CN115563465A - Characteristic coding method for individual identification of radiation source - Google Patents
Characteristic coding method for individual identification of radiation source Download PDFInfo
- Publication number
- CN115563465A CN115563465A CN202211155942.8A CN202211155942A CN115563465A CN 115563465 A CN115563465 A CN 115563465A CN 202211155942 A CN202211155942 A CN 202211155942A CN 115563465 A CN115563465 A CN 115563465A
- Authority
- CN
- China
- Prior art keywords
- time
- radiation source
- signals
- sequence
- frequency
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a characteristic coding method facing radiation source individual identification, which is used for coding IQ two-path sampled radiation source electromagnetic signals based on gram angle and/or difference field transformation. And respectively carrying out the Graham angle and/or difference field transformation on the real part, the imaginary part and the amplitude characteristic sequence of the signal to obtain a time domain and frequency domain correlation characteristic matrix containing sampling points. The obtained coding features can be used as deep neural network input, and radiation source individual recognition is achieved through training the obtained model. The invention provides an idea for extracting time domain and frequency domain characteristics of various electromagnetic signals such as radar and communication, and provides a solution for data preprocessing for individual identification tasks of various radiation sources. Compared with the traditional coding method, the method has better characteristic extraction effect on IQ two-path collected signals of the radiation source, has the characteristics of high identification precision and wide signal application range, and has important significance on individual identification application of the radiation source such as electronic countermeasure.
Description
Technical Field
The invention belongs to the technical field of signal processing, and particularly relates to a characteristic coding method for individual identification of a radiation source.
Background
The electromagnetic signal identification aims at acquiring useful information from the transmitted electromagnetic signal, providing support for applications such as situation estimation and the like, and needs to perform characteristic analysis on the captured signal to acquire various characteristics of the signal, further acquire signal fingerprint characteristics of a radiation source and perform a radiation source target individual identification task.
The general electronic signal feature extraction such as simple features of modulation parameters such as carrier frequency, bandwidth, code element rate and the like is difficult to identify the individual radiation source, and cannot meet the requirements of modern application. With the progress of computer technology, fusion algorithm, electronic information science and other technologies, signal feature analysis and feature extraction technology are rapidly developed, so that information is fully mined and utilized. The individual subtle characteristics of the radiation source target can show the unique attributes of the target, and the fingerprint characteristic extraction and application technology has great research and application values.
From the existing radiation source signal feature extraction and individual identification methods, the feature extraction method of the system is less, and most of the methods do not perform effective coding preprocessing on the signal before feature extraction. A radiation source IQ signal individual identification method based on a deep residual error neural network is proposed by an incremental radiation source individual identification method based on a knowledge distillation mechanism (CN 114492745A) and a radiation source individual identification method based on a deep residual error shrinkage network (CN 114091545A), but a oscillogram of IQ two-path signals is directly applied, the requirement on waveform resolution is high during feature extraction, and preliminary feature extraction on the radiation source signals is lacked; a radiation source individual identification method under a small sample scene (CN 114492604A) provides that signal time-frequency characteristics extracted based on different time-frequency analysis methods are respectively input into a neural network for training, after a radiation source signal is preliminarily analyzed, the characteristics are extracted by using the neural network, but a suitable scene is limited by conditions such as the number of samples and the length of the radiation source signal, and the scene adaptability is weak.
Disclosure of Invention
In order to solve the problems, the invention discloses a characteristic coding method facing to individual identification of a radiation source, which provides an idea for extracting time domain and frequency domain characteristics of various electromagnetic signals such as radar, communication and the like and provides a solution for data preprocessing for individual identification tasks facing to various radiation sources. Compared with the traditional coding method, the method has better characteristic extraction effect on IQ two-path collected signals of the radiation source, has the characteristics of high identification precision and wide signal application range, and has important significance on individual identification application of the radiation source such as electronic countermeasure.
