Radio frequency fingerprint data enhancement method, device and equipment based on distillation diffusion

Abstract

The embodiment of the invention provides a radio frequency fingerprint data enhancement method, device and equipment based on distillation diffusion. The method comprises the following steps: preprocessing a pre-collected radio frequency signal to obtain a training data set; training a pre-constructed teacher model according to the training data set to obtain an optimized teacher model; training a pre-constructed student model according to the training data set, and optimizing the student model through a student network distillation loss function in the training process to obtain an optimized student model; and iterating the optimized student model for multiple times to generate a new signal sample, and adding the generated signal sample into a training data set for training a ResNet classification model. According to the technical scheme of the embodiment of the invention, the fine features of the signal can be effectively learned, the generation of various training samples is guided, and the generalization ability of the model is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of radio frequency fingerprint recognition, and in particular, to a radio frequency fingerprint data enhancement method, device and equipment based on diffusion distillation. Background Art

[0002] Radio frequency fingerprint recognition technology is a technology that identifies the identity of wireless devices by analyzing the received radio frequency signals of transmitters and obtaining the subtle feature differences hidden in the transmitted waveforms. These subtle feature differences mainly come from the distortion characteristics of electronic components such as I / Q imbalance and power amplifier non-linearity. As an additional security layer for wireless communication, radio frequency fingerprint recognition technology is an important means to improve the physical layer security performance in wireless networks and has broad application prospects in fields such as vehicle-to-everything (V2X) wireless communication, automatic identification systems for ships, and industrial Internet. However, due to the spatio-temporal superposition characteristics of wireless signals, it is difficult to label them, resulting in insufficient generalization ability of existing models and limited authentication performance. In current technical solutions, most methods use the generative adversarial network (GAN) framework for data enhancement. However, GAN needs to reach the Nash equilibrium point in the dynamic game process between the generator and the discriminator network. The generator may generate similar samples to deceive the discriminator and cannot effectively cover the distribution of real data and retain the subtle features of the signals. Therefore, how to effectively learn the subtle features of signals and then guide the generation of diverse training samples to ensure the generalization ability of the model has become an urgent technical problem to be solved. Summary of the Invention

[0003] Embodiments of the present application provide a radio frequency fingerprint data enhancement method, device and equipment based on diffusion distillation, which can, at least to a certain extent, effectively learn the subtle features of signals, and then guide the generation of diverse training samples to ensure the generalization ability of the model.

[0004] Other features and advantages of the present application will become apparent through the following detailed description, or be learned in part through the practice of the present application.

[0005] According to one aspect of the embodiments of the present application, a radio frequency fingerprint data enhancement method based on diffusion distillation is provided, including:

[0006] Preprocessing the pre-collected radio frequency signals to obtain a training data set;

[0007] Training a pre-constructed teacher model according to the training data set to obtain an optimized teacher model;

[0008] Training a pre-constructed student model according to the training data set, and optimizing the student model through a student network distillation loss function during the training process to obtain an optimized student model;

[0009] Iteratively use the optimized student model multiple times to generate new signal samples, and add the generated signal samples to the training dataset for training the ResNet classification model.

[0010] According to one aspect of the embodiments of the present application, there is provided a radio frequency fingerprint data enhancement device based on distillation diffusion, including:

[0011] A preprocessing module for preprocessing the pre-collected radio frequency signals to obtain a training dataset;

[0012] A first optimization module for training a pre-constructed teacher model according to the training dataset to obtain an optimized teacher model;

[0013] A second optimization module for training a pre-constructed student model according to the training dataset, and during the training process, optimizing the student model through a student network distillation loss function to obtain an optimized student model;

[0014] A processing module for iteratively using the optimized student model multiple times to generate new signal samples, and adding the generated signal samples to the training dataset for training the ResNet classification model.

[0015] According to one aspect of the embodiments of the present application, there is provided a computer-readable medium having a computer program stored thereon, and when the computer program is executed by a processor, it implements the radio frequency fingerprint data enhancement method based on distillation diffusion as described in the above embodiments.

[0016] According to one aspect of the embodiments of the present application, there is provided an electronic device, including: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the radio frequency fingerprint data enhancement method based on distillation diffusion as described in the above embodiments.

