Radio frequency fingerprinting method and system based on adaptive wavelet threshold denoising algorithm

Abstract

The present application belongs to the technical field of radio frequency fingerprint identification, and discloses a radio frequency fingerprint identification method and system based on an adaptive wavelet threshold denoising algorithm.The method comprises the following steps: receiving a radio frequency signal; training the received radio frequency signal, including: performing standardization processing on the received radio frequency signal; extracting a wavelet base function and a wavelet threshold matched with the radio frequency signal; performing denoising processing on the radio frequency signal after standardization processing based on the wavelet base function and the wavelet threshold; adding the radio frequency signal before denoising processing and the radio frequency signal after denoising processing to serve as the input of a classification network to obtain the prediction probability of each category to which the radio frequency signal belongs; updating the network parameters according to the prediction probability of each category to which the radio frequency signal belongs; and repeating the training of the received radio frequency signal until the network converges.Thus, on the basis of solving the time-consuming problem of manually selecting a wavelet base in a general wavelet threshold denoising algorithm, the performance of a subsequent identification network in a low signal-to-noise ratio environment is improved.

Description

Technical Field

[0001] This invention belongs to the field of radio frequency fingerprint recognition technology, and particularly relates to a radio frequency fingerprint recognition method and system based on an adaptive wavelet threshold denoising algorithm. Technical Background

[0002] With the development of wireless communication technology and the Internet of Things (IoT), the number of IoT devices in wireless communication systems has surged. A large amount of data is transmitted between interconnected IoT devices, making the identification of legitimate and illegitimate IoT devices in the network a pressing problem.

[0003] Radio Frequency Fingerprint (RFF) is often used as a physical layer authentication technology for devices accessing a network due to its difficulty in being copied and forged. An RRF fingerprint refers to a unique "fingerprint" formed when imperfections in the manufacturing process lead to inherent defects in the components used for wireless signal transmission. These defects result in slight differences in parameters such as frequency and phase in the radio frequency signals emitted by different devices (even those from the same batch and model).

[0004] However, most existing RFID fingerprinting methods suffer from performance degradation in low signal-to-noise ratio (SNR) environments. Therefore, researchers have begun to denoise RFID signals before identification. Wavelet transform can extract time-frequency domain features of signals, and the wavelet threshold denoising (WTD) algorithm derived from it is widely used in the field of noise reduction. However, current WTD algorithms require an extremely time-consuming wavelet basis selection process, and there is a problem of poor denoising results due to mismatch between existing wavelet bases and signals. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a radio frequency fingerprint recognition method and system based on an adaptive wavelet threshold denoising algorithm. This method can adaptively select the wavelet basis, thereby improving the recognition performance of the algorithm in low signal-to-noise ratio environments without requiring manual selection of the wavelet basis.

[0006] In a first aspect, a radio frequency fingerprinting method based on an adaptive wavelet threshold denoising algorithm is provided, the method comprising the following steps:

[0007] Receive radio frequency signals;

[0008] Training the received radio frequency signal includes:

[0009] The received radio frequency signals are standardized.

[0010] Extract wavelet basis functions and wavelet thresholds that match the radio frequency signal;

[0011] Denoising of standardized radio frequency signals is performed based on wavelet basis functions and wavelet thresholds.

[0012] The radio frequency signals before and after denoising are added together and used as the input to the classification network to obtain the predicted probability of each category to which the radio frequency signal belongs.

[0013] The network parameters are updated based on the predicted probability of the radio frequency signal's category.

[0014] Repeat the training process on the received radio frequency signals until the network converges;

[0015] The optimal network parameters are output for testing.

[0016] Preferably, the standardization processing of the received radio frequency signal specifically includes:

[0017] Assume the received signal is The mean μ and standard deviation σ of the signal are calculated using the following formulas:

[0018]

[0019]

[0020] The received radio frequency signal is standardized to obtain a standardized radio frequency signal.

[0021]

[0022] Preferably, the extraction of wavelet basis functions and wavelet thresholds matching the radio frequency signal specifically includes:

[0023] One-dimensional radio frequency signal sequence A two-dimensional matrix is ​​formed by stacking segments. Data processing is performed using a two-dimensional convolutional network.

[0024]

[0025] A convolutional neural network is used to extract wavelet basis functions and wavelet threshold λ that match the input radio frequency signal sequence; wherein the wavelet basis functions correspond to a pair of filter coefficients h[n],l[n].