In order to achieve the purpose, the technical scheme of the invention is as follows:
(1) And performing form conversion on the IQ two-path sampled radiation source signals. Acquiring an IQ sampling I path one-dimensional signal as Re, a Q path one-dimensional signal as Im and the amplitude (square sum and root) of the IQ two paths of signals as Am; for Re, im and Am, common time-frequency analysis modes such as short-time Fourier transform, wavelet transform, hilbert-Huang transform and the like can be utilized. And when short-time Fourier transform and wavelet transform are carried out, calculating the power spectral density of the signal at different moments, and obtaining the maximum value of the frequency distribution of the two-dimensional transform result in each time window to obtain a one-dimensional time-frequency characteristic sequence. When Hilbert-Huang transformation is carried out, the content modal component decomposed by the empirical mode decomposition method is subjected to Hilbert transformation, and the instantaneous frequency and the instantaneous amplitude of the signal, namely the decomposed time-frequency characteristic, are finally obtained. The number of the obtained one-dimensional characteristic sequences is equal to the number of empirical mode decomposition times.
In this step, a time-frequency analysis method needs to be specifically screened according to the analysis requirement of the signal and the characteristics of the signal itself.
(2) And (3) performing Graham angle field transformation (GASF/GADF transformation) on the time domain, time domain and frequency domain characteristic results obtained in the step (1). And converting the original time sequence and time-frequency domain characteristic sequence with the length of n into an n multiplied by n characteristic coding matrix.
(3) And (3) cascading I paths, Q paths, amplitude values and time-frequency domain feature coding matrixes of the radiation source IQ signals in the step (2), selecting k features with high discrimination to form a matrix array with the size of kXnxnxn, and applying the matrix array to an individual identification task based on a deep neural network.
The concatenated feature matrix is selected specifically for the different signals received. And the signals with a large number of sampling points select more time-frequency domain characteristics for cascading, and otherwise, cascade time-domain characteristics. When time domain and time-frequency domain features with different feature sequence lengths need to be cascaded simultaneously, the size of the coding matrix with the largest feature dimension is selected as a reference, and zero filling operation is carried out on the smaller coding matrix to enable the size of the coding matrix to be the same as the reference. Under the condition that the number of samples is enough, the features with highest discrimination and best recognition effect are selected as the coding mode of the final signal as the features required by individual recognition through the combination of the transformation feature cascade.
As a specification and improvement of the present invention, the step (1) provides a solution for limiting the range and under different signal conditions at the sampling point, and improves the range of the electromagnetic signal to which the present invention is applied.
The applied radiation source IQ signal is digital signal data which is discrete in amplitude and time. The data is divided into two paths, overload wave modulation is respectively carried out on the two paths of data, and the two paths of carriers are orthogonal to each other. The two paths of signals have the same frequency and 90-degree phase difference, and the two paths of signals I and Q are simultaneously transmitted after being respectively modulated; the applied radiation source IQ signal has certain requirements on the number of sampling points. The method adopts the gram angle and/or difference field transformation, the time sequence signal with the number of sampling points being n obtains the matrix with the size of n multiplied by n after the coding pretreatment of the method, so the number of the sampling points of the required radiation source IQ signal is below 1E3 order of magnitude. For radiation source IQ signals with long sampling time and the number of sampling points higher than 1E3, the peak characteristics of the signals can be further obtained by an envelope detection method, and the characteristics and the processing method are applied to subsequent processing. In addition, the output length can be matched with the number order of magnitude of the sampling points in the step by directly using a time-frequency domain analysis method and controlling the size of a window. For the sequence data with the length of n, the converted n × n matrix can adopt the segmentation aggregation approximation to firstly reduce the sequence length and then carry out conversion. Segmenting the sequence, and then compressing the subsequences in each segment into a value by averaging; the applied radiation source IQ signal has no requirements on absolute time parameters such as sampling rate, and the method only considers the relative time relationship between different sampling points of the radiation source IQ signal during processing.
As an improvement of the present invention, the time-frequency domain analysis method in step (1) needs to perform time-frequency-amplitude-time-frequency processing on the analysis result to match the format requirement of the glam angle field transformation and the visualization requirement in step (2). The specific treatment method comprises the following steps:
performing time-frequency analysis by using short-time Fourier transform, wavelet transform and an optimization method thereof in the step (1) to obtain Sub>A three-dimensional matrix result with Sub>A format of T-F-A, and performing modulus on the obtained complex result to obtain frequency-amplitude characteristics at different sampling points; and for each different sampling point, taking the frequency with the highest amplitude at the point as the frequency characteristic of the point, or obtaining the average frequency characteristic of the sampling point in a weighting mode.
