CN116963074A - Random fence-based dual-branch enhanced radio frequency signal fingerprint identification method and device - Google Patents

Abstract

The application provides a random fence-based dual-branch enhanced radio frequency signal fingerprint identification method and device, and relates to the technical field of communication. The method comprises the steps S1 to S5. S1, acquiring a wireless radio frequency signal containing fingerprint information, and preprocessing. S2, performing enhancement processing on the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals. S3, converting the preprocessed wireless radio frequency signals through short-time Fourier transform to obtain a spectrogram. S4, respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics. S5, combining the time domain features and the frequency domain features through a self-attention mechanism, and acquiring fingerprint features based on the wireless radio frequency signals. The method is more suitable for the radio frequency fingerprint characteristics of the interference environment, and finally improves the recognition rate of the high signal-to-noise ratio signals.

Description

Random fence-based dual-branch enhanced radio frequency signal fingerprint identification method and device

Technical Field

The application relates to the technical field of communication, in particular to a method and a device for fingerprint identification of a dual-branch enhanced radio frequency signal based on a random fence.

Background

Wireless networks have brought subverted changes to society and become an integral part of everyday life. However, the explosive growth of wireless application demands and the number of wireless devices inevitably creates security and privacy concerns. Thus, authenticating the identity of a wireless device is urgent.

The wireless signals of the wireless devices have certain energy loss and interference in the transmission process. Therefore, data enhancement is required for the wireless radio frequency signal to perform fingerprint identification authentication.

At present, methods commonly used in academia at home and abroad for realizing data enhancement effect are classified into three types of data enhancement-based methods, algorithm-based methods and model-based methods. The envelope of the radio frequency signal is easily destroyed by the data enhancement filling method, and the information carried by the radio frequency signal is influenced. Whereas algorithm-based and model-based methods tend to be complex and prone to over-fitting.

In view of the above, the applicant has studied the prior art and has made the present application.

Disclosure of Invention

The application provides a dual-branch enhanced radio frequency signal fingerprint identification method and device based on a random fence, which are used for improving at least one of the technical problems.

A first aspect,

The embodiment of the application provides a dual-branch enhanced radio frequency signal fingerprint identification method based on a random fence, which comprises steps S1 to S5.

S1, acquiring a wireless radio frequency signal containing fingerprint information, and preprocessing.

S2, performing enhancement processing on the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals.

S3, converting the preprocessed wireless radio frequency signals through short-time Fourier transform to obtain a spectrogram.

S4, respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics.

S5, combining the time domain features and the frequency domain features through a self-attention mechanism, and acquiring fingerprint features based on the wireless radio frequency signals.

A second aspect,

The embodiment of the application provides a dual-branch enhanced radio frequency signal fingerprint identification device based on a random fence, which comprises:

and the preprocessing module is used for acquiring the wireless radio frequency signals containing the fingerprint information and preprocessing the wireless radio frequency signals.

The enhancement module is used for enhancing the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals.

The transformation module is used for transforming the preprocessed wireless radio frequency signals through short-time Fourier transformation to obtain a spectrogram.

The extraction module is used for respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics.

And the fusion module is used for combining the time domain features and the frequency domain features through a self-attention mechanism and acquiring fingerprint features based on the wireless radio frequency signals.

By adopting the technical scheme, the application can obtain the following technical effects:

according to the random fence-based dual-branch enhanced radio frequency signal fingerprint identification method, the random fence increases the diversity of input samples, the BBSEN symmetrical bilateral branch characteristic enhancement network respectively carries out characteristic extraction on the enhanced time domain signals and spectrograms obtained by short-time Fourier transform of the signals, then characteristic fusion is carried out to obtain radio frequency fingerprint characteristics which are more suitable for interference environments, and finally the identification rate of the signals with high signal to noise ratio is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flow chart of a fingerprint identification method.

Fig. 2 is a schematic diagram of a random fence algorithm.

FIG. 3 shows different degrees of retentionA line graph of model accuracy.

FIG. 4 is a different random segment numberA line graph of model accuracy.

Fig. 5 is a network structure diagram of a dual-branch enhanced radio frequency signal fingerprint identification method based on a random fence.