[0026] Preferably, the denoising process of the standardized radio frequency signal based on wavelet basis functions and wavelet thresholds specifically includes:

[0027] Using the wavelet basis functions h[n], l[n] to analyze the radio frequency signal sequence sstd Perform wavelet decomposition:

[0028]

[0029]

[0030] Among them, s 1,L [n],s 1,H [n] represents the low-frequency and high-frequency components of the first-order wavelet coefficients, respectively; s 1,L [n] further decomposed:

[0031]

[0032]

[0033] Among them, s β,L [n],s β,H [n] represents the low-frequency and high-frequency components of the β-order wavelet coefficients, respectively;

[0034] Using the threshold λ, the wavelet coefficients are subjected to thresholding, and the thresholding function is:

[0035]

[0036] Where ω is the original wavelet coefficient, ω th These are the wavelet coefficients after thresholding.

[0037] The wavelet coefficients after thresholding are reconstructed using wavelet functions to obtain the denoised signal s′:

[0038]

[0039] In the above formula, the superscript ' is used to distinguish the wavelet coefficients after thresholding from the original wavelet coefficients.

[0040] Preferably, the step of adding the radio frequency signals before and after denoising processing and using the sum as input to the classification network to obtain the predicted probability of each category to which the radio frequency signal belongs specifically includes:

[0041] The radio frequency signal sequence s′ and s std After addition, a two-dimensional matrix is ​​formed by stacking segments, and the data is processed by a two-dimensional convolutional network. The two-dimensional convolutional network selected in this scheme is ResNet18.

[0042] After inputting the summed RF signal sequence into ResNet18, the output is the probability predicted by the network that the summed RF signal belongs to each category. N trans This represents the number of transmitter categories.

[0043] Preferably, updating the network parameters based on the predicted probability of the radio frequency signal's category specifically includes:

[0044] Cross-entropy is chosen as the loss function, and the Adam optimizer is used for parameter updates; the formula for calculating cross-entropy is:

[0045]

[0046] Among them, y i The sign function is set to 1 if the signal belongs to category i, and 0 otherwise.

[0047] Secondly, a radio frequency fingerprinting system based on an adaptive wavelet threshold denoising algorithm is provided, the system comprising the following modules:

[0048] The receiver module is used to receive radio frequency signals;

[0049] The training module, used to train the received radio frequency signal, includes:

[0050] The standardization processing module is used to standardize the received radio frequency signals.

[0051] The matching module is used to extract wavelet basis functions and wavelet thresholds that are matched with the radio frequency signal;

[0052] The noise reduction module is used to perform noise reduction processing on the standardized radio frequency signal based on wavelet basis functions and wavelet thresholds.

[0053] The probability prediction module is used to add the radio frequency signals before and after denoising and use the sum as the input to the classification network to obtain the predicted probability of the radio frequency signal to each category.

[0054] The update module is used to update network parameters based on the predicted probability of the radio frequency signal belonging to a specific category.

[0055] The iterative training module is used to repeatedly train the received radio frequency signals until the network converges.

[0056] The output module is used to output the optimal network parameters for the testing process.

[0057] Thirdly, a computer-readable storage medium is provided for storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform any of the methods described.

[0058] Fourthly, a computing device is provided, comprising:

[0059] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods described.

[0060] The beneficial effects of this invention are as follows:

[0061] 1. By using a convolutional neural network to extract wavelet basis functions and thresholds, the process of manually selecting wavelet basis functions is eliminated;

[0062] 2. The signal after adaptive wavelet threshold denoising processing has a significantly improved recognition accuracy in low signal-to-noise ratio environments;

[0063] 3. Adaptive Wavelet Thresholding Denoising (AWTD) is portable and not limited to ResNet18 used in this invention; it can also be used as a pre-processing module for other recognition networks. Numerical results show that the AWTD module can also improve the performance of other recognition networks in low signal-to-noise ratio environments. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0065] Appendix Figure 1 This is a flowchart of the training process of the radio frequency fingerprint recognition algorithm based on the adaptive wavelet threshold denoising algorithm of the present invention;

[0066] Appendix Figure 2 This is a simplified network structure diagram of the radio frequency fingerprint recognition algorithm based on the adaptive wavelet threshold denoising algorithm of the present invention;

[0067] Appendix Figure 3 This is a network structure diagram of the adaptive wavelet threshold denoising module in this invention. Detailed Implementation

[0068] To make the objectives and technical solutions of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be described more clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0069] Since this invention is based on deep neural networks, the description of specific implementation methods will be separated into two aspects: the training process and the testing process.

[0070] like Figure 1-3 As shown, one embodiment of the present invention provides a radio frequency fingerprint recognition method based on an adaptive wavelet threshold denoising algorithm, comprising the following steps:

[0071] Receive radio frequency signals;

[0072] Training the received radio frequency signal includes:

[0073] The received radio frequency signals are standardized.

