Radar radiation source individual identification method based on multi-domain feature joint judgment

Abstract

The invention discloses a radar radiation source individual identification method based on multi-domain feature joint judgment. The technical scheme comprises the following steps: step 1, signal preprocessing; 2, original signal detection and parameter measurement; 3, signal feature extraction and multi-feature-domain conjoint analysis are carried out; and step 4, signal characteristic quantity visualized visual display is carried out. Therefore, multi-domain feature extraction and recognition of the radar radiation source individuals are realized. According to the method, the multi-domain features of the radar radiation source signals are extracted, multiple kinds of reference information are provided for distinguishing a radar carrier platform, the problem that the single feature recognition effect is poor is solved, and the radar radiation source individual recognition accuracy is improved; a set of powerful tool is provided for a user by utilizing an MATLAB GUI user interface prototype, so that radar signals can be better understood and analyzed. A user can realize comprehensive analysis from an original signal to multi-dimensional fine characteristics through the interface, so that the user can deeply research signal characteristics, and more accurate decision making and further research are facilitated.

Description

Technical field:

[0001] The invention belongs to the technical field of electronic reconnaissance, and mainly relates to a radar radiation source individual identification method based on multi-domain feature joint judgment. Background technology:

[0002] The battlefield electromagnetic environment is complex and changeable. Sorting unknown radar emitter signals has become a key link in the signal processing of electronic reconnaissance systems, which to a certain extent restricts the generation of electronic intelligence, electronic support measures and threat warning capabilities. In a complex electromagnetic environment, how to effectively distinguish the radar carrier platform of the combat opponent, accurately grasp the distribution of electromagnetic targets and complete the individual identification of radar emitters has always been a hot and difficult issue in this field.

[0003] Individual identification of radar emitters requires the use of the "unintentional modulation" features carried in radar pulse signals to uniquely identify individual radar transmitters. At present, research on individual identification of radar emitters mainly includes identification based on a single feature such as time domain features, frequency characteristics, etc. However, with the high degree of confrontation of electronic information warfare and the rapid advancement of new radar technology, the operating frequency range of radars is extending, signal parameters are more flexible, and signal waveforms change simultaneously in multiple domains such as time domain and frequency domain, which brings serious difficulties to current signal sorting and identification. Therefore, a relatively reliable solution is to combine multiple feature quantities and use multi-dimensional information to identify the emitter. Summary of the invention:

[0004] In order to overcome the above-mentioned shortcomings, the present invention provides a radar emitter individual identification method based on multi-domain feature joint judgment.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A radar emitter individual identification method based on multi-domain feature joint judgment includes the following steps:

[0007] Step 1: Signal preprocessing includes: noise reduction, time alignment and normalization, and multipath detection;

[0008] Step 2: Original signal detection and parameter measurement, including original signal analysis, pulse and parameter extraction, and pulse storage;

[0009] Step 3: Signal feature extraction domain analysis, including feature extraction and joint analysis of multiple feature quantities in time domain, frequency domain, time-frequency domain, cyclic spectrum and cyclic bispectrum;

[0010] Step 4: Signal feature visualization, by developing a set of feature visualization display tools with the help of MATLAB graphical user interface software platform.

[0011] In the step 1), the denoising adopts a wavelet threshold denoising algorithm, so as to process the measured radar radiation source signal and realize signal-noise separation in the wavelet domain.

[0012] The specific steps of time alignment and normalization in step 1) are as follows: sorting the intercepted pulses according to the signal-to-noise ratio, selecting the pulse with the largest signal-to-noise ratio as the first reference pulse, and calculating the cross-correlation value between the kth input pulse and each aligned pulse; aligning the pulses according to the time delay corresponding to the maximum cross-correlation value;

[0013] The pulse data is normalized to limit the data within a certain range, and the signal amplitude is normalized to 1 using the variance normalization method.