Fig. 6 is a network structure diagram of an improved self-attention mechanism.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. It will be apparent that the described embodiments are only some, but not all, embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

Embodiment 1,

Referring to fig. 1 to 6, a first embodiment of the present application provides a dual-branch enhanced rf signal fingerprint identification method based on a random fence, which can be performed by a dual-branch enhanced rf signal fingerprint identification device (hereinafter referred to as a fingerprint identification device) based on a random fence. In particular, the steps S1 to S5 are implemented by one or more processors in the fingerprint recognition device.

S1, acquiring a wireless radio frequency signal containing fingerprint information, and preprocessing. Preferably, step S1 specifically includes: acquiring a wireless radio frequency signal containing fingerprint information, performing unified noise reduction processing through a singular value decomposition method, synchronizing the starting points of the radio frequency signals through a transient detection method based on phase characteristics, and performing normalization processing through a maximum and minimum normalization method to acquire a preprocessed wireless radio frequency signal.

Specifically, due to the influence of environmental factors, a series of pretreatment works are carried out on the received signals, wherein the pretreatment works mainly comprise unified noise reduction treatment on the collected signals by a singular value decomposition (Singular Value Decomposition, SVD) method; synchronizing the start of the radio frequency signal by using a transient detection method (Transient Detection using Phase Characteristic, TDPC) based on the phase characteristics; the maximum and minimum normalization is used to make the amplitude of the signals collected in different backgrounds in the range of (0, 1).

It is understood that the fingerprint recognition device may be an electronic device with computing performance, such as a portable notebook computer, a desktop computer, a server, a smart phone, or a tablet computer.

S2, performing enhancement processing on the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals. Specifically, a random fence characteristic enhancement algorithm is used for processing signals, so that characteristics and diversity of the signals are increased.

On the basis of the above embodiment, in an alternative embodiment of the present application, step S2 specifically includes steps S21 to S24.

S21, according to the random segment numberDividing the pretreated wireless radio frequency signal into +.>Fragments. Wherein, the liquid crystal display device comprises a liquid crystal display device,wherein->The length of the preprocessed radio frequency signal, </i >>Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment.

S22, respectively toThe segments perform the steps of performing enhancement processing by a random fence feature enhancement algorithm, obtaining an enhanced signal,

s23, according to the preset reservation degree of the original signalAnd acquiring a filling area and an original signal area of the fragment.

S24, carrying out enhancement processing through a random fence characteristic enhancement algorithm according to the filling area and the original signal area. The definition formula of the random fence characteristic enhancement algorithm is as follows:

in the method, in the process of the application,for enhanced signal->Is->No. 5 of the individual fragment>Time domain signal values at each moment,Is a random weight->Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment.

In this embodiment, the random fence is directed to the characteristics of the radio frequency signal (RFF) and incorporates the idea of re-balancing, using a mask that is a set of segments that meet a uniform distribution. In this configuration, the degree of retention of the signal can be controlled by varying the density and length of the fragments. For the part that needs to be filled, the filling value needed by the random fence is determined by the previous time and the next time of the sequence.

FIG. 2 is a schematic diagram of a random fence algorithm, where the random mask used by the random fence is neither a large continuous area nor a "0" value filling method. In this configuration, by judging the density and length of the threshold control mask filling segments, the regularization operation can be performed on the network while retaining important features of the signal. The random fence appropriately adds masks to unbalanced signal data, so that not only is minority class data expanded, but also the effect of data enhancement is achieved.

Specifically, for each sample of each class，/>Representation sample->Is a length of (c). Mask of random fence algorithm (RRE)>Is a set of fragments that meet a uniform distribution. We will add a sample->Uniformly divide into->Fragments, each fragment having the sequence number +.>Wherein->。/>Representing a certain moment of the signal, wherein ∈ ->。/>Indicate->Section, th->Time domain signal values of time instants. Selecting a random pad area length +.>And original signal area length +>Therefore, there are:

the application usesTo represent the degree of retention of the original signal:

for the firstFor a segment signal, the filling area +.>Is defined by the value of the preceding moment +.>And the value of the latter instant ∈ ->And a random weight->And (5) determining. Wherein (1)>The formula is defined as follows:

in the method, in the process of the application,representing the segment length of the signal>Smaller means +.>And->The more the value of (2)Large, the longer the continuous portion that needs to be randomly filled, resulting in more information being lost.

