A unique identification method for radio signals based on I / Q component weight features

Abstract

This invention provides a method for unique identification of radio signals based on I / Q component weighting features. The method includes: S11, acquiring I / Q data for the monitoring frequency band of the target radio device at each detection node; S12, processing the I / Q data to obtain an evaluation value; S13, determining whether the target radio device has a signal based on the evaluation value. If yes, proceed to step S14; otherwise, repeat step S11; S14, performing secondary processing on the I / Q data to obtain a radio frequency fingerprint feature value, matching the radio frequency fingerprint feature value with pre-stored radio frequency fingerprint feature values ​​in a radio frequency fingerprint feature database to identify the signal; S15, if signal identification fails in step S14, storing the radio frequency fingerprint feature value in the radio frequency fingerprint feature database. This invention directly extracts features from I / Q data, fully utilizing the phase and amplitude information of the signal, thereby achieving more accurate signal identification.

Description

Technical Field

[0001] This invention relates to the field of fingerprint feature extraction technology, and in particular to a method for unique identification of radio signals based on I / Q component weight features. Background Technology

[0002] Traditional signal recognition methods include spectral analysis, time-domain analysis, matched filter detection, energy detection, and deep learning-based recognition methods.

[0003] Spectrum analysis is a technique that decomposes complex signals into simpler ones. Its principle is to identify information about these signals at different frequencies, such as amplitude, power, intensity, or phase. A spectrum refers to the frequency domain representation of a signal in the time domain, obtained by performing a Fourier transform on the signal. The result is two graphs plotted with amplitude and phase on the vertical axis and frequency on the horizontal axis, respectively. By performing spectrum analysis, the distribution of the signal in the frequency domain can be obtained. However, the disadvantages of spectrum analysis include its complexity and the potential impact on its accuracy for certain types of signals (such as non-stationary or transient signals).

[0004] Time-domain analysis refers to analyzing the stability, transient, and steady-state performance of a control system under given input, based on the time-domain expression of the output. Processing signals in the time domain, such as filtering, amplification, statistical feature calculation, and correlation analysis, is collectively referred to as time-domain signal analysis. Time-domain analysis methods can effectively improve the signal-to-noise ratio and determine the similarity and correlation of signal waveforms at different times. The disadvantages of time-domain analysis are that it cannot directly provide frequency information and it presents challenges in extracting the time-series features of complex signals. Matched filters are a signal processing technique that works by convolving the received signal with its complex conjugate (time-reversed) and then integrating the result to enhance the signal and suppress noise. Its disadvantages include high computational complexity and the need for accurate prior information about the user signal, leading to difficulties in practical applications.

[0005] Energy detection is a type of incoherent signal detection, suitable for situations where the signal detector is unaware of prior information about the authorized user's signal. Its drawbacks include the difficulty in setting threshold values, and in environments with low signal-to-noise ratios, energy detection can easily identify very high-energy interference noise as useful signals while excluding weak signals with low energy, leading to reduced detection accuracy.

[0006] Deep learning-based recognition methods are advanced signal recognition techniques that have emerged in recent years. They primarily use deep learning models such as convolutional neural networks (CNNs) to extract features and classify signals. This recognition method is complex and computationally resource-intensive, requiring a large amount of training data to optimize the model. Summary of the Invention

[0007] The purpose of this invention is to provide a unique identification method for radio signals based on I / Q component weighting features. By directly extracting features from I / Q data, the phase and amplitude information of the signal can be fully utilized, thereby achieving more accurate identification of the signal.

[0008] To achieve the above objectives, the present invention provides a method for unique identification of radio signals based on I / Q component weighting features, the method comprising:

[0009] S11. Collect I / Q data for the monitoring frequency band of the target wireless equipment at each detection node;

[0010] S12. Process the I / Q data to obtain the value to be evaluated;

[0011] S13. Determine whether the target wireless device has a signal based on the value to be evaluated. If yes, proceed to step S14; otherwise, repeat step S11.

[0012] S14. Perform secondary processing on the I / Q data to obtain the radio frequency fingerprint feature value. Match the radio frequency fingerprint feature value with the pre-stored radio frequency fingerprint feature value in the radio frequency fingerprint feature database to identify the signal.

