signal detection and classification model to obtain a detection result and a classification result;

[0040] Wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network;

[0041] The feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map;

[0042] The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result;

[0043] The classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

[0044] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements any of the above electromagnetic signal detection and classification methods.

[0045] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements any of the above electromagnetic signal detection and classification methods.

[0046] The electromagnetic signal detection and classification method and device provided by the present invention extract electromagnetic signal features from the electromagnetic signal data to be processed through a feature extraction backbone network to obtain feature maps of multiple spatial frequencies, thereby reducing feature redundancy and saving storage and computing resources. The detection branch network locates the electromagnetic signal region in the electromagnetic signal data to be processed, thereby filtering the noise background region and reducing the interference of noise on the classification feature map. The classification branch network realizes the classification of electromagnetic signals, and finally realizes the functions of electromagnetic signal detection and classification. Through the above pre-trained signal detection and classification model, the electromagnetic signal data to be processed is automatically detected and classified, and has high classification accuracy for the electromagnetic signal data with complex interference factors and various signal types. The entire detection and classification process is more intelligent. Description of the Drawings

[0047] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the accompanying drawings required for describing the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0048] Figure 1 It is one of the schematic flowcharts of the electromagnetic signal detection and classification method provided by the embodiments of the present invention;

[0049] Figure 2 It is the second schematic flowchart of the electromagnetic signal detection and classification method provided by the embodiments of the present invention;

[0050] Figure 3 It is the schematic structural diagram of the signal detection and classification model provided by the embodiments of the present invention;

[0051] Figure 4 It is the schematic structural diagram of the feature extraction backbone network provided by the embodiments of the present invention;

[0052] Figure 5 It is the implementation schematic diagram of the detection branch network and the classification branch network provided by the embodiments of the present invention;

[0053] Figure 6 It is the waveform schematic diagram corresponding to the electromagnetic signal data to be processed provided by the embodiments of the present invention.

[0054] Figure 7 It is the electromagnetic signal constellation diagram corresponding to the electromagnetic signal data to be processed provided by the embodiments of the present invention;

[0055] Figure 8 It is the waveform schematic diagram corresponding to the initial electromagnetic signal data provided by the embodiments of the present invention;

[0056] Figure 9 It is the visualization schematic diagram of the electromagnetic signal and the detection and classification results provided by the embodiments of the present invention;

[0057] Figure 10 It is the schematic structural diagram of the electromagnetic signal detection and classification device provided by the embodiments of the present invention;

[0058] Figure 11 It is the schematic physical structure diagram of an electronic device provided by the embodiments of the present invention. Detailed implementation manners

[0059] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0060] Currently, due to the great success of deep learning in aspects such as computer vision, speech recognition, natural language processing, and human-computer gaming, some scholars have gradually introduced deep learning methods into the electromagnetic field, enabling communication terminals to have the ability of self-learning and reconstruction to cope with the problems and challenges brought by complex interference factors and numerous signal types.

[0061] In addition, to achieve the detection and classification of electromagnetic signals in complex background noise, it is necessary to map the original signal data to generalize the differences between electromagnetic signals and background noise. Traditional electromagnetic feature extraction is based on model-driven methods and is not applicable to the case of complex background noise. The present invention uses a data-driven deep neural network to extract signal features. The following will specifically describe the electromagnetic signal detection and classification method and device provided by the present invention with reference to the accompanying drawings.

[0062] Figure 1 is one of the schematic flowcharts of the electromagnetic signal detection and classification method provided by the embodiments of the present invention; Figure 2 is the second schematic flowchart of the electromagnetic signal detection and classification method provided by the embodiments of the present invention.

[0063] As Figure 1 shown in and 2, the electromagnetic signal detection and classification method includes:

[0064] Step 101, obtain electromagnetic signal data to be processed.

[0065] In this step, the original electromagnetic signal data is received through a common electromagnetic signal receiving device (i.e., the receiving end). The original electromagnetic signal data is electromagnetic signal data with strong authenticity and rich signal types in a complex scenario. And the original electromagnetic signal data is subjected to radio frequency front-end processing and data acquisition processing to obtain the electromagnetic signal data to be processed.