During training, the application hopes to increase the diversity of samples by adding more randomness to the samples, but also does not hope to make continuous parts of random filling too long to cause information loss, so the application randomly selects from a range meeting uniform distribution:

In the method, in the process of the application,is->Lower limit of the value of->Is->An upper limit of the value of (2);

as can be seen from the above-mentioned theory,and->From which a unique mask may be determined. The effect of the random barrier under different super parameters will be verified in the experimental part, thereby deciding +.>And->Is a solution to the optimization of (3).

FIGS. 3 and 4 are different viewsValue sum->Values, the obtained accuracy curves were obtained by performing experiments on signals of ship wireless authentication (Automatic Identification System, AIS) equipment, customer premises (Customer Premise Equipment, CPE) equipment, internet of things (Internet of Things, ioT) equipment, mobile Phone (MP) equipment, wireless interphone (Wireless Intercom, WI) equipment, and high frequency (Very High Frequency, VHF) wireless equipment, respectively.

As shown in FIG. 4, a device is providedIs the total length of the signal. When->The range is concentrated in the larger section +.>When the recognition accuracy reaches the highest; while->The interval is smaller than->In this case, the recognition accuracy is greatly lowered. Because of->Smaller means that the original signal will have too long a continuous area covered, resulting in a corrupted signal structure. In combination with the experimental results of the inventors, in the examples of the present application,/->、/>。

Figures 3 and 4 illustrate the use of differentValue sum->The values train the model and observe the effect of the random fence on the model performance in order to select the most appropriate parameters. When->Smaller means less retention of the original signal and therefore may cause a lack of fit resulting in a significant drop in recognition accuracy. When->When the accuracy increases gradually, the accuracy also increases gradually, and at +.>When the maximum value is reached, the subsequent accuracy will follow +.>Is decreased by an increase in (a). />The reason for the reduced accuracy at larger values may be that the deleted information is too small and too many overlapping samples are generated resulting in an overfitting.

The degree of retention of the signal can be controlled by varying the density and length of the segments using the method of increasing the mask of the random fence. For the part needing to be filled, the filling value needed by the random fence is determined by the previous moment and the next moment of the sequence, so that the signal only presents a small-amplitude distortion phenomenon on part of the fragments, and the target area is not completely deleted like the traditional mask adding mode, the information such as the signal envelope is destroyed, and the rest information cannot be subjected to downstream tasks. The random fence enhances the characteristics of the input signals, so that the model learns the radio frequency fingerprints with the more distinguishing degree to improve the recognition rate in the interference environment.

S3, converting the preprocessed wireless radio frequency signals through short-time Fourier transform to obtain a spectrogram. Preferably, step S3 specifically includes: and transforming the preprocessed wireless radio frequency signal by selecting short-time Fourier transform of a Kaiser window function, and obtaining a spectrogram of the signal.

Specifically, a Kaiser window function which has smaller side lobe, narrower main lobe width and finer analysis on amplitude in a frequency domain and a time domain is selected by utilizing short-time Fourier transform, and a time domain signal processed by a random fence is converted into a two-dimensional signal spectrogram to be used as the input of a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network.

S4, respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics. Preferably, step S4 specifically includes steps S41 to S42.

S41, inputting the enhanced signal into an upper branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to acquire a time domain characteristic.

S42, inputting the spectrogram into a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network to acquire frequency domain characteristics.

The convolution layers of the upper branch network and the lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network adopt a weight sharing connection mode.

Specifically, as shown in fig. 5, the input of the upper branch network of the BBSEN symmetrical bilateral branch feature enhancement network is that an enhancement signal is obtained after random fence enhancement, and the branches mainly complete time domain feature extraction; the input of the lower branch network is a spectrogram obtained by short-time Fourier transform of the signal, and the frequency domain feature extraction is mainly completed by the branches. In order to avoid the rapid increase of model parameters caused by a double-branch network and ensure that two branches better perform information interaction during training, the BBSEN symmetrical double-side branch characteristic enhancement network adopts a weight sharing connection mode in a convolution layer.

Compared with the traditional deep learning algorithm which simply uses a time domain signal or a spectrogram as input, the dual-branch network of the dual-branch signal data enhancement network (BBSEN) provided by the application respectively performs characteristic extraction on the time domain signal and the spectrogram, and performs information fusion on a final output layer by using a self-attention mechanism. The dual-branch neural network is widely applied to various tasks, can be used for data interfered by factors such as noise and the like, and can also be used for high-quality data which is not interfered, so that the recognition rate of a model is improved.