[0013] S15. If signal recognition fails in step S14, the radio frequency fingerprint feature value is stored in the radio frequency fingerprint feature database.

[0014] Furthermore, in step S11, each detection node consists of a main control module and a software radio module. The software radio module is used to acquire I / Q data, receive and preprocess the I / Q data, and the main control module and the software radio module are signal connected.

[0015] Furthermore, in step S12, the I / Q data is processed, specifically including:

[0016] The Q values ​​in the collected I / Q data are subtracted from the end to the beginning. The absolute values ​​of all the subtracted values ​​are then summed to obtain the value to be evaluated.

[0017] Furthermore, in step S13, determining whether the target wireless device has a signal based on the value to be evaluated specifically includes:

[0018] The value to be evaluated is compared with a preset threshold. If the value to be evaluated is less than the preset threshold, it is determined that no signal has been detected, and step S11 is executed again. If the value to be evaluated is greater than or equal to the preset threshold, step S14 is executed.

[0019] Furthermore, in step S14, the collected data undergoes secondary processing, specifically including:

[0020] S21. Statistically analyze the count value of each I / Q data point and use the count value as the feature value corresponding to each I / Q data point;

[0021] S22. Set the weight values ​​for I / Q data;

[0022] S23. Multiply each feature value by its corresponding weight value and then sum them up to get the x-value on the new coordinate axis.

[0023] S24. Sum the absolute values ​​of the horizontal and vertical coordinates of all points that have appeared in the two-dimensional distribution feature, and use the sum as the y-value on the new coordinate axis. The two-dimensional distribution feature is composed of I and Q.

[0024] S25. Process all I / Q data from one sampling through steps S21 to S24, fill in new coordinates, and transform them into a curve that is approximately an 8th-degree polynomial consisting of 9 parameters. Fit the curve to obtain the characteristic distribution curve of the signal.

[0025] S26. The vertex coordinates of the characteristic distribution curve and the 9 parameters are used as the radio frequency fingerprint feature values ​​of the signal source.

[0026] Furthermore, statistical analysis is performed on the count value for each I / Q data point, specifically including:

[0027] S31. Describe each I / Q component of the I / Q data using the complex form I+jQ, with I as the horizontal axis and Q as the vertical axis, to obtain the two-dimensional distribution of each I / Q component.

[0028] S32. The number of points with the same horizontal and vertical coordinates is taken as the count value of the I / Q component;

[0029] S33. Use the count value as the feature value corresponding to each I / Q data.

[0030] Furthermore, when setting the weight values ​​for I / Q data, the stronger the signal energy, the closer the I and Q values ​​of the I / Q data are to the four coordinate points (1, -1), (-1, -1), (1, 1), and (-1, 1), and the smaller the weight value; the weaker the signal energy, the more concentrated the I and Q values ​​of the I / Q data are at the coordinate point (0, 0), and the larger the weight value. The weights are set according to this rule.

[0031] Furthermore, the expression for the curve, which consists of an approximate 8th-degree polynomial composed of 9 parameters, is as follows:

[0032] Curve = a + b*X + c*X 2 +d*X 3 +e*X 4

[0033] +f*X 5 +g*X 6 +h*X 7 +i*X 8

[0034] Where Curve is an 8th-degree polynomial curve, a, b, c, d, e, f, g, h, i are constant terms, i.e., 9 parameters, and X is the independent variable.

[0035] Furthermore, the radio frequency fingerprint feature value is matched with the radio frequency fingerprint feature value in the radio frequency fingerprint feature database to identify the signal, specifically including:

[0036] S41. First, compare the vertex coordinates of the radio frequency fingerprint feature value with the vertex coordinates in the radio frequency fingerprint feature database one by one;

[0037] S42. If the distance between the vertex coordinates of the radio frequency fingerprint feature value and the vertex coordinates in the radio frequency fingerprint feature database is greater than a preset threshold, the matching fails and the radio frequency fingerprint feature value is stored in the radio frequency fingerprint feature database.

