Radar signal identification method, device and equipment based on software and deep learning

Abstract

The invention relates to a radar signal identification method, device and equipment based on software and deep learning, and the method comprises the steps: firstly constructing a deep learning-based convolutional neural network model, and feeding a large amount of different types of radar signal data into the convolutional neural network model for training in combination with the convenience and flexibility of software radio. And the convolutional neural network model automatically performs signal feature extraction, so that the recognition accuracy of the radar signal is greatly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of radar signal processing, and relates to a radar signal recognition method, device and equipment based on software and deep learning. Background Art

[0002] Radar signal recognition plays an important role in electronic countermeasures. It can effectively obtain the valuable information transmitted by the opponent's radar, which is beneficial to the deployment of targeted countermeasure plans. The main task is to identify the type of the opponent's radar signal in order to determine the parameter information of the radar signal. However, in reality, there are many types and systems of radar signals. From the duration of transmission, they are divided into pulse signals and continuous wave signals. From the modulation mode, there are generally frequency modulation (such as linear frequency modulation, nonlinear frequency modulation, step frequency and frequency agility) and phase modulation (two-phase and multi-phase modulation). Nonlinear frequency modulation is divided into cosine frequency modulation, tangent frequency modulation, inverse tangent frequency modulation, hyperbolic frequency modulation and parabolic frequency modulation. Two-phase coding sequences are divided into Barker codes, M sequences and L sequences. Multi-phase coding sequences include Frank codes and P-phase codes (including P1, P2, P3 and P4 phase codes). Based on whether the phase is coherent or not, they can be divided into coherent and non-coherent signals. Based on the way the pulse repetition frequency changes, they can be divided into repetition frequency jitter, repetition frequency slip and repetition frequency variation. The above types of signals will produce various radar signals through free combination. Faced with such complex and diverse radar signals, it is still difficult to improve the recognition accuracy using traditional radar signal recognition methods. Summary of the invention

[0003] In response to the above-mentioned problems existing in traditional radar signal recognition, the present invention proposes a radar signal recognition method based on software and deep learning, a radar signal recognition device based on software and deep learning, and a computer device, which can greatly improve the accuracy of radar signal recognition.

[0004] In order to achieve the above objectives, the embodiments of the present invention adopt the following technical solutions: On the one hand, a radar signal recognition method based on software and deep learning is provided, comprising the steps of: Acquire radar signals to be identified through software radio frequency equipment; The radar signal to be identified is input into the trained radar signal identification model, the signal type of the radar signal to be identified is output, and the signal parameters of the radar signal to be identified are determined; wherein the training acquisition process of the radar signal identification model includes: The radar signal received by the antenna is obtained through the software radio RF device, and the radar signal is collected for a fixed duration, labeled, and stored in the radar signal database according to the signal category; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to data output; Take the labeled radar signals from the radar signal database as training sets and input them into the convolutional neural network model for model training; When the convolutional neural network model is trained and converged to a set degree, radar signals without labels are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches the set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

[0005] On the other hand, a radar signal recognition device based on software and deep learning is also provided, including: A signal acquisition module, used for acquiring radar signals to be identified through software radio frequency equipment; The identification output module is used to input the radar signal to be identified into the trained radar signal identification model, output the signal type of the radar signal to be identified and determine the signal parameters of the radar signal to be identified; wherein the training acquisition process of the radar signal identification model includes: The radar signal received by the antenna is obtained through the software radio RF device, and the radar signal is collected for a fixed duration, labeled, and stored in the radar signal database according to the signal category; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to data output; Take the labeled radar signals from the radar signal database as training sets and input them into the convolutional neural network model for model training; When the convolutional neural network model is trained and converged to a set degree, radar signals without labels are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches the set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

[0006] On the other hand, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned radar signal recognition method based on software and deep learning when executing the computer program.