[0074] Extract wavelet basis functions and wavelet thresholds that match the radio frequency signal;

[0075] Denoising of standardized radio frequency signals is performed based on wavelet basis functions and wavelet thresholds.

[0076] The radio frequency signals before and after denoising are added together and used as the input to the classification network to obtain the predicted probability of each category to which the radio frequency signal belongs.

[0077] The network parameters are updated based on the predicted probability of the radio frequency signal's category.

[0078] Repeat the training process on the received radio frequency signals until the network converges;

[0079] The optimal network parameters are output for testing.

[0080] Specifically, the input signal is standardized, including the following steps:

[0081] (1.1) Assume the received signal is The mean μ and standard deviation σ of the signal are calculated using the following formulas:

[0082]

[0083]

[0084] (1.2) Standardize the received signal to obtain a standardized signal.

[0085]

[0086] Specifically, the process involves using a convolutional neural network to extract wavelet basis functions and thresholds that match the signal, including the following steps:

[0087] (2.1) A one-dimensional signal sequence A two-dimensional matrix is ​​formed by stacking segments. In order to facilitate subsequent data processing for two-dimensional convolutional networks:

[0088]

[0089] (2.2) Utilizing, for example Figure 3 The convolutional neural network shown extracts wavelet basis functions (corresponding to a pair of filter coefficients h[n], l[n]) and threshold λ that match the input signal sequence.

[0090] Specifically, based on the wavelet threshold denoising algorithm, for signal s std Denoising involves the following steps:

[0091] (3.1) Using the wavelet basis functions h[n], l[n] obtained in step (2.2) to analyze the signal sequence s std Perform wavelet decomposition:

[0092]

[0093]

[0094] Among them, s 1,L [n],s 1,H [n] represents the low-frequency and high-frequency components of the first-order wavelet coefficients, respectively. 1,L [n] can be further decomposed:

[0095]

[0096]

[0097] Among them, s β,L [n],s β,H [n] represents the low-frequency and high-frequency components of the β-order wavelet coefficients, respectively.

[0098] (3.2) Using the threshold λ obtained in step (2.2), the wavelet coefficients obtained in (3.1) are subjected to thresholding to reduce the noise component in the received signal. The thresholding function is:

[0099]

[0100] Where ω is the original wavelet coefficient, ω th These are the wavelet coefficients after thresholding.

[0101] (3.3) Perform wavelet reconstruction on the wavelet coefficients after thresholding to obtain the denoised signal s′:

[0102]

[0103] In the above formula, the superscript ' is used to distinguish the wavelet coefficients after thresholding from the original wavelet coefficients.

[0104] Specifically, the residual neural network ResNet18 is used to classify the wavelet-denoised signal, including the following steps:

[0105] (4.1) Combine the signal sequence s′ with s std After addition, the process in step (2.1) is performed, and the result is used as input to the ResNet18 network.

[0106] (4.2) After inputting the signal sequence into ResNet18, the output network predicts the probability that the signal belongs to each category. N trans Number of transmitter categories

[0107] Specifically, the network parameters are updated. Cross-entropy is chosen as the loss function, and the Adam optimizer is used for parameter updates. The formula for calculating cross-entropy is:

[0108]

[0109] Among them, y i The sign function is set to 1 if the signal belongs to category i, and 0 otherwise.

[0110] During the training phase, the relevant parameters of the network are determined, and the testing process is the same as the training process (because the network parameters are determined during training, so parameter updates are not required). The following modifications are made to "update network parameters according to the predicted probability of the radio frequency signal category":

[0111] Step 5: Output the predicted category of the signal.

[0112] The category with the highest predicted probability is used as the category predicted by the network.

[0113]

[0114] Another embodiment of the present invention provides a radio frequency fingerprinting system based on an adaptive wavelet threshold denoising algorithm, the system comprising the following modules:

[0115] The receiver module is used to receive radio frequency signals;

[0116] The training module, used to train the received radio frequency signal, includes:

[0117] The standardization processing module is used to standardize the received radio frequency signals.

[0118] The matching module is used to extract wavelet basis functions and wavelet thresholds that are matched with the radio frequency signal;

[0119] The noise reduction module is used to perform noise reduction processing on the standardized radio frequency signal based on wavelet basis functions and wavelet thresholds.

[0120] The probability prediction module is used to add the radio frequency signals before and after denoising and use the sum as the input to the classification network to obtain the predicted probability of the radio frequency signal to each category.