S5, combining the time domain features and the frequency domain features through a self-attention mechanism to acquire fingerprint features based on the wireless radio frequency signals. Preferably, the self-attention mechanism comprises a time domain signal self-attention module and a spectrogram self-attention module. Step S5 specifically includes steps S51 to S53.

S51, respectively characterizing time domains through two full connection layersAnd frequency domain features->Input time domain signal self-attention module and spectrogram self-attention module, obtain time domain characteristic +.>And frequency domain features->And (5) a weight vector.

S52, the time domain featuresAnd frequency domain features->The weight vectors are concatenated and fed to a two-dimensional convolution layer and then passed through the activation function +.>And +.>After which the adaptive weights are finally generated>。

S53, according to the self-adaptive weightFusing the temporal features->And frequency domain features->And acquiring fingerprint characteristics based on the wireless radio frequency signals. Wherein, the fusion formula is:

in the method, in the process of the application,is a classifier->Representing transpose, < >>Is self-adaptive weight->Time domain features, < >>Frequency domain characteristics.

Specifically, after the feature extraction of the two branches is completed, the feature fusion is completed through a self-attention mechanism (merging branches) (namely, the self-attention mechanism is adopted to acquire the local features of the radio frequency signals). The combining branch is mainly composed of two multipliers and a self-attention module.

The application improves the self-attention module and extends the self-attention module to a time domain signal self-attention module and a spectrogram self-attention module. Fig. 6 is a schematic diagram of an improved self-attention mechanism.

Concrete embodimentsIn terms of the characteristics of the two paths, the two paths are respectivelyAnd->The application realizes a cumulative learning strategy through a self-attention mechanism, so that the upper branch and the lower branch divert 'attention' in a training stage, namely, an adaptive parameter +.>To control the characteristics of the upper branch and the lower branch>And->And takes the weighted feature vector as a fusion feature as a classifier +.>Is input to the computer. The output logic formula (i.e., fusion formula) is as follows:

according to the embodiment of the application, the self-attention module is expanded to the time domain signal self-attention module and the spectrogram self-attention module. Namely: features derived from convolutional encoders using two fully concatenated layersAnd->After being respectively input into the two modules, the self-adaptive balanced signal characteristics are contributed to the time domain and the frequency domain, the obtained weight vectors are connected in series and fed into a two-dimensional convolution layer, and the two-dimensional convolution layer is subjected to an activation function +.>And +.>After which the adaptive weights are finally generated>。

According to the random fence-based dual-branch enhanced radio frequency signal fingerprint identification method, the random fence increases the diversity of input samples, the BBSEN symmetrical bilateral branch characteristic enhancement network respectively carries out characteristic extraction on the enhanced time domain signals and spectrograms obtained by short-time Fourier transform of the signals, then characteristic fusion is carried out to obtain radio frequency fingerprint characteristics which are more suitable for interference environments, and finally the identification rate of the signals with high signal to noise ratio is improved.

The conventional wireless device identification based on the physical layer radio frequency fingerprint often faces the problem of being influenced by factors such as interference, noise and the like, and the fingerprint identification method can effectively improve the radio frequency fingerprint identification performance of the wireless signal in an interference environment.

Embodiment II,

The embodiment of the application provides a dual-branch enhanced radio frequency signal fingerprint identification device based on a random fence, which comprises:

and the preprocessing module is used for acquiring the wireless radio frequency signals containing the fingerprint information and preprocessing the wireless radio frequency signals.

The enhancement module is used for enhancing the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals.

The transformation module is used for transforming the preprocessed wireless radio frequency signals through short-time Fourier transformation to obtain a spectrogram.

And the extraction module is used for respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics.

And the fusion module is used for merging the time domain features and the frequency domain features through a self-attention mechanism to acquire fingerprint features based on the wireless radio frequency signals.

In an alternative embodiment of the present application based on the above embodiment, the preprocessing module is specifically configured to: acquiring a wireless radio frequency signal containing fingerprint information, performing unified noise reduction processing through a singular value decomposition method, synchronizing the starting points of the radio frequency signals through a transient detection method based on phase characteristics, and performing normalization processing through a maximum and minimum normalization method to acquire a preprocessed wireless radio frequency signal.