[0066] Step 102, use a pre-trained signal detection and classification model to process the electromagnetic signal data to be processed, and obtain a detection result and a classification result.

[0067] As Figure 3 shown, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network.

[0068] The feature extraction backbone network extracts features from the electromagnetic signal data to be processed, obtaining a mixed component feature map and a low-frequency component feature map.

[0069] The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result.

[0070] The classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

[0071] In this step, a pre-trained signal detection and classification model is used to detect and classify the electromagnetic signal data to be processed. Among them, the feature extraction backbone network extracts electromagnetic signal features from the electromagnetic signal data to be processed to obtain feature maps of multiple spatial frequencies, thereby reducing feature redundancy and saving storage and computing resources. The detection branch network locates the electromagnetic signal region in the electromagnetic signal data to be processed, thereby filtering the noise background region and reducing the interference of noise on the classification feature map. The classification branch network realizes the classification of electromagnetic signals, and finally realizes the functions of electromagnetic signal detection and classification.

[0072] It should be noted that the trained signal detection and classification model is pre-trained according to the training set and the corresponding electromagnetic signal labels.

[0073] The electromagnetic signal detection and classification method provided by the embodiment of the present invention extracts electromagnetic signal features from the electromagnetic signal data to be processed through the feature extraction backbone network to obtain feature maps of multiple spatial frequencies, thereby reducing feature redundancy and saving storage and computing resources. The detection branch network locates the electromagnetic signal region in the electromagnetic signal data to be processed, thereby filtering the noise background region and reducing the interference of noise on the classification feature map. The classification branch network realizes the classification of electromagnetic signals, and finally realizes the functions of electromagnetic signal detection and classification. Through the above pre-trained signal detection and classification model, the electromagnetic signal data to be processed is automatically detected and classified, and has high classification accuracy for the electromagnetic signal data to be processed with complex interference factors and various signal types. The entire detection and classification process is more intelligent.

[0074] Figure 4 It is a schematic structural diagram of the feature extraction backbone network provided by the embodiment of the present invention; as Figure 4 shown, the feature extraction backbone network includes a multi-scale high-low frequency grouped convolution layer.

[0075] Correspondingly, the feature extraction backbone network extracts features from the electromagnetic signal data to be processed, obtaining a mixed component feature map and a low-frequency component feature map, including:

[0076] Obtain the mixed component feature map and the low-frequency component feature map from the electromagnetic signal data to be processed, and use them as the mixed component feature map and the low-frequency component feature map at the previous moment respectively.

[0077] Perform downsampling on the mixed component feature map at the previous moment to obtain the downsampled mixed component feature map.

[0078] Perform upsampling on the low-frequency component feature map at the previous moment to obtain the upsampled low-frequency component feature map.

[0079] Obtain the mixed component feature map at the current moment according to the mixed component feature map at the previous moment and the upsampled low-frequency component feature map.

[0080] Obtain the low-frequency component feature map at the current moment according to the low-frequency component feature map at the previous moment and the downsampled mixed component feature map.

[0081] Specifically, the feature extraction backbone network is multi-scale high-low frequency grouped convolution. This feature extraction backbone network has made adjustments in the convolution operation to reduce memory and computational costs. The input feature map corresponding to the electromagnetic signal data to be processed is divided into two groups: the mixed component and the low-frequency component. During the forward propagation of the signal detection and classification model, the feature map corresponding to the mixed component (i.e., Mixture Frequency) at the current moment is obtained by superimposing the convolution results of the feature map of the mixed component at the previous moment and the upsampled feature map of the low-frequency component feature map at the previous moment. The feature map corresponding to the low-frequency component (i.e., LowFrequency) at the current moment is obtained by superimposing the convolution results of the low-frequency component at the previous moment and the downsampled feature map of the mixed component feature map at the previous moment, realizing the update and interaction between different frequency components.

[0082] Among them, the channel allocation of the multi-scale high-low frequency grouped convolution is as follows:

[0083] c = ac+(1 - a)c = c1 + c2, a ∈ [0, 1]

[0084] In the formula, c is the total number of convolution channels, a is the channel ratio controlling the low-frequency component, c1 is the number of low-frequency component channels, and c2 is the number of high-frequency component channels.