[0017] According to one aspect of the embodiments of the present application, there is provided a computer program product or a computer program, the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the radio frequency fingerprint data enhancement method provided in the above embodiments.

[0018] In the technical solutions provided by some embodiments of the present application, a training data set is obtained by preprocessing the pre-collected radio frequency signals; the pre-constructed teacher model is trained according to the training data set to obtain an optimized teacher model, and then the pre-constructed student model is trained according to the training data set. During the training process, the student model is optimized through the student network distillation loss function to obtain an optimized student model. The optimized student model is used for multiple iterations to generate new signal samples, and the generated signal samples are added to the training data set for the training of the ResNet classification model. In this way, by adopting the method of knowledge distillation diffusion, through training the teacher model and the student model, the student model can effectively learn the subtle features of the signals, thereby ensuring the effectiveness and diversity of the signal samples generated by the student model, and thus ensuring the generalization ability of the trained classification model.

[0019] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings:

[0021] Figure 1 shows a schematic flow chart of a radio frequency fingerprint data enhancement method based on distillation diffusion according to an embodiment of the present application;

[0022] Figure 2 shows a schematic flow chart of model training of a radio frequency fingerprint data enhancement method based on distillation diffusion according to an embodiment of the present application;

[0023] Figure 3 shows a block diagram of a radio frequency fingerprint data enhancement device based on distillation diffusion according to an embodiment of the present application;

[0024] Figure 4 shows a schematic structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present application. DETAILED DESCRIPTION

[0025] Now, the exemplary embodiments will be described more fully with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present application will be more complete and comprehensive, and the concept of the exemplary embodiments will be fully conveyed to those skilled in the art.

[0026] In addition, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of the embodiments of the present application. However, those skilled in the art will realize that the technical solutions of the present application can be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, well-known methods, devices, implementations, or operations are not shown or described in detail to avoid obscuring aspects of the present application.

[0027] The block diagrams shown in the drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0028] The flowcharts shown in the drawings are merely illustrative and do not necessarily include all the content and operations / steps, nor are they necessarily executed in the described order. For example, some operations / steps may be decomposed, while some operations / steps may be combined or partially combined, so the actual execution order may change according to the actual situation.

[0029] Figure 1 A schematic flow diagram of a radio frequency fingerprint data enhancement method based on distillation diffusion according to an embodiment of the present application is shown.

[0030] This method can be applied to a terminal device or a server. Among them, the terminal device may include, but is not limited to, one or more of a smart phone, a tablet computer, a portable computer, and a desktop computer, and the server may be a physical server or a cloud server.

[0031] As Figure 1 shown, the radio frequency fingerprint data enhancement method based on distillation diffusion at least includes steps S110 to S140, which are introduced in detail as follows (hereinafter, this method is described by taking its application to a terminal device as an example, hereinafter referred to as "terminal" for short):

[0032] In step S110, the pre-collected radio frequency signals are preprocessed to obtain a training data set.

[0033] In this embodiment, the terminal can pre-acquire radio frequency signals corresponding to multiple wireless devices and preprocess each radio frequency signal to obtain a training data set for subsequent model training.

[0034] In an embodiment of the present application, preprocessing the pre-collected radio frequency signals to obtain a training data set includes:

[0035] Collect a set of radio frequency signal data from different LoRa wireless devices;

[0036] Divide the set of radio frequency signal data into labeled signal samples and unlabeled signal samples, and intercept the effective signal length to generate a training data set.

[0037] In this embodiment, the terminal can pre-acquire a set of radio frequency signal data from different LoRa devices. This set of radio frequency signal data can include several radio frequency signals. It should be understood that LoRa (Long Range) is a long-distance wireless communication technology and belongs to a type of spread spectrum technology. It expands the signal in the frequency domain through spread spectrum technology, thereby achieving long-distance transmission and high-capacity data communication.

[0038] After obtaining the set of radio frequency signal data, the terminal can divide each radio frequency signal into labeled signal samples and unlabeled signal samples, and also intercept the effective signal length for each radio frequency signal, thereby obtaining a training data set.

[0039] In one example, based on the radio frequency signal, 5000 sampling points in a signal sample with 14400 sampling points can be intercepted as the effective signal length. Training samples with this signal length have a relatively high degree of distinguishability, which can not only improve the recognition accuracy of the classification model after training, but also save a certain amount of training evaluation time.