In an alternative embodiment of the present application based on the above embodiment, the enhancement module specifically includes:

a dividing unit for dividing the random segment numberDividing the pretreated wireless radio frequency signal into +.>Fragments. Wherein (1)>Wherein->The length of the preprocessed radio frequency signal, </i >>Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment.

Repeating units for respectively toThe segments perform the steps of performing enhancement processing by a random fence feature enhancement algorithm, obtaining an enhanced signal,

a region selection unit for selecting a region based on a predetermined retention level of the original signalAnd acquiring a filling area and an original signal area of the fragment.

And the enhancement processing unit is used for carrying out enhancement processing through a random fence characteristic enhancement algorithm according to the filling area and the original signal area. The definition formula of the random fence characteristic enhancement algorithm is as follows:

in the method, in the process of the application,for enhanced signal->Is->No. 5 of the individual fragment>Time domain signal values at each moment,Is a random weight->Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment.

In an alternative embodiment of the present application based on the above embodiment, the transformation module is specifically configured to: and transforming the preprocessed wireless radio frequency signal by selecting short-time Fourier transform of a Kaiser window function, and obtaining a spectrogram of the signal.

In an alternative embodiment of the present application, based on the above embodiment, the extracting module specifically includes:

the time domain extraction unit is used for inputting the enhanced signal into an upper branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to acquire a time domain characteristic.

The frequency domain unit is used for inputting the spectrogram into a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network to obtain frequency domain characteristics.

The convolution layers of the upper branch network and the lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network adopt a weight sharing connection mode.

In an alternative embodiment of the present application, the self-focusing mechanism includes a time-domain signal self-focusing module and a spectrogram self-focusing module.

In an alternative embodiment of the present application based on the above embodiment, the fusion module specifically includes:

a weight vector acquisition unit for respectively characterizing the time domain through two full connection layersAnd frequency domain featuresInput time domain signal self-attention module and spectrogram self-attention module, obtain time domain characteristic +.>And frequency domain features->And (5) a weight vector.

A weight acquisition unit for obtaining the time domain characteristicsAnd frequency domain features->The weight vectors are concatenated and fed to a two-dimensional convolution layer and then passed through the activation function +.>And +.>And finally generating the self-adaptive weight。

A fusion unit for adapting the weight according to the self-adapting weightFusing the temporal features->And frequency domain featuresAnd acquiring fingerprint characteristics based on the wireless radio frequency signals. Wherein, the fusion formula is:

in the method, in the process of the application,is a classifier->Representing transpose, < >>Is self-adaptive weight->Time domain features, < >>Frequency domain characteristics.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The apparatus and method embodiments described above are merely illustrative, for example, flow diagrams and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

In addition, functional modules in the embodiments of the present application may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.

The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, an electronic device, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The terminology used in the embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be understood that the term "and/or" as used herein is merely one relationship describing the association of the associated objects, meaning that there may be three relationships, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.

Depending on the context, the word "if" as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to detection". Similarly, the phrase "if determined" or "if detected (stated condition or event)" may be interpreted as "when determined" or "in response to determination" or "when detected (stated condition or event)" or "in response to detection (stated condition or event), depending on the context.

References to "first\second" in the embodiments are merely to distinguish similar objects and do not represent a particular ordering for the objects, it being understood that "first\second" may interchange a particular order or precedence where allowed. It is to be understood that the "first\second" distinguishing aspects may be interchanged where appropriate, such that the embodiments described herein may be implemented in sequences other than those illustrated or described herein.

The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims (7)

1. A random fence-based dual-branch enhanced radio frequency signal fingerprint identification method is characterized by comprising the following steps of:

acquiring a wireless radio frequency signal containing fingerprint information and preprocessing;

the preprocessed wireless radio frequency signals are enhanced through a random fence characteristic enhancement algorithm, and enhanced signals are obtained;

transforming the preprocessed wireless radio frequency signals through short-time Fourier transform to obtain a spectrogram;

respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of a BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics;

and combining the time domain features and the frequency domain features through a self-attention mechanism to acquire fingerprint features based on the wireless radio frequency signals.

2. The method for identifying the fingerprint of the dual-branch enhanced radio frequency signal based on the random fence according to claim 1, wherein the steps of acquiring the radio frequency signal containing fingerprint information and preprocessing are performed, and specifically include:

acquiring a wireless radio frequency signal containing fingerprint information, performing unified noise reduction processing through a singular value decomposition method, synchronizing the starting points of the radio frequency signals through a transient detection method based on phase characteristics, and performing normalization processing through a maximum and minimum normalization method to acquire a preprocessed wireless radio frequency signal.