[0007] One of the above technical solutions has the following advantages and beneficial effects: The above-mentioned radar signal recognition method, device and equipment based on software and deep learning first constructs a convolutional neural network model based on deep learning, and then feeds a large amount of radar signal data of different categories into the convolutional neural network model for training in combination with the convenience and flexibility of software radio, so that the convolutional neural network model can automatically extract the features of the signal, thereby greatly improving the recognition accuracy of the radar signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the conventional technology, the drawings required for use in the embodiments or the conventional technology descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0009] Figure 1 1 is a flowchart of a radar signal recognition method based on software and deep learning in one embodiment; Figure 2 A schematic diagram of a flow chart of a training process of a radar signal recognition model in one embodiment; Figure 3 A schematic diagram of the structure of a training system for a radar signal recognition method based on software and deep learning in one embodiment; Figure 4 Schematic diagram of the module structure of a radar signal recognition device based on software and deep learning in one embodiment. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and Examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0011] It should be noted that the reference to "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The presentation of this phrase at various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It will be appreciated by those skilled in the art that the embodiments described herein may be combined with other embodiments. The term "and / or" used in the specification of the present invention and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0012] The following will describe the implementation of the present invention in detail with reference to the accompanying drawings in the embodiment diagram of the present invention.

[0013] In one embodiment, Figure 1 As shown, a radar signal recognition method based on software and deep learning is provided, which may include the following processing steps S12 and S14: S12, acquiring a radar signal to be identified through a software radio frequency device; S14, inputting the radar signal to be identified into the trained radar signal identification model, outputting the signal type of the radar signal to be identified and determining the signal parameters of the radar signal to be identified. Figure 2 As shown, the training acquisition process of the radar signal recognition model includes steps S101 to S109: S101, obtain the radar signal received by the antenna through the software radio RF device, collect the radar signal for a fixed duration, label it, and store it in the radar signal database according to the signal category; wherein, the radar signal can be collected by collecting the baseband signal after down-conversion by the receiving antenna. The signal category of the radar signal is a priori knowledge. After the signal category is determined at the transmitting end, it is necessary to adjust as many different parameters of the signal as possible to obtain different radar signals. Labeling means labeling the collected signal with the corresponding signal category. The radar signal database may include the absolute path of the collected data of the radar signal and the corresponding signal category, as well as the collection duration (such as a fixed duration set according to the collection needs) and the corresponding IQ sample number (referring to the result of the system down-converting the radio frequency (RF) signal to a lower intermediate frequency (IF) or baseband when receiving the signal, and then sampling it).

[0014] S103, constructing a one-dimensional convolutional neural network model according to the set radar signal recognition index; the convolutional neural network model can specifically include a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to output; when these existing network layers in the field are used, their specific network parameters can be set according to the needs of the actual application, as long as they can be used to perform the required recognition tasks.

[0015] S105, taking out the labeled radar signals from the radar signal database as training sets and inputting them into the convolutional neural network model for model training.

[0016] S107, when the convolutional neural network model is trained to converge to a set degree, unlabeled radar signals are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; the set degree of model convergence can be selected according to actual training needs, for example, comprehensively considering factors such as model training rounds, output effect and training time, and selecting the corresponding convergence degree that can better balance the above factors to achieve the optimal training effect.

[0017] S109, if the recognition accuracy of the convolutional neural network model reaches a set threshold, the model training is terminated and the convolutional neural network model is used as a radar signal recognition model; That is, if the accuracy of the convolutional neural network model's output at its Softmax (multi-classification) layer reaches a set threshold (e.g., 99.5%), the model training is terminated and the convolutional neural network model is used as a radar signal recognition model.

[0018] In some embodiments, if the recognition accuracy of the convolutional neural network model does not reach the set threshold, return to step S105 for iterative training until the recognition accuracy of the convolutional neural network model reaches the set threshold to obtain a radar signal recognition model.

[0019] It can be understood that by adopting software-defined radio technology, various radar signal waveforms are generated on the host computer using software programming, and the waveform data is loaded into a unified hardware platform (existing software radio RF equipment in this field) for real-time transmission, and radiated at the air interface after connecting to a power amplifier and a horn antenna. Another device can collect the received radar signals and label them according to the signal type, so as to train the radar signal recognition model through supervised learning.

[0020] Specifically, the first step is to transmit various radar signals based on software radio and store various collected signal files according to signal categories.

[0021] The second step is to build a one-dimensional convolutional neural network model. The network structure can be innovated according to specific application needs, such as but not limited to adding transfer learning, attention mechanism layers and other functions. The network model is mainly adjusted to adapt to the current application scenario. By adopting the convolutional neural network model structure of this embodiment, the residual layer is introduced to alleviate the network degradation problem, and the batch normalization layer is introduced to speed up the network training while avoiding the problems of gradient disappearance and explosion, so that the model training converges faster, the feature extraction is deeper, and the classification effect is better.