[0038] S43. If the distance between the vertex coordinates of the RF fingerprint feature value and the vertex coordinates in the RF fingerprint feature database is less than or equal to a preset threshold, continue to match the curve of an approximate 8th degree polynomial consisting of 9 parameters, with the average error between the horizontal coordinates of 210000 and 500000. If the average error is greater than the preset threshold, the matching fails, and the RF fingerprint feature value is stored in the RF fingerprint feature database. If the average error is less than or equal to the preset threshold, the matching succeeds.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] This invention provides a unique identification method for radio signals based on I / Q component weighting features. It utilizes only the raw I / Q data, omitting energy data. The feature curves extracted from the I / Q data remain unchanged despite environmental variations and center frequency shifts, exhibiting excellent stability. Compared to electromagnetic spectrum energy detection methods, it demonstrates a higher recognition rate. Furthermore, compared to deep learning-based signal identification methods, this method requires less computation and has lower complexity. In contrast to methods using other signal features as RF fingerprints, directly extracting features from I / Q data plays a crucial role in identifying signal modulation schemes, transmission sources, and other information.

[0041] Compared with traditional signal identification methods, the feature extraction of this invention is more direct and efficient, and can make full use of the phase and amplitude information of the signal, thereby achieving more accurate signal identification. Especially in complex communication environments, such as those with multipath interference and noise interference, it can still maintain high accuracy. Radio measurement tasks require improving the capabilities of each electromagnetic spectrum monitoring terminal device. The I / Q data distribution characteristics proposed in this invention are baseband signals, from which the radio frequency hardware characteristics of radio devices can be effectively extracted. This is suitable for building distributed electromagnetic spectrum monitoring terminals that use radio devices as monitoring objects. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 This is a schematic flowchart of a method for unique identification of radio signals based on I / Q component weight features, provided in an embodiment of the present invention. Detailed Implementation

[0044] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0045] Reference Figure 1 This embodiment provides a method for unique identification of radio signals based on I / Q component weighting features. The method includes:

[0046] S11. I / Q data are acquired for the monitoring frequency band of the target wireless device at each detection node. In step S11, each detection node consists of a main control module and a software radio module. The software radio module is used to acquire I / Q data, receive and preprocess the I / Q data, and the main control module and the software radio module are signal-connected.

[0047] In this embodiment, each detection node consists of a main control module and a software-defined radio (SDR) module. The main control module can be a Windows platform or an embedded hardware platform, while the SDR module can be any type of SDR product. The SDR module is primarily responsible for acquiring I / Q data, while the main control module controls the SDR module, enabling it to receive and preprocess I / Q data. This embodiment can combine multiple monitoring nodes to collaboratively acquire and process RF I / Q data.

[0048] S12. Process the I / Q data to obtain the value to be evaluated. Step S12, processing the I / Q data, specifically includes:

[0049] The Q values ​​in the collected I / Q data are subtracted from the end to the beginning. The absolute values ​​of all the subtracted values ​​are then summed to obtain the value to be evaluated.

[0050] In this embodiment, the value to be evaluated increases as the signal energy intensity increases. At a signal-to-noise ratio of 0dB, this value is still different from the value obtained from pure noise. The distinction between signal and noise is achieved by setting a threshold.

[0051] S13. Determine whether the target wireless device has a signal based on the value to be evaluated. If yes, proceed to step S14; otherwise, repeat step S11. Step S13, determining whether the target wireless device has a signal based on the value to be evaluated, specifically includes:

[0052] The value to be evaluated is compared with a preset threshold. If the value to be evaluated is less than the preset threshold, it is determined that no signal has been detected, and step S11 is executed again. If the value to be evaluated is greater than or equal to the preset threshold, step S14 is executed.

[0053] In this embodiment, if no signal is detected, the signal of the target wireless device continues to be collected until it is determined that there is a signal in the monitoring frequency band, and then the signal is identified.

[0054] S14. Perform secondary processing on the I / Q data to obtain RF fingerprint feature values. Match the RF fingerprint feature values ​​with pre-stored RF fingerprint feature values ​​in the RF fingerprint feature database to identify the signal. Step S14 involves secondary processing of the acquired data, specifically including:

[0055] S21. Statistically analyze the count value of each I / Q data point and use the count value as the feature value corresponding to each I / Q data point. Extract the I / Q data collected by the software radio module, and statistically analyze the count value of each I / Q data point. In the I / Q data of the RF baseband signal, each I / Q component is described in the complex form I+jQ. With I as the horizontal axis and Q as the vertical axis, the two-dimensional distribution of each I / Q component can be obtained (all I and Q values ​​are between -1 and 1). One I / Q data point contains 131072 I / Q components. In the coordinate system composed of I and Q, the number of points with the same horizontal and vertical coordinates is used as the count value of that I / Q component. The count value is used as the feature value of the I / Q component under the same horizontal and vertical coordinates. By setting the count value, one I / Q data point can be transformed into N feature values.