[0040] In step S120, train the pre-constructed teacher model according to the training data set to obtain an optimized teacher model.

[0041] In this embodiment, those skilled in the art can pre-construct a radio frequency fingerprint distillation diffusion model (RF distillation diffusion model, RFDD). This radio frequency fingerprint distillation diffusion model includes a teacher model, a student model, and a ResNet classification network module. The terminal can first train the teacher model according to the training data set to obtain an optimized teacher model.

[0042] In some embodiments of the present application, the teacher model includes a forward diffusion model, a reverse generation model, and a UNet noise network module;

[0043] Training the pre-constructed teacher model according to the training data set to obtain an optimized teacher model includes:

[0044] Randomly extract signal samples from the training data set and add noise to them through the forward diffusion model;

[0045] Input the noisy signal samples into the UNet noise network module to obtain denoising predictions;

[0046] Calculate the MSE loss based on the denoising predictions and the original signal samples, and construct a teacher network loss function; optimize the teacher model according to the teacher network loss function to obtain an optimized teacher model.

[0047] In this embodiment, the forward diffusion model is shown as the following formula (1):

[0048]

[0049] where the latent variable at time step t is h t , is the cumulative noise intensity, is the cumulative signal intensity, is the Gaussian model distribution, x is the signal sample, and Ι is the unit vector.

[0050] The reverse generation model is shown as the following formula (2):

[0051]

[0052] where h s is the latent variable at the s-th time step, h t is the latent variable at the t-th time step, and the classifier-free guidance diffusion score As shown in formula (3), predict the noise at a given time step As shown in formula (4), the variance hyperparameter σ related to the constant η t As shown in formula (5):

[0053]

[0054] where c ∈ [c min, c max is the conditional guidance strength, c min, c max are the minimum and maximum values of the value range of c respectively, is the diffusion score obtained by inputting the vectors of [h t , y] and into the network respectively, y is the sample label, and the learnable parameter is the unified pseudo-label of the unlabeled signal.

[0055]

[0056] When η = 1, the sampling process is random sampling, and when η = 0, the sampling process is deterministic sampling.

[0057] Thus, referring to Figure 2When training the teacher model, the preprocessed training dataset contains labeled signal samples and unlabeled signal samples. Signal samples are randomly selected from them, random time steps are sampled, and the noisy signal samples are obtained through the forward diffusion model (i.e., Equation 1). Then, they are input into the UNet noise network module to obtain the corresponding denoising prediction. Next, the MES loss is calculated based on the denoising prediction and the original signal sample (i.e., the signal sample before adding noise), thereby constructing the teacher network loss function.

[0058] In one example, the teacher network loss function is constructed according to the following formula:

[0059]

[0060] where \(E\) is the mathematical expectation, is the denoising prediction estimated by the UNet noise network module, and \(x\) is the original signal sample.

[0061] Then, the terminal optimizes the teacher model according to this teacher network loss function to obtain the optimized teacher model for guiding the training of the student model in the subsequent steps.

[0062] Please continue to refer to Figure 1 In step S130, the pre - constructed student model is trained according to the training dataset, and during the training process, the student model is optimized through the student network distillation loss function to obtain the optimized student model.

[0063] In this embodiment, after the training of the teacher model is completed, the terminal can train the student model according to the training dataset, and during the training process, the student model is optimized through the student network distillation loss function. The student network distillation loss function can be used to measure the difference between the latent variables generated by the student model and the latent variables generated by the teacher model. By minimizing this student network distillation loss function, the student model can learn the sampling knowledge of the teacher model, reduce the sampling steps, and at the same time maintain the high quality of the generated signal samples.

[0064] In some embodiments of the present application, when training the student model, it includes:

[0065] Sampling random distillation time steps, and at the beginning of each distillation step, initializing the student model with the same parameters as the optimized teacher model;

[0066] Performing distillation with the sampling time steps from large to small, and halving the sampling time each time for guiding the learning of the student model;

[0067] For any signal sample in the training dataset, given a sampling time step, sample a set of sample distillation time steps from a uniform distribution, which includes the starting point, intermediate point, and ending point of sample distillation;

[0068] Use the reverse generation model of the optimized teacher model to perform two-step deterministic sampling to obtain the latent variables from the starting point to the intermediate point of sample distillation, and the latent variables from the intermediate point to the ending point;

[0069] Train the student model to learn the output of the optimized teacher model in two-step deterministic sampling, and optimize the student model based on the student network distillation loss function.