[0022] The third step is to feed the signal file labeled in the first step to the model for large-scale training. The trained model is tested and the recognition accuracy is calculated. The generalization ability of the model is evaluated by splitting the training set and test set data in the existing cross-fold method. If the generalization ability does not meet the given design requirements, return to the second step.

[0023] The above radar signal recognition method based on software and deep learning first constructs a convolutional neural network model based on deep learning, and then feeds a large amount of radar signal data of different categories to the convolutional neural network model for training in combination with the convenience and flexibility of software radio, so that the convolutional neural network model can automatically extract the features of the signal, thereby greatly improving the recognition accuracy of the radar signal. Compared with traditional technologies, the above scheme makes full use of the flexibility of software radio, facilitates the use of various radar signal types, and has richer signal parameters, such as signal bandwidth, time width and signal-to-noise ratio, which can be adjusted at any time. For the convolutional neural network model of deep learning, the data level of the training data, the richness of the coverage scene and the balance of the samples are the most important, which directly affects the final signal recognition effect. Therefore, combined with the flexibility of software radio, the above training requirements can be well met.

[0024] Convolutional neural network models have achieved a lot of success in the field of image recognition. In order to learn from their advantages for radar signal recognition, various convolutional neural networks are conveniently used, such as residual neural networks, dilated convolutions, and attention mechanisms, which can be improved in combination with the characteristics of the radar signals to be recognized in actual application scenarios. The radar signal data used to train the model are all real signals directly obtained from the air interface, not simulated signal data, so they have a good practical reference value for real radar signal recognition.

[0025] In one embodiment, Figure 3As shown, the radar signal received by the (directional receiving) antenna is transmitted by the target device through the air interface. The target device includes a host computer, a software radio RF device, a power amplifier (i.e., a PA) and a directional transmitting antenna connected in sequence. A linear frequency modulation signal generator, a nonlinear frequency modulation signal generator, a multi-phase coded signal generator, a stepped frequency signal generator and a frequency agile signal generator are deployed in the host computer.

[0026] It can be understood that data can be transmitted between the host computer and the software radio RF device through interfaces such as USB3.0, Gigabit Ethernet port or 10 Gigabit optical fiber; similarly, on the side of receiving and identifying radar signals, data can be transmitted between the signal acquisition device and the software radio RF device through interfaces such as USB3.0, Gigabit Ethernet port or 10 Gigabit optical fiber, wherein the signal acquisition device, CNN network model (i.e., convolutional neural network model) and model evaluation module can be located on the same computer / server, so as to respectively undertake the radar signal acquisition function in the above-mentioned radar signal recognition method based on software and deep learning (the specific implementation of the acquisition function can be understood by referring to the existing radar signal acquisition process in the art), convolutional neural network model training and operation, and model accuracy evaluation and other functions.

[0027] By using the above-mentioned target device to directly send radar signal data for training from the air interface instead of using simulated signal data, various types of signal generators deployed by software programming in the host computer can be combined to generate various radar signals corresponding to actual real radar signal recognition scenarios. The signal type configuration is flexible and diverse, and various real radar signal data can be provided without relying on complicated and diverse hardware circuits. In this way, the obtained radar signal recognition model can be trained based on real radar signals, so that the radar signal recognition model is more suitable for the recognition and processing of real radar signals in real application scenarios, and ultimately the accuracy of radar signal recognition is further improved.

[0028] In one embodiment, the convolutional neural network model also introduces transfer learning and attention mechanism layers.

[0029] It can be understood that in this embodiment, transfer learning and attention mechanism layers are also introduced for the convolutional neural network model to achieve stronger migration application capabilities for different signal recognition, and enable the model to process input data more flexibly and focus on specific information or features. Specifically, the top layer of the pre-trained convolutional neural network model can be replaced with a structure suitable for the new recognition task, such as a classification layer, and then the model can be fine-tuned on the data set of the new recognition task to meet the needs of the new recognition task. The model can then be trained using the data set of the new recognition task, and the parameters can be adjusted to optimize its performance. An attention module can also be inserted between specific layers or multiple layers in the backbone network of the convolutional neural network model, or the attention mechanism can be integrated into the entire structure of the model, so that the model can process input data more flexibly and focus on specific information or features.