When time-frequency analysis is performed by using empirical mode decomposition, hilbert transform and an optimization method thereof in the step (1), the difference of the decomposable times caused by different characteristics of the same type of signals needs to be considered. Therefore, the signal decomposition times with the minimum decomposition times in all the signals needing encoding preprocessing are taken as the uniform decomposition times. And after the transformed result is obtained, all the multi-level decomposition results of the single signal are used as a time-frequency characteristic of the signal, and the next step is carried out.
As an improvement of the present invention, in the radiation source individual identification method based on feature coding preprocessing in step (2), when the sum-difference field transform and the difference field transform of the gram angle are used, the specific processing method is as follows:
scaling the data range obtained by the analysis in the step (1) to [ -1,1]Or [0,1](ii) a And converting the zoomed sequence data into a polar coordinate system, namely, taking the numerical value as a cosine value of an included angle and taking the time stamp as a radius. If the data scaling range is [ -1,1]Then the converted angle range is [0, π](ii) a If the zoom range is [0,1 ]]Then the converted angle range is
Then, an inner product-like operation is performed in the rectangular coordinate system, the scaled one-dimensional sequence data is converted from the rectangular coordinate system to the polar coordinate system, and then the time correlations at different time points are identified by considering the angles (sum and difference) between different points.
The obtained feature coded image has the following features:
(1) The coded images are square matrixes which are distributed from the upper left corner to the lower right corner in time sequence
(2) The image is a heat image, and the gray scale represents the characteristic value of the point
(3) Each point of the image comprises amplitude-time related information and frequency-time related information which are contained in the time point of the sequence after time domain and time-frequency domain analysis
(4) The matrix slice of the image matrix also has the characteristics of containing the time sequence distribution from the upper left corner to the lower right corner, and the image is not sparse. The method has the characteristics of strong information fusion and high robustness in an individual identification task applied to a deep residual error neural network, and is verified in an experiment.
As a refinement of the present invention, the transformations performed in step (2) are all transformations of a single result obtained in step (1) for a single signal. Therefore, in order to obtain a complete pre-processing coding result of a single signal, the transformation results need to be cascaded. When time domain and time-frequency domain features with different feature sequence lengths need to be concatenated simultaneously, the size of the coding matrix with the largest feature dimension is selected as a reference, zero padding operation is carried out on the smaller coding matrix, the size of the coding matrix is enabled to be the same as the reference, and a directly used standard format is provided for individual identification by using the deep residual error neural network in the step (3).
The invention has the beneficial effects that:
the coding preprocessing method of the radiation source IQ signal provided by the invention provides an idea for extracting time domain and time-frequency domain characteristics of the radiation source IQ signal and provides a solution for data preprocessing for artificial intelligence engineering facing various signals. Compared with the traditional coding method, the method has better characteristic extraction effect on IQ two-path signals of the radiation source, has the characteristics of good visualization effect, strong interpretability and wide signal application range, and has important significance on individual identification application of the radiation source and modern electronic countermeasure.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of time-frequency analysis after the preliminary decomposition according to the present invention;
FIG. 3 is a feature encoding flow diagram according to the present invention;
FIG. 4 is an Im sequence image after initial decomposition in an example of the invention;
fig. 5 is a Re sequence image after preliminary decomposition in an example of the present invention;
FIG. 6 is an Am sequence image after initial decomposition in an example of the invention;
FIG. 7 is a result image of Im sequence encoding preprocessing after preliminary decomposition in an example of the present invention;
FIG. 8 is a result image of Re sequence coding preprocessing after preliminary decomposition in an example of the present invention;
FIG. 9 is a result image of the pre-processing of the preliminary decomposed Am sequence encoding in an example of the present invention;
FIG. 10 is a graph illustrating the results of an encoding scheme not employed in the present invention;
FIG. 11 is a diagram illustrating the results of an encoding scheme employing the present invention.
Detailed Description
The present invention will be further explained with reference to the accompanying drawings and the following detailed description of the preamble signal for a civil aviation ADS-B signal, which is to be understood as merely illustrative and not restrictive.