[0014] In the step 1), the multipath detection is performed by using the measured signal multipath detection based on the Wigner-Ville distribution; first, a sliding window is used to intercept the signal, and a threshold value is pre-set so that the signal in the final sliding window can extract the tail effective detection data containing less noise; secondly, the Wigner-Ville distribution of the tail of the extracted signal data is calculated, the detection parameters are extracted, and the optimal detection threshold is set.

[0015] In the step 2), during the original signal analysis process, the amplitude and time are analyzed from the overall perspective of the received signal. After a pulse signal is intercepted from the original signal, the amplitude and flatness angle of the single pulse are analyzed; the main characteristics of the target radar signal are established as a theoretical basis for subsequent pulse detection and extraction, and a large amount of data is used to analyze the applicability of these characteristics.

[0016] In the step 2), during the pulse and parameter extraction process, a pulse detection processing technique is used to detect a specific type of pulse signal by setting a threshold; during the pulse extraction process, the original signal is first subjected to wavelet noise reduction processing; secondly, a threshold is selected; finally, the amplitude of the signal is monitored, and when the signal exceeds or drops to the set threshold, pulse detection is triggered;

[0017] During the parameter extraction process, the number of pulses, the starting and ending positions, and the pulse point coordinates are extracted; first, the average value and threshold of the signal are calculated, and then the pulse points are filtered according to the threshold; secondly, the starting and ending positions of the detected pulses are saved in an array, and the amplitude and width of the pulses are calculated; finally, the pulse data is sorted, short pulses are removed, and the relevant information of the pulses is stored, including the pulse sequence number, starting time, amplitude, and width. In step 3), for the extraction of individual features in the time domain, the Hilbert transform method is used to extract the different transient information presented by the radar signal envelope in time. First, the input data is preprocessed to exclude the zero-value elements on both sides, and the valid data retained is from the first non-zero element to the last non-zero element; then the optimized data is subjected to the Hilbert transform to obtain the instantaneous frequency and amplitude values. In the process of obtaining the envelope, the amplitude value is used; finally, the retained data is subjected to a sliding window median filter operation to obtain the final envelope result;

[0018] To extract individual features in the frequency domain, accurate signal carrier frequency estimation is used to perform carrier frequency difference analysis on radar signals with the same system and equivalent signal strength, including two algorithms: the dual-line amplitude method and the single-line phase method.

[0019] For the extraction of individual features in the time-frequency domain, the time-frequency localized area information is extracted by wavelet packet transform algorithm, the fuzzy function algorithm is used to obtain fuzzy function slices based on different frequency deviation parameter settings, and the finite number of inherent mode function components obtained after decomposing the multi-component signal by the classical mode decomposition algorithm are used to reconstruct the time-frequency distribution of the radiation source signal, so as to extract the time-frequency diagram as the classification feature;

[0020] Extract individual features of the cyclic spectrum, and extract the transformed plane slices intercepted after the cyclic spectrum transformation as feature vectors;

[0021] For the individual feature extraction of the cyclic bispectrum, the diagonal slices of the cyclic bispectrum are extracted using the cyclic bispectrum function, and the singular value decomposition is performed at the same time, and the extracted singular values ​​or singular vectors are used as the individual features of the radiation source.

[0022] The specific process of realizing signal feature visualization includes:

[0023] First, data is read. Currently, there are two types of data files that can be read in: the first is the original signal data file, with the suffix .bxsldmn, and the second is the file that has been through pulse detection and parameter measurement, with the suffix .pls. The file is saved in the format of the pulse saved in step 2);

[0024] The second is data viewing. For the original signal data bxsldmn, after reading in, it needs to go through pulse detection and extract the pulse signal before subsequent feature viewing can be performed. For the pls file, it saves the data after pulse detection and extraction, so there is no need to go through the pulse detection step, and you can directly enter the pulse signal viewing page;

[0025] After the above process, enter the feature result view. In the feature signal visualization window, the interface content includes pulse signal envelope, spectrum, power spectrum, fuzzy function, SVD singular value, cyclic spectrum, wavelet packet decomposition, time-frequency reconstruction, third-order cumulant, fourth-order cumulant and other feature information;

[0026] The interface also sets up save configuration options, and selects the output directory to save the feature information that needs to be saved.