[0030] It should be understood that although Figure 1 and Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of the steps, and the steps can be executed in other orders. Moreover, Figure 1 and Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0031] In one embodiment, Figure 4 As shown, a radar signal recognition device based on software and deep learning is also provided, including a signal acquisition module 11 and an identification output module 13. The signal acquisition module 11 is used to acquire the radar signal to be identified through a software radio frequency device. The identification output module 13 is used to input the radar signal to be identified into a trained radar signal recognition model, output the signal type of the radar signal to be identified, and determine the signal parameters of the radar signal to be identified.

[0032] The training and acquisition process of the radar signal recognition model includes: The radar signal received by the antenna is obtained through the software radio RF device, and the radar signal is collected for a fixed duration, labeled, and stored in the radar signal database according to the signal category; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to data output; Take the labeled radar signals from the radar signal database as training sets and input them into the convolutional neural network model for model training; When the convolutional neural network model is trained and converged to a set degree, radar signals without labels are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches the set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

[0033] The above-mentioned radar signal recognition device based on software and deep learning first constructs a convolutional neural network model based on deep learning, and then feeds a large amount of radar signal data of different categories into the convolutional neural network model for training in combination with the convenience and flexibility of software radio, so that the convolutional neural network model can automatically extract the features of the signal, thereby greatly improving the recognition accuracy of the radar signal.

[0034] In one embodiment, the training acquisition process of the radar signal recognition model further includes: If the recognition accuracy of the convolutional neural network model does not reach the set threshold, return to the step of taking out the labeled radar signal from the radar signal database as the training set and inputting the convolutional neural network model for model training to perform iterative training until the recognition accuracy of the convolutional neural network model reaches the set threshold and then obtains the radar signal recognition model.

[0035] In one embodiment, the convolutional neural network model also introduces transfer learning and attention mechanism layers.

[0036] In one embodiment, the radar signal received by the antenna is transmitted by the target device through the air interface. The target device includes a host computer, a software radio frequency device, a power amplifier and a directional transmitting antenna connected in sequence. A linear frequency modulation signal generator, a nonlinear frequency modulation signal generator, a multi-phase coded signal generator, a stepped frequency signal generator and a frequency agile signal generator are deployed in the host computer.

[0037] The specific definition of the radar signal recognition device based on software and deep learning can be found in the above definition of the radar signal recognition method based on software and deep learning, which will not be repeated here. The above modules can be embedded in or independent of the device with radar signal processing function in hardware form, or can be stored in the memory of the above device in software form, so that the processor can call and execute the operations corresponding to the above modules. The above devices can be but not limited to various types of radar signal processing devices already available in the art.

[0038] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the following processing steps when executing the computer program: obtaining a radar signal to be identified through a software radio frequency device; inputting the radar signal to be identified into a trained radar signal recognition model, outputting a signal type of the radar signal to be identified and determining a signal parameter of the radar signal to be identified.

[0039] The training and acquisition process of the radar signal recognition model includes: The radar signal received by the antenna is obtained through the software radio RF device, and the radar signal is collected for a fixed duration, labeled, and stored in the radar signal database according to the signal category; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to data output; Take the labeled radar signals from the radar signal database as training sets and input them into the convolutional neural network model for model training; When the convolutional neural network model is trained and converged to a set degree, radar signals without labels are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches the set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

[0040] In one embodiment, when the processor executes the computer program, it can also implement the steps or sub-steps added in each embodiment of the above-mentioned radar signal recognition method based on software and deep learning.

[0041] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following processing steps are implemented: obtaining a radar signal to be identified through a software radio frequency device; inputting the radar signal to be identified into a trained radar signal recognition model, outputting a signal type of the radar signal to be identified and determining a signal parameter of the radar signal to be identified.

[0042] The training and acquisition process of the radar signal recognition model includes: The radar signal received by the antenna is obtained through the software radio RF device, and the radar signal is collected for a fixed duration, labeled, and stored in the radar signal database according to the signal category; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer from data input to data output; Take the labeled radar signals from the radar signal database as training sets and input them into the convolutional neural network model for model training; When the convolutional neural network model is trained and converged to a set degree, radar signals without labels are taken from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches the set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

[0043] In one embodiment, when the computer program is executed by a processor, the added steps or sub-steps in each embodiment of the above-mentioned radar signal recognition method based on software and deep learning can also be implemented.