As shown in fig. 1, the method for preprocessing the encoding of the radiation source IQ signal according to the present invention can be roughly divided into several steps of preliminary decomposition, time-frequency analysis (optional), feature encoding, and use in artificial intelligence engineering:
primary decomposition: and decomposing two paths of information of the signals, taking a radiation source IQ signal I in-phase as a real part Re, taking Q quadrature as an imaginary part Im, and taking a geometric average value of the two paths of signals of the radiation source IQ as Am. Time series of Re, im and Am were obtained.
And (3) time-frequency analysis: and in order to further obtain the time-frequency domain characteristics of the radiation source IQ signals, the obtained signal time sequence is subjected to time-frequency analysis processing. The method comprises the following steps of calculating signal power spectrums at different moments by applying short-time Fourier transform, and obtaining a one-dimensional time-frequency characteristic sequence by taking the maximum value of frequency distribution of a two-dimensional analysis result in each time window; and (3) obtaining a signal decomposition result and a time-frequency characteristic after decomposition according to the time-domain characteristic of the data by applying empirical mode decomposition and Hilbert transform, and screening the result with proper empirical mode decomposition times. The time-frequency joint analysis method applicable to the step also comprises other common methods such as wavelet transformation and the like.
After time-frequency analysis is carried out by utilizing short-time Fourier transform, wavelet transform and an optimization method thereof, sub>A three-dimensional matrix result with Sub>A format of T-F-A is obtained, and Sub>A module is taken for the obtained complex result to obtain frequency-amplitude characteristics at different sampling points; and for each different sampling point, taking the frequency with the highest amplitude at the point as the frequency characteristic of the point, or obtaining the average frequency characteristic of the sampling point in a weighting mode.
When time-frequency analysis is carried out by using empirical mode decomposition, hilbert transform and an optimization method thereof, different characteristics of similar signals need to be considered, and the difference of the decomposable times is caused. Therefore, the signal decomposition times with the minimum decomposition times in all the signals needing encoding preprocessing are taken as the uniform decomposition times. And after the transformed result is obtained, all the multi-level decomposition results of the single signal are used as a time-frequency characteristic of the signal, and the next step is carried out.
The specific flow is shown in fig. 2.
Preprocessing characteristic codes: and (2) performing Graham angular field transformation (GASF/GADF transformation) on different results obtained by time domain and frequency domain analysis in the step (1) to obtain a characteristic coding matrix.
Scaling the data range obtained by the analysis in the step (1) to [ -1,1]Or [0,1](ii) a And converting the zoomed sequence data into a polar coordinate system, namely, taking the numerical value as a cosine value of an included angle and taking the time stamp as a radius. If the data scaling range is [ -1,1]Then the converted angleIn the range of [0, π](ii) a If the zoom range is [0,1 ]]Then the converted angle range is
Then, an inner product-like operation is performed in the rectangular coordinate system, the scaled one-dimensional sequence data is converted from the rectangular coordinate system to the polar coordinate system, and then the time correlations at different time points are identified by considering the angles (sum and difference) between different points.
The encoding preprocessing flow is shown in fig. 3.
The preliminary decomposition part of the invention decomposes two paths of information of signals, and takes a radiation source IQ signal I in-phase as a real part Re, takes Q quadrature as an imaginary part Im, and takes a geometric mean value of the two paths of signals of the radiation source IQ as Am. And obtaining Im, re and Am time sequences. Fig. 4, 5, and 6 show the Im, re, am results after the initial decomposition of the IQ signal of the exemplary radiation source, respectively.
The time-frequency analysis technical scheme can apply short-time Fourier transform to calculate the signal power spectrum at different moments, and the maximum value of the frequency distribution of the two-dimensional analysis result in each time window is taken to obtain a one-dimensional time-frequency characteristic sequence; and obtaining a signal decomposition result and a time-frequency characteristic after decomposition according to the time-domain characteristic of the data by applying empirical mode decomposition and Hilbert transform, and screening to obtain a result with proper empirical mode decomposition times. The time-frequency joint analysis method applicable to the step also comprises other common methods such as wavelet transformation and the like.