[0070] In this embodiment, as Figure 2 shown, when training the student model, sample random distillation time steps. At the beginning of each distillation step, the student model is initialized with the same parameters as the teacher model. The sampling time steps are distilled from large to small, and each time the distillation halves the sampling time step to guide the student model to learn. For any training data x ∼ p(x), given a sampling time step K ∈ {1, 2,..., T}, sample g ∼ U{1, 2,..., K} from a uniform distribution to obtain a set of sample distillation time steps τ = {t1, t2, t3}, where t1 = gT / K is the starting point of sample distillation, t2 = (g - 0.5)T / K is the intermediate point of sample distillation, and t3 = (g - 1)T / K is the ending point of sample distillation.

[0071] Execute a two-step distillation process based on the set of sample distillation time steps τ. First, obtain the latent variable at time step t1 through the forward diffusion process of Equation (1) Subsequently, train the student model to learn the output of the teacher model in two-step deterministic sampling. Specifically, the teacher model performs two-step deterministic sampling according to Equation (2) and to obtain the latent variable from the starting point t1 to the intermediate point t2 of sample distillation and the latent variable from the intermediate point t2 to the ending point t3 The student model estimates the latent variable obtained by one-step sampling from the starting point t1 to the ending point t3 through the network Set the denoising prediction to be equal to Combined with Equation (2), the denoising prediction of the student model can be obtained as shown in the following equation:

[0072]

[0073] where is the cumulative signal intensity corresponding to time step t3, is the cumulative signal intensity corresponding to time step t1,​​ is the cumulative noise intensity corresponding to time step t3, is the cumulative noise intensity corresponding to time step t1.

[0074] Finally, the distillation loss function of the student network is constructed by the following formula to optimize the student model, so as to obtain the optimized student model:

[0075]

[0076] where E is the mathematical expectation, is the denoising prediction estimated by UNet, is the distillation target that the network needs to fit.

[0077] Please continue to refer to Figure 1 , in step S140, the optimized student model is iterated multiple times to generate new signal samples, and the generated signal samples are added to the training dataset for training the ResNet classification model.

[0078] In this embodiment, the RF fingerprint distillation diffusion model further includes a ResNet classification model. After the training of the student model is completed, the terminal can use the optimized student model to iterate multiple times to generate new signal samples and add them to the training dataset to participate in the training of the ResNet classification model, improving the training effect of the ResNet classification model.

[0079] In some embodiments of the present application, when training the ResNet classification model, the classification error of the model is calculated through the cross-entropy loss function to optimize the ResNet classification model.

[0080] In some embodiments of the present application, the method further includes:

[0081] Obtain the RF signal to be recognized;

[0082] Input the RF signal to be recognized into the optimized ResNet classification model so that the ResNet classification model outputs the corresponding classification result.

[0083] In this embodiment, when RF signal recognition is required, the terminal can input the obtained RF signal to be recognized into the optimized ResNet classification model so that the ResNet classification model outputs the corresponding classification result.

[0084] Thus, according to Figure 1In the illustrated embodiment, a training data set is obtained by preprocessing the pre-collected radio frequency signals; the pre-constructed teacher model is trained according to the training data set to obtain an optimized teacher model, and then the pre-constructed student model is trained according to the training data set. During the training process, the student model is optimized through the student network distillation loss function to obtain an optimized student model. The optimized student model is used for multiple iterations to generate new signal samples, and the generated signal samples are added to the training data set for the training of the ResNet classification model. In this way, by adopting the method of knowledge distillation diffusion, through training the teacher model and the student model, the student model can effectively learn the subtle features of the signals, thereby ensuring the effectiveness and diversity of the signal samples generated by the student model, and thus ensuring the generalization ability of the trained classification model.

[0085] The following describes the apparatus embodiments of the present application, which can be used to execute the radio frequency fingerprint data enhancement method based on distillation diffusion in the above embodiments of the present application. For the details not disclosed in the apparatus embodiments of the present application, please refer to the embodiments of the radio frequency fingerprint data enhancement method based on distillation diffusion in the above of the present application.