[0044] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus dynamic random access memory (RambusDRAM, referred to as RDRAM) and interface dynamic random access memory (DRDRAM).

[0045] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0046] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of protection of the invention. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the attached claims.

Claims

1. A radar signal recognition method based on software and deep learning, characterized in that: Includes steps: Acquire radar signals to be identified through software radio frequency equipment; The radar signal to be identified is input into a trained radar signal identification model, the signal type of the radar signal to be identified is output, and the signal parameters of the radar signal to be identified are determined; wherein the training acquisition process of the radar signal identification model includes: Acquire radar signals received by the antenna through software radio frequency equipment, collect the radar signals for a fixed duration, annotate the radar signals with labels, and store them in a radar signal database according to signal categories; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer in sequence from data input to data output; Taking out the labeled radar signals from the radar signal database as training sets and inputting them into the convolutional neural network model for model training; When the convolutional neural network model is trained to converge to a set degree, radar signals without labels are taken out from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches a set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

2. The radar signal recognition method based on software and deep learning according to claim 1, characterized in that: The training acquisition process of the radar signal recognition model also includes: If the recognition accuracy of the convolutional neural network model does not reach the set threshold, return to the step of taking out the labeled radar signal from the radar signal database as a training set and inputting it into the convolutional neural network model for model training to perform iterative training until the recognition accuracy of the convolutional neural network model reaches the set threshold and then obtains the radar signal recognition model.

3. The radar signal recognition method based on software and deep learning according to claim 1 or 2, characterized in that: The convolutional neural network model also introduces transfer learning and attention mechanism layers.

4. A radar signal recognition device based on software and deep learning, characterized in that: include: A signal acquisition module, used for acquiring radar signals to be identified through software radio frequency equipment; The identification output module is used to input the radar signal to be identified into the trained radar signal identification model, output the signal type of the radar signal to be identified and determine the signal parameters of the radar signal to be identified; wherein the training acquisition process of the radar signal identification model includes: Acquire radar signals received by the antenna through software radio frequency equipment, collect the radar signals for a fixed duration, annotate the radar signals with labels, and store them in a radar signal database according to signal categories; A one-dimensional convolutional neural network model is constructed according to the set radar signal recognition index; the convolutional neural network model includes a one-dimensional convolution layer, a pooling layer, a batch normalization layer, multiple groups of residual layers, an excitation layer, an average pooling layer and a fully connected layer in sequence from data input to data output; Taking out the labeled radar signals from the radar signal database as training sets and inputting them into the convolutional neural network model for model training; When the convolutional neural network model is trained to converge to a set degree, radar signals without labels are taken out from the radar signal database as a test set and input into the convolutional neural network model for model testing; If the recognition accuracy of the convolutional neural network model reaches a set threshold, the model training is terminated and the convolutional neural network model is used as the radar signal recognition model.

5. The radar signal recognition device based on software and deep learning according to claim 4, characterized in that: The training acquisition process of the radar signal recognition model also includes: If the recognition accuracy of the convolutional neural network model does not reach the set threshold, return to the step of taking out the labeled radar signal from the radar signal database as a training set and inputting it into the convolutional neural network model for model training to perform iterative training until the recognition accuracy of the convolutional neural network model reaches the set threshold and then obtains the radar signal recognition model.

6. The radar signal recognition device based on software and deep learning according to claim 4 or 5, characterized in that: The convolutional neural network model also introduces transfer learning and attention mechanism layers.

7. The radar signal recognition device based on software and deep learning according to claim 6, characterized in that: The radar signal received by the antenna is transmitted by the target device through the air interface. The target device includes a host computer, a software radio frequency device, a power amplifier and a directional transmitting antenna connected in sequence. A linear frequency modulation signal generator, a nonlinear frequency modulation signal generator, a multi-phase coded signal generator, a stepped frequency signal generator and a frequency agile signal generator are deployed in the host computer.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the radar signal recognition method based on software and deep learning as described in any one of claims 1 to 3 are implemented.

Images