[0056] S22. Set the weight values ​​for I / Q data. By analyzing the two-dimensional distribution of I / Q data, it is found that the stronger the signal energy intensity, the more dispersed the two-dimensional distribution of I / Q data becomes, eventually concentrating on the four coordinate points (1, -1), (-1, -1), (1, 1), and (-1, 1). The weaker the signal energy intensity, the more concentrated the two-dimensional distribution of I / Q data is on the (0, 0) coordinate. The closer to (0, 0), the larger the weight value; the closer to the four coordinate points (1, -1), (-1, -1), (1, 1), and (-1, 1), the smaller the weight value. The weights are set according to this rule.

[0057] S23. Multiply each feature value by its corresponding weight value and sum them to obtain the x-value on the new coordinate axis. After processing an I / Q data point in S21, N feature values ​​are obtained. Each feature value is multiplied by its corresponding weight value and summed to obtain the x-value on the new coordinate axis. The x-value is used to describe the distribution characteristics of the I / Q data. The larger the x-value, the weaker the signal strength.

[0058] S24. Summate the absolute values ​​of the horizontal and vertical coordinates of all points that have appeared in the two-dimensional distribution feature, and use this sum as the y-value on the new coordinate axis. The two-dimensional distribution feature consists of I and Q. After the I / Q data is transformed by S21, sum the absolute values ​​of the coordinates of all points that have appeared in the coordinate system composed of I and Q, and use this sum as the y-value on the new coordinate axis. That is, the stronger the signal energy, the wider the distribution range, and the weaker the signal energy, the narrower the distribution range. Based on this rule, the range of the signal distribution (i.e., the sum of the absolute values ​​of the horizontal and vertical coordinates of the points that have appeared) is used as the y-value in the new coordinate system.

[0059] S25. Process all I / Q data from a single sampling through steps S21 to S24, fill in the new coordinate transformation to form a curve containing an approximate 8th-degree polynomial with 9 parameters, and fit the curve to obtain the characteristic distribution curve of the signal. One I / Q data point is transformed into a point on the new coordinate axis. A single sampling includes multiple I / Q data points. Transforming all I / Q data from a single sampling through steps S21 to S24 forms a curve containing an approximate 8th-degree polynomial with 9 parameters. Fitting the 8th-degree polynomial yields the characteristic distribution curve of the signal.

[0060] S26. The vertex coordinates of the characteristic distribution curve and the 9 parameters are used as the radio frequency fingerprint feature values ​​of the signal source. The vertex coordinates of the fitted characteristic distribution curve and the 9 parameters are used as the radio frequency fingerprint feature values ​​of the signal source.

[0061] Statistical analysis of the count value for each I / Q data point, specifically including:

[0062] S31. Describe each I / Q component of the I / Q data using the complex form I+jQ, with I as the horizontal axis and Q as the vertical axis, to obtain the two-dimensional distribution of each I / Q component.

[0063] S32. The number of points with the same horizontal and vertical coordinates is taken as the count value of the I / Q component.

[0064] S33. Use the count value as the feature value corresponding to each I / Q data.

[0065] In this embodiment, the I / Q data collected by the software radio module is extracted, and the count value of each I / Q data point is statistically analyzed. In the I / Q data of the RF baseband signal, each I / Q component is described by the complex number I+jQ. Using I as the horizontal axis and Q as the vertical axis, a two-dimensional distribution of each I / Q component can be obtained (all I and Q values ​​are between -1 and 1). One I / Q data point contains 131072 I / Q components. In the coordinate system composed of I and Q, the number of points with the same horizontal and vertical coordinates is taken as the count value of that I / Q component. The count value is used as the feature value of the I / Q component under the same horizontal and vertical coordinates. By setting the count value, one I / Q data point can be transformed into N feature values.