3. The method for identifying the fingerprint of the dual-branch enhanced radio frequency signal based on the random fence according to claim 1, wherein the method for enhancing the preprocessed radio frequency signal by the random fence characteristic enhancement algorithm is characterized by comprising the following steps:

based on random segment numbersDividing the pretreated wireless radio frequency signal into +.>Fragments; wherein (1)>Wherein->The length of the preprocessed radio frequency signal, </i >>Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment;

respectively toThe segments perform the steps of performing enhancement processing by a random fence feature enhancement algorithm, obtaining an enhanced signal,

according to the preset reservation degree of the original signalAcquiring a filling area and an original signal area of a fragment;

performing enhancement processing through a random fence characteristic enhancement algorithm according to the filling area and the original signal area; the definition formula of the random fence characteristic enhancement algorithm is as follows:

in the method, in the process of the application,for enhanced signal->Is->No. 5 of the individual fragment>Time domain signal values at each moment,Is a random weight->Length of randomly filled region in a segment, +.>Is the length of the original signal region in the segment.

4. The method for identifying the fingerprint of the dual-branch enhanced radio frequency signal based on the random fence according to claim 3, wherein the pre-processed radio frequency signal is transformed by short-time fourier transform to obtain a spectrogram, and the method specifically comprises the following steps:

and transforming the preprocessed wireless radio frequency signal by selecting short-time Fourier transform of a Kaiser window function, and obtaining a spectrogram of the signal.

5. The method for fingerprint identification of a dual-branch enhanced radio frequency signal based on a random fence according to claim 1, wherein the enhanced signal and the spectrogram are input into an upper branch network and a lower branch network of a BBSEN symmetric bilateral branch feature enhanced network respectively, and time domain features and frequency domain features are acquired, specifically comprising:

inputting the enhanced signal into an upper branch network of a BBSEN symmetrical bilateral branch characteristic enhanced network to acquire a time domain characteristic;

inputting the spectrogram into a lower branch network of a BBSEN symmetrical bilateral branch characteristic enhancement network to obtain frequency domain characteristics;

the convolution layers of the upper branch network and the lower branch network of the BBSEN symmetrical bilateral branch characteristic enhancement network adopt a weight sharing connection mode.

6. The method for fingerprint identification of a dual-branch enhanced radio frequency signal based on a random fence according to claim 1, wherein the self-attention mechanism comprises a time-domain signal self-attention module and a spectrogram self-attention module;

combining the time domain features and the frequency domain features through a self-attention mechanism to acquire fingerprint features based on wireless radio frequency signals, wherein the method specifically comprises the following steps:

time domain features are respectively carried out through two full connection layersAnd frequency domain features->Input time domain signal self-attention module and spectrogram self-attention module, obtain time domain characteristic +.>And frequency domain features->A weight vector;

characterizing the time domainAnd frequency domain features->The weight vectors are concatenated and fed to a two-dimensional convolution layer and then passed through the activation function +.>And +.>After which the adaptive weights are finally generated>；

According to the adaptive weightFusing the temporal features->And frequency domain features->Acquiring fingerprint characteristics based on wireless radio frequency signals; wherein, the fusion formula is:

in the method, in the process of the application,is a classifier->Representing transpose, < >>Is self-adaptive weight->Time domain features, < >>Frequency domain characteristics.

7. A dual-branch enhanced radio frequency signal fingerprint identification device based on a random fence, comprising:

the preprocessing module is used for acquiring a wireless radio frequency signal containing fingerprint information and preprocessing the wireless radio frequency signal;

the enhancement module is used for enhancing the preprocessed wireless radio frequency signals through a random fence characteristic enhancement algorithm to obtain enhancement signals;

the transformation module is used for transforming the preprocessed wireless radio frequency signals through short-time Fourier transformation to obtain a spectrogram;

the extraction module is used for respectively inputting the enhanced signal and the spectrogram into an upper branch network and a lower branch network of the BBSEN symmetrical bilateral branch characteristic enhanced network to obtain time domain characteristics and frequency domain characteristics;

and the fusion module is used for merging the time domain features and the frequency domain features through a self-attention mechanism to acquire fingerprint features based on the wireless radio frequency signals.