[0121] The update module is used to update network parameters based on the predicted probability of the radio frequency signal belonging to a specific category.

[0122] The iterative training module is used to repeatedly train the received radio frequency signals until the network converges.

[0123] The output module is used to output the optimal network parameters for the testing process.

[0124] Embodiments of this application may be provided as methods or computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of this application may be implemented in various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0125] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0126] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0128] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A radio frequency fingerprint recognition method based on an adaptive wavelet threshold denoising algorithm, characterized in that, The method includes the following steps: Receive radio frequency signals; Training the received radio frequency signal includes: The received radio frequency signals are standardized. Extracting the wavelet basis function and wavelet threshold that match the radio frequency signal, specifically including: One-dimensional radio frequency signal sequence A two-dimensional matrix is ​​formed by stacking segments. Data processing is performed using a two-dimensional convolutional network. Wavelet basis functions and wavelet thresholds that match the input RF signal sequence are extracted using a convolutional neural network. Wherein, the wavelet basis function corresponds to a pair of filter coefficients. ; Denoising of standardized radio frequency signals is performed based on wavelet basis functions and wavelet thresholds, specifically including: Using the wavelet basis function For standardized radio frequency signal sequences Perform wavelet decomposition: in, These are the low-frequency and high-frequency components of the first-order wavelet coefficients, respectively; Further breakdown: in, They are The low-frequency and high-frequency components of the wavelet coefficients. Using the threshold Thresholding is applied to the wavelet coefficients. The thresholding function is as follows: in, These are the original wavelet coefficients. These are the wavelet coefficients after thresholding. The wavelet coefficients after thresholding are reconstructed using wavelet convolution to obtain the denoised signal. : In the above formula, the superscript ' is used to distinguish the wavelet coefficients after thresholding from the original wavelet coefficients; The radio frequency signals before and after denoising are added together and used as the input to the classification network to obtain the predicted probability of each category to which the radio frequency signal belongs. The network parameters are updated based on the predicted probability of the radio frequency signal's category. Repeat the training process on the received radio frequency signals until the network converges; The optimal network parameters are output for testing.

2. The radio frequency fingerprint recognition method based on adaptive wavelet threshold denoising algorithm according to claim 1, characterized in that, The standardization process for the received radio frequency signal specifically includes: Assume the received signal is The mean of the signal is calculated using the following formula. and standard deviation : The received radio frequency signal is standardized to obtain a standardized radio frequency signal. : 。 3. The radio frequency fingerprint recognition method based on adaptive wavelet threshold denoising algorithm according to claim 1, characterized in that, The step of adding the radio frequency signals before and after denoising processing and using the sum as input to the classification network to obtain the predicted probability of each category to which the radio frequency signal belongs specifically includes: radio frequency signal sequence and After addition, a two-dimensional matrix is ​​formed by stacking segments, and the data is processed by a two-dimensional convolutional network. The two-dimensional convolutional network selected in this scheme is ResNet18. After the summed RF signal sequences are input into ResNet18, the output network predicts the probability that the original RF signals belong to each category. , This represents the number of transmitter categories.

4. The radio frequency fingerprint recognition method based on adaptive wavelet threshold denoising algorithm according to claim 1, characterized in that, The step of updating network parameters based on the predicted probability of the radio frequency signal's category specifically includes: Cross-entropy is chosen as the loss function, and the Adam optimizer is used for parameter updates; the formula for calculating cross-entropy is: in, The sign function is used if the signal belongs to a category. If the result is 1, then take 1; otherwise, take 0.

5. A system for implementing the radio frequency fingerprint recognition method based on the adaptive wavelet threshold denoising algorithm as described in any one of claims 1-4, characterized in that, The system includes the following modules: The receiver module is used to receive radio frequency signals; The training module, used to train the received radio frequency signal, includes: The standardization processing module is used to standardize the received radio frequency signals. The matching module is used to extract wavelet basis functions and wavelet thresholds that are matched with the radio frequency signal; The noise reduction module is used to perform noise reduction processing on the standardized radio frequency signal based on wavelet basis functions and wavelet thresholds. The probability prediction module is used to add the radio frequency signals before and after denoising and use the sum as the input to the classification network to obtain the predicted probability of the radio frequency signal to each category. The update module is used to update network parameters based on the predicted probability of the radio frequency signal belonging to a specific category. The iterative training module is used to repeatedly train the received radio frequency signals until the network converges. The output module is used to output the optimal network parameters for the testing process.

6. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform the method according to any one of claims 1-4.

7. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing the method according to any one of claims 1-4.

Images