[0066] When setting the weight values ​​for I / Q data, the stronger the signal energy, the closer the I and Q values ​​of the I / Q data are to the four coordinate points (1, -1), (-1, -1), (1, 1), and (-1, 1), and the smaller the weight value; the weaker the signal energy, the more concentrated the I and Q values ​​of the I / Q data are at the coordinate point (0, 0), and the larger the weight value. The weights are set according to this rule.

[0067] In this embodiment, by analyzing the two-dimensional distribution of I / Q data, it was found that the stronger the signal energy intensity, the more dispersed the two-dimensional distribution of I / Q data becomes, eventually concentrating on four coordinate points: (1, -1), (-1, -1), (1, 1), and (-1, 1). Conversely, the weaker the signal energy intensity, the more concentrated the two-dimensional distribution of I / Q data becomes at the coordinate point (0, 0), with a larger weight value closer to (0, 0) and a larger weight value closer to (1, -1), (-1, -1), (1, 1), and (-1, 1). 1) The smaller the weight value of the four coordinate points, that is, in the coordinate system obtained with I as the horizontal axis and Q as the vertical axis, the stronger the signal energy, the closer the I and Q values ​​of the I / Q data are to 1 and -1, and the weaker the signal energy, the closer the I and Q values ​​of the I / Q data are to 0. The closer the point (I, Q) is to (0, 0), the larger the weight value of the coordinate point. The closer the point (I, Q) is to (1, -1), (-1, -1), (1, 1), (-1, 1), the smaller the weight value of the four coordinate points. The weights are set according to this rule.

[0068] The expression for a curve that is an approximate 8th-degree polynomial consisting of 9 parameters is:

[0069] Curve = a + b*X + c*X 2 +d*X 3 +e*X 4

[0070] +f*X 5 +g*X 6 +h*X 7 +i*X 8

[0071] Where Curve is an 8th-degree polynomial curve, a, b, c, d, e, f, g, h, i are constant terms, i.e., 9 parameters, and X is the independent variable.

[0072] In this embodiment, the degree of the polynomial is directly related to the shape of the curve. By using an nth-degree polynomial to fit a function to the new coordinate point, n+1 variables can be generated (in this embodiment, the 8th-degree polynomial fitting yields nine new parameters: a, b, c, d, e, f, g, h, and i). These fitted variables are the eigenvalues. Through testing, it was found that the curve fitted using the 8th-degree polynomial has the highest similarity to the new coordinate point.

[0073] The process involves matching the radio frequency fingerprint feature value with the radio frequency fingerprint feature value in the radio frequency fingerprint feature database to identify the signal, specifically including:

[0074] S41. First, compare the vertex coordinates of the radio frequency fingerprint feature value with the vertex coordinates in the radio frequency fingerprint feature database one by one;

[0075] S42. If the distance between the vertex coordinates of the radio frequency fingerprint feature value and the vertex coordinates in the radio frequency fingerprint feature database is greater than a preset threshold, the matching fails and the radio frequency fingerprint feature value is stored in the radio frequency fingerprint feature database.

[0076] S43. If the distance between the vertex coordinates of the RF fingerprint feature value and the vertex coordinates in the RF fingerprint feature database is less than or equal to a preset threshold, continue to match the curve of an approximate 8th degree polynomial consisting of 9 parameters, with the average error between the horizontal coordinates of 210000 and 500000. If the average error is greater than the preset threshold, the matching fails, and the RF fingerprint feature value is stored in the RF fingerprint feature database. If the average error is less than or equal to the preset threshold, the matching succeeds.

[0077] S15. If signal recognition fails in step S14, the RFID fingerprint feature value is stored in the RFID fingerprint feature database. In this embodiment, if signal recognition fails, the recognized signal and its tag are stored together in the RFID fingerprint database, and the signal type in the database is updated.