[0027] Due to the adoption of the above-mentioned technical solution, the present invention has the following advantages:

[0028] 1) Extracting multi-domain features of radar emitter signals to provide multiple reference information for identifying radar carrier platforms helps solve the problem of poor single feature recognition and improve the accuracy of individual radar emitter recognition.

[0029] 2) The MATLAB GUI user interface prototype provides users with a powerful set of tools to better understand and analyze radar signals. Through this interface, users can achieve a comprehensive analysis from raw signals to features, enabling them to study signal characteristics more deeply, which helps to make more accurate decisions and further research. Description of the drawings:

[0030] Figure 1 The flowchart of realizing radar radiation source signal extraction in the example of the present invention.

[0031] Figure 2 This is a flow chart of the wavelet threshold denoising algorithm used in the example of the present invention.

[0032] Figure 3 Flowchart for pulse alignment for the correlation process used in examples of the present invention.

[0033] Figure 4 This is a flow chart of the multipath detection algorithm used in the example of the present invention.

[0034] Figure 5 This is the basic process of calculating the envelope by Hilbert transform used in the examples of the present invention.

[0035] Figure 6 This is the prototype page for visualizing the original signal implemented in the example of the present invention.

[0036] Figure 7This is a prototype page for visualizing the features implemented in the example of the present invention.

[0037] Figure 8 The figure is a specific flow chart for visualizing the signal characteristics implemented in the example of the present invention. Specific implementation method:

[0038] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0039] like Figure 1 As shown, the radar emitter individual identification method based on multi-domain feature joint judgment provided by the present invention includes the following steps:

[0040] Step 1: Signal preprocessing mainly includes noise reduction, time alignment and normalization, and multipath detection.

[0041] Since wavelet transform has the characteristics of low entropy, multi-resolution, decorrelation and flexible basis function, wavelet threshold denoising algorithm is adopted in the denoising process to process the measured radar radiation source signal and realize signal-noise separation in the wavelet domain. The flow chart is shown in Figure 2 .

[0042] In the normalization process, Figure 3 As shown, the signal frequency, time delay and signal amplitude are processed. Frequency normalization transforms the carrier frequencies of different signals to the same frequency band, usually performed at intermediate frequency sampling. Time normalization uses correlation processing to align pulse waveforms, where the intercepted pulses are first sorted by signal-to-noise ratio, and the pulse with the largest signal-to-noise ratio is selected as the first reference pulse. For the kth input pulse, it is necessary to calculate its cross-correlation value with each aligned pulse, and finally align the pulses according to the time delay corresponding to the maximum cross-correlation value. Amplitude energy normalization is to normalize the pulse data, thereby limiting the data to a certain range, and the variance normalization method is used to normalize the signal amplitude to 1.

[0043] In the multipath detection process, if Figure 4As shown in the figure, the measured signal multipath detection based on Wigner-Ville distribution is adopted. The idea is to first extract the effective detection data at the tail of the signal, that is, the signal segment where the frequency drift exists. Due to the existence of noise at the tail of the signal, it will seriously affect the multipath detection performance. Here, a sliding window is used to intercept the signal, and the threshold value is pre-set so that the signal in the final sliding window can extract the tail data of effective detection and contain less noise. Secondly, the Wigner-Ville distribution of the extracted signal data tail is calculated, the detection parameters are extracted, and the optimal detection threshold is set.

[0044] Step 2: For original signal detection and parameter measurement, it mainly includes original signal analysis, pulse and parameter extraction and pulse preservation.

[0045] In the process of analyzing the original signal, the amplitude and time are analyzed from the overall perspective of the received signal. After a pulse signal is intercepted from the original signal, the amplitude and flat angle of a single pulse are analyzed. Through the above analysis, the main characteristics of the target radar signal are established as the theoretical basis for subsequent pulse detection and extraction, and a large amount of data is used to analyze the applicability of these characteristics.