[0086] Figure 3 The block diagram of a radio frequency fingerprint data enhancement apparatus based on distillation diffusion according to an embodiment of the present application is shown.

[0087] Refer to Figure 3 As shown, a radio frequency fingerprint data enhancement apparatus based on distillation diffusion according to an embodiment of the present application includes:

[0088] A preprocessing module, configured to preprocess the pre-collected radio frequency signals to obtain a training data set;

[0089] A first optimization module, configured to train the pre-constructed teacher model according to the training data set to obtain an optimized teacher model;

[0090] A second optimization module, configured to train the pre-constructed student model according to the training data set, and during the training process, optimize the student model through the student network distillation loss function to obtain an optimized student model;

[0091] A processing module, configured to use the optimized student model for multiple iterations to generate new signal samples, and add the generated signal samples to the training data set for the training of the ResNet classification model.

[0092] In some embodiments of the present application, the teacher model includes a forward diffusion model and a UNet noise network module;

[0093] Training the pre-constructed teacher model according to the training data set to obtain an optimized teacher model, including:

[0094] Randomly extracting signal samples from the training data set and adding noise to them through the forward diffusion model;

[0095] Inputting the noise-added signal samples into the UNet noise network module to obtain denoising predictions;

[0096] Calculating the MSE loss based on the denoising prediction and the original signal samples, and constructing a teacher network loss function; optimizing the teacher model according to the teacher network loss function to obtain an optimized teacher model.

[0097] In some embodiments of the present application, the teacher model further includes a reverse diffusion model;

[0098] Then, when training the student model, it includes:

[0099] Sampling random distillation time steps, and at the beginning of each distillation step, initializing the student model with the same parameters as the optimized teacher model;

[0100] Performing distillation by sampling time steps from large to small, and halving the sampling time each time for guiding the student model to learn;

[0101] For any signal sample in the training data set, given the sampling time step, uniformly sampling to obtain a sample distillation time step set, which includes the starting point, intermediate point, and ending point of sample distillation;

[0102] Using the reverse generation model of the optimized teacher model to perform two-step deterministic sampling to obtain the latent variables from the starting point to the intermediate point and from the intermediate point to the ending point of sample distillation;

[0103] Training the student model to learn the output of the optimized teacher model in two-step deterministic sampling, and optimizing the student model based on the student network distillation loss function.

[0104] In some embodiments of the present application, when training the ResNet classification model, calculating the model classification error through the cross-entropy loss function to optimize the ResNet classification model.

[0105] In some embodiments of the present application, the processing module is further configured to:

[0106] Obtain the radio frequency signal to be recognized;

[0107] Inputting the radio frequency signal to be recognized into the optimized ResNet classification model so that the ResNet classification model outputs the corresponding classification result.

[0108] In some embodiments of the present application, pre-collected radio frequency signals are preprocessed to obtain a training data set, including:

[0109] Collect radio frequency signal data sets from different LoRa wireless devices;

[0110] Divide the radio frequency signal data set into labeled signal samples and unlabeled signal samples, and intercept the effective signal length to generate a training data set.

[0111] Figure 4 The structural diagram of a computer system of an electronic device suitable for implementing the embodiments of the present application is shown.

[0112] It should be noted that Figure 4 The computer system of the shown electronic device is only an example and should not bring any restrictions to the functions and usage scope of the embodiments of the present application.

[0113] As Figure 4 shown, the computer system includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage section 408 into the random access memory (RAM) 403, such as executing the method described in the above embodiments. In the RAM 403, various programs and data required for system operation are also stored. The CPU 401, ROM 402, and RAM 403 are connected to each other through a bus 404. The input / output (I / O) interface 405 is also connected to the bus 404.

[0114] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including, for example, a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that the computer program read from it can be installed into the storage section 408 as needed.

[0115] In particular, according to an embodiment of the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium, and the computer program contains a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication section 409, and / or installed from the removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the system of the present application are executed.

[0116] It should be noted that the computer-readable medium shown in the embodiments of the present application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, apparatus, or device. In the present application, the computer-readable signal medium can include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable computer program is carried. Such a propagated data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, and the computer-readable medium can send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0117] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. Among them, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the above-mentioned module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, as well as the combination of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0118] The units involved in the embodiments described in the present application can be implemented in software or in hardware, and the described units can also be provided in a processor. Among them, the names of these units do not, in some cases, constitute a limitation on the unit itself.