[0085] The process of high-low frequency grouped convolution is as follows:

[0086]

[0087]

[0088] In the formula, represents the feature map of the i-th channel of the mixed component, The $i$-th channel feature map representing the low-frequency component, Pooling represents the downsampling process, and Upsample represents the upsampling process.

[0089] The final mixed component feature map at the current moment and the low-frequency component feature map at the current moment obtained through the above multi-scale high-low frequency grouped convolution calculation, the detection branch network and the classification branch network perform electromagnetic signal detection and classification based on the final mixed component feature map at the current moment and the low-frequency component feature map at the current moment.

[0090] The electromagnetic signal detection and classification method provided by the embodiment of the present invention obtains the mixed component feature map at the current moment according to the mixed component feature map at the previous moment and the upsampled low-frequency component feature map; obtains the low-frequency component feature map at the current moment according to the low-frequency component feature map at the previous moment and the downsampled mixed component feature map, realizing the update and interaction between different frequency components.

[0091] Figure 5 It is a schematic diagram of the implementation of the detection branch network and the classification branch network provided by the embodiment of the present invention; as Figure 5 shown, the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result, including:

[0092] Performing masking processing on the low-frequency component feature map to obtain a detection result.

[0093] Correspondingly, the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result, including:

[0094] Performing Hadamard product on the mixed component feature map and the detection result to obtain a classification result.

[0095] In this embodiment, the electromagnetic signal detection and classification are combined into one, divided into two branch networks of detection and classification, and the detection task and the classification task are respectively executed. The process of executing the detection task and the classification task is realized based on the classification region screening architecture of the Mask mechanism, which completes detection and classification based on two groups of feature maps of mixed components and low-frequency components obtained by multi-scale high-low frequency grouped convolution.

[0096] It should be noted that compared with background noise, most of the energy of electromagnetic signals is concentrated in the low-frequency band. Therefore, the low-frequency feature map is used for detection; and the more information-rich mixed component feature map is used in the classification process.

[0097] A Mask is constructed in the detection branch network, and the value in the Mask represents the probability that the data at that moment is an electromagnetic signal. In the classification branch network, the detection result of the detection task is multiplied by the mixed component feature map in the branch network through Hadamard product to obtain the feature map corresponding to the classification result (i.e., Classification). This is used to limit the magnitude of the values in the noise position feature map. During the forward inference process of the detection branch network, the calculation of the Mask is as follows:

[0098]

[0099] In the formula, Mask i represents the i-th mask, that is, the objectivity score of the electromagnetic signal. The closer this score is to 1, the greater the probability that the moment is an electromagnetic signal, and the closer it is to 0, the more likely it is background noise at that moment, FM low represents the output low-frequency component feature map.

[0100] The feature map of the classification branch network is calculated as follows:

[0101]

[0102] In the formula, FM mix represents the output mixed component feature map, and ⊙ represents the matrix Hadamard product.

[0103] The electromagnetic signal detection and classification method provided by the embodiments of the present invention performs masking processing on the low-frequency component feature map to obtain a detection result; performs Hadamard product on the mixed component feature map and the detection result to obtain a classification result, thereby being able to filter the noise background area, reduce the interference of noise on the classification feature map, and realize the functions of electromagnetic signal detection and classification.

[0104] Further, the obtaining of the electromagnetic signal data to be processed includes:

[0105] Obtain the initial electromagnetic signal data.

[0106] Perform radio frequency front-end processing on the initial electromagnetic signal data to obtain the signal data after radio frequency front-end processing. The radio frequency front-end processing at least includes gain adjustment, spectrum shifting, and low-pass filtering.

[0107] Perform digital sampling on the signal data after radio frequency front-end processing, and segment and save the sampled signal data according to the sampling information to obtain the electromagnetic signal data to be processed.

[0108] Specifically, initial electromagnetic signal data is obtained through a common electromagnetic signal receiving device, and this initial electromagnetic signal data is sent by a common electromagnetic signal transmitting device (i.e., the transmitting end).