[0046] In the process of pulse and parameter extraction, pulse detection, a signal processing technique, is used to detect specific types of pulse signals by setting thresholds. In the pulse extraction process, the original signal is first subjected to wavelet noise reduction. Secondly, a threshold is selected, usually a voltage or power value. Then, by monitoring the amplitude of the signal, pulse detection is triggered when the signal exceeds or drops to the set threshold.

[0047] In the parameter extraction process, the number of pulses, the starting and ending positions, and the pulse point coordinates are extracted. First, the average value and threshold of the signal are calculated, and then the pulse points are filtered according to the threshold. Secondly, the starting and ending positions of the detected pulses are saved in an array, and the amplitude and width of the pulses are calculated. The amplitude is expressed in decibels and the width is expressed in points. Finally, the pulse data is sorted, short pulses are removed, and the relevant information of the pulses is stored, including the pulse number, starting time, amplitude, and width.

[0048] During the pulse saving process, the obtained pulse signal data and the corresponding pulse basic parameters are saved in a specific format to facilitate direct reading and use of the data source in subsequent analysis.

[0049] Step 3: For signal feature extraction domain analysis, it mainly includes feature extraction and joint analysis of multiple feature quantities in multiple aspects, including time domain, frequency domain, time-frequency domain, cyclic spectrum and cyclic bispectrum.

[0050] For the extraction of individual features in the time domain, such as Figure 5As shown in the figure, the Hilbert transform method is used to extract the different transient information presented in the radar signal envelope in time. First, the input data is preprocessed to exclude the 0-value elements on both sides, and the valid data retained is from the first non-0 element to the last non-0 element. Then the optimized data is subjected to the Hilbert transform to obtain the instantaneous frequency and amplitude values. In the process of finding the envelope, the amplitude value is mainly used. Finally, the retained data is subjected to the sliding window median filter operation to obtain the final envelope result.

[0051] The extraction of individual features in the frequency domain is carried out through accurate signal carrier frequency estimation, and carrier frequency difference analysis is performed on radar signals with the same system and equivalent signal strength. It mainly includes two algorithms: the dual-line amplitude method and the single-line phase method.

[0052] The individual features in the time-frequency domain are extracted by using the wavelet packet transform algorithm (see formula 1 for details) to extract the time-frequency localized area information, and the fuzzy function algorithm (see formula 2 for details) is used to obtain the fuzzy function slices based on different frequency offset parameter settings. The finite number of intrinsic mode function (IMF) components obtained after decomposing the multi-component signal using the classical mode decomposition (EMD) algorithm are used to reconstruct the time-frequency distribution of the radiation source signal to extract the time-frequency diagram as the classification feature.

[0053]

[0054]

[0055] Among them, h is a low-pass filter, τ is the time delay, and ξ is the frequency difference.

[0056] Individual feature extraction of cyclic spectrum, extracting the transformed plane slice intercepted after cyclic spectrum transformation as feature vector.

[0057] The individual features of the cyclic bispectrum are extracted by using the cyclic bispectrum function (see Formula 3 for details) to extract the diagonal slices of the cyclic bispectrum, and singular value decomposition (SVD, see Formula 4 for details) is performed at the same time, and the extracted singular values ​​or singular vectors are used as individual features of the radiation source.

[0058]

[0059]

[0060] in, is the third-order cumulant of the cyclostationary signal; U is an m×p orthogonal matrix, V is a p×n orthogonal matrix, and Z is a diagonal matrix. The non-zero diagonal elements of the matrix σ 1 ≥σ 2 ≥…≥σ p >0 is called the non-zero singular value of the matrix A, u 1 ,u 2 ,…,up and v 1 ,v 2 ,…,v p They are called the left and right singular vectors of matrix A respectively.

[0061] Step 4: For signal feature visualization, a set of feature visualization display tools is developed using the MATLAB graphical user interface (GUI) software platform.

[0062] The visualization design takes into account both original signal viewing and feature viewing.