[0119] On the other hand, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or may exist separately without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the methods described in the above embodiments.

[0120] It should be noted that although several modules or units of the devices for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of the two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0121] Through the description of the above embodiments, those skilled in the art can easily understand that the example embodiments described here can be implemented by software or by a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (such as a personal computer, a server, a touch terminal, or a network device, etc.) to execute the methods according to the embodiments of the present application.

[0122] Those skilled in the art will readily conceive of other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common general knowledge or conventional technical means in the technical field not disclosed in the present application.

[0123] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A radio frequency fingerprint data enhancement method based on distillation diffusion, characterized in that, Including: Preprocessing the pre-collected radio frequency signals to obtain a training data set; Training a pre-constructed teacher model according to the training data set to obtain an optimized teacher model; Training a pre-constructed student model according to the training data set, and optimizing the student model through a student network distillation loss function during the training process to obtain an optimized student model; Using the optimized student model to iterate multiple times to generate new signal samples, and adding the generated signal samples to the training data set for training the ResNet classification model.

2. The method according to claim 1, wherein The teacher model includes a forward diffusion model and a UNet noise network module; Training a pre-constructed teacher model according to the training data set to obtain an optimized teacher model, including: Randomly extracting signal samples from the training data set and adding noise to them through the forward diffusion model; Inputting the noise-added signal samples into the UNet noise network module to obtain a denoising prediction; Calculating the MSE loss according to the denoising prediction and the original signal samples, and constructing a teacher network loss function; optimizing the teacher model according to the teacher network loss function to obtain an optimized teacher model.

3. The method according to claim 1, characterized in that The teacher model further includes a reverse diffusion model; Then, when training the student model, it includes: Sampling random distillation time steps, and initializing the student model with the same parameters as the optimized teacher model at the beginning of each distillation step; Sampling time steps for distillation from large to small, and halving the sampling time each time for guiding the student model to learn; For any signal sample in the training data set, given a sampling time step, uniformly sampling to obtain a sample distillation time step set, which includes the starting point, middle point, and ending point of sample distillation; Using the reverse generation model of the optimized teacher model to perform two-step deterministic sampling to obtain the latent variables from the starting point to the middle point and from the middle point to the ending point of sample distillation; Training the student model to learn the output of the optimized teacher model in two-step deterministic sampling, and optimizing the student model based on the student network distillation loss function.

4. The method according to claim 1, wherein When training the ResNet classification model, calculating the model classification error through the cross-entropy loss function to optimize the ResNet classification model.

5. The method according to claim 4, wherein The method further includes: Obtaining a radio frequency signal to be recognized; Inputting the radio frequency signal to be recognized into the optimized ResNet classification model so that the ResNet classification model outputs a corresponding classification result.

6. The method according to any one of claims 1-5, characterized in that, Preprocessing the pre-collected radio frequency signals to obtain a training data set, including: Collecting a radio frequency signal data set from different LoRa wireless devices; Dividing the radio frequency signal data set into labeled signal samples and unlabeled signal samples, and intercepting the effective signal length to generate a training data set.

7. A radio frequency fingerprint data enhancement device based on distillation diffusion, characterized in that Including: A preprocessing module for preprocessing the pre-collected radio frequency signals to obtain a training data set; A first optimization module for training a pre-constructed teacher model according to the training data set to obtain an optimized teacher model; A second optimization module, configured to train a pre-constructed student model according to the training data set, and optimize the student model through a student network distillation loss function during the training process to obtain an optimized student model; A processing module, configured to use the optimized student model to iterate multiple times to generate new signal samples, and add the generated signal samples to the training data set for training a ResNet classification model.

8. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for enhancing radio frequency fingerprint data based on distillation diffusion according to any one of claims 1 to 6.

9. An electronic device, characterized in that, Comprising: One or more processors; A storage device, configured to store one or more programs, when the one or more programs are executed by the one or more processors, enabling the one or more processors to implement the method for enhancing radio frequency fingerprint data based on distillation diffusion according to any one of claims 1 to 6.

Images