[0078] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for unique identification of radio signals based on I / Q component weighting features, characterized in that, The method includes: S11. Collect I / Q data for the monitoring frequency band of the target wireless equipment at each detection node; S12. Process the I / Q data to obtain the value to be evaluated; S13. Determine whether the target wireless device has a signal based on the value to be evaluated. If yes, proceed to step S14; otherwise, repeat step S11. S14. Perform secondary processing on the I / Q data to obtain the radio frequency fingerprint feature value. Match the radio frequency fingerprint feature value with the pre-stored radio frequency fingerprint feature value in the radio frequency fingerprint feature database to identify the signal. In step S14, the collected data undergoes secondary processing, specifically including: S21. Statistically analyze the count value of each I / Q data point and use the count value as the feature value corresponding to each I / Q data point; S22. Set the weight values ​​for I / Q data; S23. Multiply each feature value by its corresponding weight value and then sum them up to get the x-value on the new coordinate axis. S24. Sum the absolute values ​​of the horizontal and vertical coordinates of all points that have appeared in the two-dimensional distribution feature, and use the sum as the y-value on the new coordinate axis. The two-dimensional distribution feature is composed of I and Q. S25. Process all I / Q data from one sampling through steps S21 to S24, fill in new coordinates, and transform them into a curve that is approximately an 8th-degree polynomial consisting of 9 parameters. Fit the curve to obtain the characteristic distribution curve of the signal. S26. Use the vertex coordinates of the characteristic distribution curve and 9 parameters as the radio frequency fingerprint feature values ​​of the signal source; S15. If signal recognition fails in step S14, the radio frequency fingerprint feature value is stored in the radio frequency fingerprint feature database.

2. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, In step S11, each detection node consists of a main control module and a software radio module. The software radio module is used to collect I / Q data, receive and preprocess the I / Q data, and the main control module and the software radio module are signal connected.

3. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, In step S12, the I / Q data is processed, specifically including: The Q values ​​in the collected I / Q data are subtracted from the end to the beginning. The absolute values ​​of all the subtracted values ​​are then summed to obtain the value to be evaluated.

4. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, In step S13, determining whether the target wireless device has a signal based on the value to be evaluated specifically includes: The value to be evaluated is compared with a preset threshold. If the value to be evaluated is less than the preset threshold, it is determined that no signal has been detected, and step S11 is executed again. If the value to be evaluated is greater than or equal to the preset threshold, step S14 is executed.

5. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, Statistical analysis of the count value for each I / Q data point, specifically including: S31. Describe each I / Q component of the I / Q data using the complex form I+jQ, with I as the horizontal axis and Q as the vertical axis, to obtain the two-dimensional distribution of each I / Q component. S32. The number of points with the same horizontal and vertical coordinates is taken as the count value of the I / Q component; S33. Use the count value as the feature value corresponding to each I / Q data.

6. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, When setting the weight values ​​for I / Q data, the stronger the signal energy, the closer the I and Q values ​​of the I / Q data are to the four coordinate points (1, -1), (-1, -1), (1, 1), and (-1, 1), and the smaller the weight value; the weaker the signal energy, the more concentrated the I and Q values ​​of the I / Q data are at the coordinate point (0, 0), and the larger the weight value. The weights are set according to this rule.

7. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, The expression for a curve that is an approximate 8th-degree polynomial consisting of 9 parameters is: in, The curve is an 8th degree polynomial. , , , , , , , , These are constant terms, i.e., 9 parameters. is the independent variable.

8. The method for unique identification of radio signals based on I / Q component weighting features according to claim 1, characterized in that, The process involves matching the radio frequency fingerprint feature value with the radio frequency fingerprint feature value in the radio frequency fingerprint feature database to identify the signal, specifically including: S41. First, compare the vertex coordinates of the radio frequency fingerprint feature value with the vertex coordinates in the radio frequency fingerprint feature database one by one. S42. If the distance between the vertex coordinates of the radio frequency fingerprint feature value and the vertex coordinates in the radio frequency fingerprint feature database is greater than a preset threshold, the matching fails and the radio frequency fingerprint feature value is stored in the radio frequency fingerprint feature database. S43. If the distance between the vertex coordinates of the RF fingerprint feature value and the vertex coordinates in the RF fingerprint feature database is less than or equal to a preset threshold, continue to match the curve of an approximate 8th degree polynomial consisting of 9 parameters, with the average error between the horizontal coordinates of 210000 and 500000. If the average error is greater than the preset threshold, the matching fails, and the RF fingerprint feature value is stored in the RF fingerprint feature database. If the average error is less than or equal to the preset threshold, the matching succeeds.

Images