[0063] In the original signal visualization page, such as Figure 6 As shown, the specific functional architecture design is as follows:

[0064] (1) File selection function, which supports users to select and read the corresponding raw data file;

[0065] (2) Data basic information display function, which provides the corresponding information of the original data file, including file size, data length, acquisition time, and file storage location;

[0066] (3) Original signal display function, which can intuitively display the original signal data in time domain and frequency domain through drawing;

[0067] (4) Visualization data point range configuration function. Since the length of the original signal is often relatively long, directly displaying all the data will cause the displayed data to be compressed in the time dimension, which will not produce a good display effect. Therefore, the visualization data point range configuration function is set to select the data segment that needs to be visualized.

[0068] In the signal feature visualization page, such as Figure 7 As shown, the specific functional architecture design is as follows:

[0069] (1) Feature selection function: select the features to be visualized, such as: visualize envelope, spectrum, power spectrum, fuzzy function;

[0070] (2) Feature visualization function, which visualizes the features selected in the feature selection module in a graphical way;

[0071] (3) Parameter configuration function: Since some features actually have some configurable parameters, some parameters of the selected features can be set in this module.

[0072] like Figure 8 As shown in the figure, the specific process of realizing signal feature visualization includes:

[0073] First, data reading. Currently, there are two types of data files that can be read in: the first is the original signal data file, with the suffix .bxsldmn, and the second is the file after pulse extraction and parameter measurement, with the suffix .pls. The file is saved in the format of the pulse saved in step 2.

[0074] The second is data viewing. For the original signal data bxsldmn, after reading it in, it needs to undergo pulse detection and extract the pulse signal before subsequent feature viewing can be performed. For the pls file, it saves the data after pulse detection and extraction, so there is no need to go through the pulse detection step and you can directly enter the pulse signal viewing page.

[0075] After the above process, enter the feature result view. In the feature signal visualization window, the interface content includes ten feature information: pulse signal envelope, spectrum, power spectrum, fuzzy function, SVD singular value, cyclic spectrum, wavelet packet decomposition, time-frequency reconstruction, third-order cumulant, and fourth-order cumulant.

[0076] The interface also sets up save configuration options, and selects the output directory to save the feature information that needs to be saved.

[0077] The present invention is not limited to the above-mentioned specific implementation methods. For ordinary technicians in the technical field to which the present application belongs, several simple deductions or substitutions can be made without departing from the inventive concept of the present application.

[0078] The parts not described in detail in the above content are prior art, so they are not described in detail.

Claims

1. A radar emitter individual identification method based on multi-domain feature joint judgment, characterized by: The steps include: Step 1: Signal preprocessing includes: noise reduction, time alignment and normalization, and multipath detection; Step 2: Original signal detection and parameter measurement, including original signal analysis, pulse and parameter extraction, and pulse storage; Step 3: Signal feature extraction domain analysis, including feature extraction and joint analysis of multiple feature quantities in time domain, frequency domain, time-frequency domain, cyclic spectrum and cyclic bispectrum; Step 4: Signal feature visualization, by developing a set of feature visualization display tools with the help of MATLAB graphical user interface software platform.

2. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 1 is characterized in that: In the step 1), the denoising adopts a wavelet threshold denoising algorithm, so as to process the measured radar radiation source signal and realize signal-noise separation in the wavelet domain.

3. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 2 is characterized in that: The specific steps of time alignment and normalization in step 1) are as follows: sort the intercepted pulses according to the signal-to-noise ratio, select the pulse with the largest signal-to-noise ratio as the first reference pulse, and for the kth input pulse, calculate its cross-correlation value with each aligned pulse; The pulses are aligned according to the time delay corresponding to the maximum cross-correlation value; The pulse data is normalized to limit the data within a certain range, and the signal amplitude is normalized to 1 using the variance normalization method.

4. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 3 is characterized in that: In the step 1), the multipath detection is performed by using the measured signal multipath detection based on the Wigner-Ville distribution; firstly, the signal is intercepted by using a sliding window, and a threshold value is pre-set so that the signal in the final sliding window can extract the tail effective detection data containing less noise; secondly, the Wigner-Ville distribution of the tail of the extracted signal data is calculated, the detection parameters are extracted, and the optimal detection threshold is set.

5. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 1 is characterized in that: In the process of analyzing the original signal in step 2), the amplitude and time are analyzed from the overall perspective of the received signal. After a pulse signal is intercepted from the original signal, the amplitude and flatness of the single pulse are analyzed. The main characteristics of the target radar signal are established as a theoretical basis for subsequent pulse detection and extraction, and a large amount of data is used to analyze the applicability of these characteristics.

6. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 5 is characterized by: In the step 2), during the pulse and parameter extraction process, a pulse detection processing technology is used to detect a specific type of pulse signal by setting a threshold; during the pulse extraction process, the original signal is first subjected to wavelet noise reduction processing; Secondly, select a threshold value; finally, monitor the amplitude of the signal, and trigger pulse detection when the signal exceeds or drops to the set threshold; During the parameter extraction process, the pulse arrival time, number, start position, end position, and physical coordinates of the pulse point are extracted; first, the average value and threshold of the signal are calculated, and then the pulse point is filtered according to the threshold; Secondly, the detected pulse arrival time, starting position and ending position are saved in an array, and the pulse amplitude and pulse width are calculated; finally, the pulse data is sorted, short pulses are removed, and the relevant information of the pulse is stored, including pulse sequence number, starting time, amplitude, pulse width, arrival time, etc.

7. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 1 is characterized in that: In step 3), for the extraction of individual features in the time domain, the Hilbert transform method is used to extract the different transient information presented by the radar signal envelope in time. First, the input data is preprocessed to exclude the 0-value elements on both sides, and the valid data retained is from the first non-0 element to the last non-0 element; then the optimized data is subjected to the Hilbert transform to obtain the instantaneous frequency and amplitude values. In the process of finding the envelope, the amplitude value is used; finally, the retained data is subjected to the sliding window median filter operation to obtain the final envelope result; Extract individual features in the frequency domain and perform carrier frequency difference analysis on radar signals with the same system and equivalent signal strength through accurate signal carrier frequency estimation; For the extraction of individual features in the time-frequency domain, the time-frequency localized area information is extracted by wavelet packet transform algorithm, the fuzzy function algorithm is used to obtain fuzzy function slices based on different frequency deviation parameter settings, and the finite number of inherent mode function components obtained after decomposing the multi-component signal by the classical mode decomposition algorithm are used to reconstruct the time-frequency distribution of the radiation source signal, so as to extract the time-frequency diagram as the classification feature; Extract individual features of the cyclic spectrum, and extract the transformed plane slices intercepted after the cyclic spectrum transformation as feature vectors; For the individual feature extraction of the cyclic bispectrum, the diagonal slices of the cyclic bispectrum are extracted using the cyclic bispectrum function, and the singular value decomposition is performed at the same time, and the extracted singular values ​​or singular vectors are used as the individual features of the radiation source.

8. The radar emitter individual identification method based on multi-domain feature joint judgment according to claim 1 is characterized in that: The specific process of realizing signal feature visualization includes: First, data reading. Currently, there are two types of data files that can be read in: the first is the original signal data file, with the suffix .bxsldmn; the second is the file after pulse extraction and parameter measurement, with the suffix .pls. The file is saved in the same format as the pulse saved in step 2); The second is data viewing. For the original signal data bxsldmn, after reading in, it needs to go through pulse detection and extract the pulse signal before subsequent feature viewing can be performed. For the pls file, it saves the data after pulse detection and extraction, so there is no need to go through the pulse detection step, and you can directly enter the pulse signal viewing page; After the above process, enter the feature result view. In the feature signal visualization window, the interface content includes pulse signal envelope, spectrum, power spectrum, fuzzy function, SVD singular value, cyclic spectrum, wavelet packet decomposition, time-frequency reconstruction, third-order cumulant, fourth-order cumulant and other feature information; The interface also sets up save configuration options, and selects the output directory to save the feature information that needs to be saved.

Images