[0109] After receiving the initial electromagnetic signal data, the initial electromagnetic signal is down-converted to an intermediate frequency through a radio frequency front-end circuit to obtain the signal data processed by the radio frequency front-end. Among them, the radio frequency front-end processing (i.e., frequency conversion processing) includes gain adjustment, spectrum shifting, low-pass filtering, etc.

[0110] Digital sampling is performed on the signal data processed by the radio frequency front-end. After digital sampling, the electromagnetic signal data stream is read based on the USRP Source module. The output format of the USRP Source module is Complex to meet the transmission requirements of the I / Q two-way electromagnetic signals.

[0111] In this embodiment, the data stream is sampled and segmented through a custom sample sampling module. This custom sample sampling module has a clock synchronization function. The custom parameters include sampling information such as the number of sample sampling points, the number of samples, and the file saving location.

[0112] In addition, since a large amount of information can be collected by radio frequency signals in a short time, to reduce resource usage, a sample sampling interval time is set. The collected electromagnetic signal data is saved in the NumPy format, which is convenient for waveform visualization and the use of deep learning models.

[0113] Based on the electromagnetic signal data stream obtained through the above-mentioned radio frequency front-end processing and digital sampling, and according to the parameters such as the number of sample sampling points, the number of samples, the file saving location, and the sampling interval time mentioned above, the collected electromagnetic signal data is segmented and saved, providing input for the signal detection and classification model and the visualization of the electromagnetic signal waveform. The obtained electromagnetic signal data to be processed is as Figure 6 shown. The real and imag waveforms are the in-phase and quadrature electromagnetic signals respectively, that is, the I / Q two-way electromagnetic signals.

[0114] The electromagnetic signal detection and classification method provided by the embodiment of the present invention obtains electromagnetic signal data with strong authenticity and rich signal types by collecting wireless electromagnetic signals in real time and truly.

[0115] Further, before obtaining the initial electromagnetic signal data, the method further includes:

[0116] Using a preset signal generation method to generate initial electromagnetic signal data in real time.

[0117] Among them, the signal generation method at least includes binary source data generation, signal modulation and pulse shaping, and addition of interference factors.

[0118] The real-time generation of initial electromagnetic signal data using a preset signal generation method specifically includes:

[0119] Obtain information in byte form and perform serial-to-parallel conversion on it to obtain binary source data.

[0120] Modulate the binary source data using a preset digital modulation method to obtain modulated source data.

[0121] Perform pulse shaping on the modulated source data using a preset pulse shaping method to obtain pulse-shaped source data.

[0122] Add multiple interference factors to the pulse-shaped source data to obtain initial electromagnetic signal data.

[0123] Among them, the interference factors at least include thermal noise, frequency offset, clock offset, and multipath fading.

[0124] Specifically, the generation of binary source data is completed through GNU Radio software. Extract text information from Shakespeare's literary works, and perform serial-to-parallel conversion on the text information represented in byte form to obtain a binary information stream (i.e., binary source data).

[0125] Modulate the binary source data through a custom modulation method class. The digital modulation methods include BPSK, QPSK, 8PSK, PAM4, QAM16, QAM64, GFSK, CPGSK, etc. Introduce various constellation diagrams (the corresponding constellation diagrams in the electromagnetic signal receiving device are as shown in Figure 7 to reduce code redundancy. And use a root-raised cosine filter to achieve pulse shaping to obtain a baseband signal. During the pulse shaping process, it is achieved by setting parameters such as the Nyquist sampling frequency, roll-off factor, energy gain, and the number of resampling filters.

[0126] Add actual interference factors such as thermal noise, frequency offset, clock offset, and multipath fading to the pulse-shaped source data. Then set the communication hardware parameters through the USRP sink module. The communication hardware parameters include the center frequency, gain, antenna selection, etc., to observe the waveform in real time, as shown in Figure 8 shown.

[0127] The electromagnetic signal detection and classification method provided by the embodiments of the present invention makes the electromagnetic signal data highly authentic and rich in signal types by real-time and real sending of wireless electromagnetic signals.

[0128] Further, after processing the to-be-processed electromagnetic signal data using the pre-trained signal detection and classification model to obtain the detection result and classification result, the method further includes:

[0129] Display the electromagnetic signal data to be processed, as well as the detection results and classification results, in a pre-constructed visualization interface, such as Figure 9 shown. The left part is the signal detection result. Among them, the light-colored part is the detected electromagnetic signal, and the dark-colored part is the noise background area; the right part is the classification result based on the detected electromagnetic signals, and this classification result includes four categories: ASK signal, PSK signal, QAM signal, and AM signal.

[0130] Specifically, the pre-constructed visualization interface is developed based on the Linux system operation interface. The functions of each module of the system are written in the Python language, and the human-computer interaction interface design is implemented through QT GUI. The electromagnetic signal waveform data and the detection and classification results are displayed through the visualization interface.

[0131] Next, the electromagnetic signal detection and classification device provided by the present invention will be described. The electromagnetic signal detection and classification device described below can be mutually corresponding and referenced to the electromagnetic signal detection and classification method described above.

[0132] Figure 10 is a schematic structural diagram of the electromagnetic signal detection and classification device provided by an embodiment of the present invention, such as Figure 10 shown. An electromagnetic signal detection and classification device includes:

[0133] A data acquisition module 1101, configured to acquire electromagnetic signal data to be processed.

[0134] In this module, the original electromagnetic signal data is received through a common electromagnetic signal receiving device. The original electromagnetic signal data is electromagnetic signal data with strong authenticity and rich signal types in a complex scenario. And the original electromagnetic signal data is subjected to radio frequency front-end processing and data acquisition processing to obtain the electromagnetic signal data to be processed.

[0135] A detection and classification module 1102, configured to process the electromagnetic signal data to be processed by using a pre-trained signal detection and classification model to obtain detection results and classification results.

[0136] Among them, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; the feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain detection results; the classification branch network performs classification based on the mixed component feature map and the detection results to obtain classification results.

[0137] In this module, a pre-trained signal detection and classification model is used to detect and classify the electromagnetic signal data to be processed. Among them, the electromagnetic signal features of the electromagnetic signal data to be processed are extracted through the feature extraction backbone network to obtain feature maps of multiple spatial frequencies, thereby reducing feature redundancy and saving storage and computing resources. The electromagnetic signal region in the electromagnetic signal data to be processed is located through the detection branch network, thereby filtering the noise background region and reducing the interference of noise on the classification feature map. The classification of electromagnetic signals is achieved through the classification branch network, and finally the functions of electromagnetic signal detection and classification are realized.

[0138] The electromagnetic signal detection and classification device provided by the embodiment of the present invention extracts the electromagnetic signal features of the electromagnetic signal data to be processed through the feature extraction backbone network to obtain feature maps of multiple spatial frequencies, thereby reducing feature redundancy and saving storage and computing resources. The electromagnetic signal region in the electromagnetic signal data to be processed is located through the detection branch network, thereby filtering the noise background region and reducing the interference of noise on the classification feature map. The classification of electromagnetic signals is achieved through the classification branch network, and finally the functions of electromagnetic signal detection and classification are realized. Through the above-mentioned pre-trained signal detection and classification model, the electromagnetic signal data to be processed is automatically detected and classified, and it has high classification accuracy for the electromagnetic signal data to be processed with complex interference factors and various signal types, and the entire detection and classification process is more intelligent.

[0139] Figure 11 The following is a schematic physical structure diagram of an electronic device provided by an embodiment of the present invention, as Figure 11 shown, the electronic device may include: a processor 1210, a communication interface 1220, a memory 1230, and a communication bus 1240. Among them, the processor 1210, the communication interface 1220, and the memory 1230 complete mutual communication through the communication bus 1240. The processor 1210 can call the logical instructions in the memory 1230 to execute the electromagnetic signal detection and classification method, and the electromagnetic signal detection and classification method includes: obtaining the electromagnetic signal data to be processed; using a pre-trained signal detection and classification model to process the electromagnetic signal data to be processed to obtain a detection result and a classification result; wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; the feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result; the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

[0140] In addition, when the logical instructions in the above-mentioned memory 1230 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0141] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is used to implement the electromagnetic signal detection and classification method provided by the above method. The electromagnetic signal detection and classification method includes: obtaining electromagnetic signal data to be processed; using a pre-trained signal detection and classification model to process the electromagnetic signal data to be processed to obtain a detection result and a classification result; wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; the feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; the detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result; the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

[0142] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0143] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiments.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; 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 each embodiment of the present invention.

Claims

1. An electromagnetic signal detection and classification method, characterized in that, including: Obtaining electromagnetic signal data to be processed; Processing the electromagnetic signal data to be processed by using a pre-trained signal detection and classification model to obtain a detection result and a classification result; Wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; The feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result; The classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

2. The electromagnetic signal detection and classification method according to claim 1, characterized in that The feature extraction backbone network includes a multi-scale high-low frequency grouped convolutional layer; Correspondingly, the feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map, including: Obtaining a mixed component feature map and a low-frequency component feature map from the electromagnetic signal data to be processed, and respectively serving as the mixed component feature map and the low-frequency component feature map at the previous moment; Performing downsampling on the mixed component feature map at the previous moment to obtain a downsampled mixed component feature map; Performing upsampling on the low-frequency component feature map at the previous moment to obtain an upsampled low-frequency component feature map; Obtaining the mixed component feature map at the current moment according to the mixed component feature map at the previous moment and the upsampled low-frequency component feature map; Obtaining the low-frequency component feature map at the current moment according to the low-frequency component feature map at the previous moment and the downsampled mixed component feature map.

3. The electromagnetic signal detection and classification method according to claim 2, wherein, The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result, including; Performing masking processing on the low-frequency component feature map to obtain a detection result; Correspondingly, the classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result, including: Performing Hadamard product on the mixed component feature map and the detection result to obtain a classification result.

4. The electromagnetic signal detection and classification method according to claim 1, characterized in that The obtaining of the electromagnetic signal data to be processed includes: Obtaining initial electromagnetic signal data; Performing radio frequency front-end processing on the initial electromagnetic signal data to obtain signal data after radio frequency front-end processing, where the radio frequency front-end processing at least includes gain adjustment, spectrum shifting, and low-pass filtering; Performing digital sampling on the signal data after radio frequency front-end processing, and segmenting and storing the sampled signal data according to sampling information to obtain the electromagnetic signal data to be processed.

5. The electromagnetic signal detection and classification method according to claim 4, wherein Before the obtaining of the initial electromagnetic signal data, the method further includes: Generating initial electromagnetic signal data in real time by using a preset signal generation method; Wherein, the signal generation method at least includes binary source data generation, signal modulation and pulse shaping, and interference factor addition.

6. The electromagnetic signal detection and classification method according to claim 5, characterized in that The generating of the initial electromagnetic signal data in real time by using a preset signal generation method includes: Obtaining information in byte form and performing serial-to-parallel conversion on it to obtain binary source data; Modulate the binary source data using a preset digital modulation method to obtain modulated source data; Perform pulse shaping on the modulated source data using a preset pulse shaping method to obtain pulse-shaped source data; Add multiple interference factors to the pulse-shaped source data to obtain initial electromagnetic signal data; wherein, the interference factors at least include thermal noise, frequency offset, clock offset, and multipath fading.

7. The electromagnetic signal detection and classification method according to claim 1, wherein After processing the electromagnetic signal data to be processed using a pre-trained signal detection and classification model to obtain a detection result and a classification result, the method further includes: Display the electromagnetic signal data to be processed, as well as the detection result and the classification result, in a pre-constructed visualization interface.

8. An electromagnetic signal detection and classification device, characterized in that, Comprising: A data acquisition module for acquiring electromagnetic signal data to be processed; A detection and classification module for processing the electromagnetic signal data to be processed using a pre-trained signal detection and classification model to obtain a detection result and a classification result; Wherein, the pre-trained signal detection and classification model includes a feature extraction backbone network, a detection branch network, and a classification branch network; The feature extraction backbone network extracts features from the electromagnetic signal data to be processed to obtain a mixed component feature map and a low-frequency component feature map; The detection branch network performs electromagnetic signal detection based on the low-frequency component feature map to obtain a detection result; The classification branch network performs classification based on the mixed component feature map and the detection result to obtain a classification result.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the electromagnetic signal detection and classification method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the electromagnetic signal detection and classification method according to any one of claims 1 to 7.

Images