Performing time-frequency analysis by using short-time Fourier transform, wavelet transform and an optimization method thereof in the step (1) to obtain Sub>A three-dimensional matrix result with Sub>A format of T-F-A, and performing modulus on the obtained complex result to obtain frequency-amplitude characteristics at different sampling points; and for each different sampling point, taking the frequency with the highest amplitude at the point as the frequency characteristic of the point, or obtaining the average frequency characteristic of the sampling point in a weighting mode.
When time-frequency analysis is performed by using empirical mode decomposition, hilbert transform and an optimization method thereof in the step (1), the difference of the decomposable times caused by different characteristics of the same type of signals needs to be considered. Therefore, the signal decomposition times with the minimum decomposition times in all the signals needing encoding preprocessing are taken as the uniform decomposition times. And after the transformed result is obtained, the multi-level decomposition results of the single signal are all used as a time-frequency characteristic of the signal, and the next step is carried out.
The characteristic coding preprocessing of the invention comprises the following steps: and (2) carrying out Graham angle and/or difference field transformation on different results obtained by time domain and frequency domain analysis in the step (1) to obtain a characteristic coding matrix. Scaling the data range obtained by the analysis in the step (1) to [ -1,1]Or [0,1](ii) a And converting the scaled sequence data into a polar coordinate system, namely, taking the numerical value as a cosine value of an included angle and taking the time stamp as a radius. If the data scaling range is [ -1,1]Then the converted angle range is [0, π](ii) a If the zoom range is [0,1 ]]Then the converted angle range is
Then, an inner product-like operation in a rectangular coordinate system is performed, the scaled one-dimensional sequence data is converted from the rectangular coordinate system to a polar coordinate system, and then the time correlations at different time points are identified by considering the angles (sum, difference) between the different points.
In the example, the image obtained by performing the encoding preprocessing adopted by the present invention on the Im, re, am sequence obtained by the preliminary decomposition is shown in fig. 7, fig. 8, and fig. 9.
The obtained feature coded image has the following features:
(1) The coded images are square matrixes which are distributed from the upper left corner to the lower right corner in time sequence
(2) The image is a heat image, and the gray level depth represents the value of the point
(3) Each point of the image comprises amplitude-time related information and frequency-time related information which are contained in the time point of the sequence after time domain and time-frequency domain analysis
(4) The matrix slice of the image matrix also has the characteristic of containing the features distributed from the upper left corner to the lower right corner in a time sequence, and the image is not sparse. The method has the characteristics of strong information fusion and high robustness in an individual identification task applied to a deep residual error neural network, and is verified in an experiment.
The individual identification task is to cascade the results of the coding preprocessing, train through a convolutional neural network and obtain the individual identification result of the radiation source. And taking the feature coding preprocessing result as a training sample, sending the training sample into a deep neural network for further feature extraction, and finally obtaining an individual recognition result. After the encoding method of the present invention is not trained, a confusion matrix of the signal classification result is obtained as shown in fig. 10, and the accuracy is 89.7%. After the characteristic coding mode adopted by the invention is utilized, the confusion matrix of the signal classification result is obtained as shown in fig. 11, and the accuracy is 93.2%. The method realizes the improvement of the accuracy rate of the large-amplitude identification under the condition of the characteristic coding preprocessing.
Claims (3)
1. A characteristic coding method for individual identification of radiation sources is characterized by comprising the following steps:
(1) Carrying out form conversion on IQ two-path sampled radiation source signals; acquiring IQ sampling I path one-dimensional signals as Re, Q path one-dimensional signals as Im and amplitude values of IQ two paths of signals as Am; aiming at Re, im and Am, utilizing a short-time Fourier transform (STFT), a Wavelet Transform (WT) and a Hilbert-Huang transform (HHT) common time-frequency analysis mode; when short-time Fourier transform and wavelet transform are carried out, the power spectral density of signals at different moments is calculated, the maximum value of the frequency distribution of a two-dimensional transform result in each time window is taken, and a one-dimensional time-frequency characteristic sequence is obtained; when Hilbert-Huang transformation is carried out, IMF (intrinsic mode function) components decomposed by EMD (empirical mode decomposition) are subjected to Hilbert transformation, and signal instantaneous frequency and instantaneous amplitude are finally obtained, namely the time-frequency characteristics after decomposition; the obtained number of the one-dimensional characteristic sequences is equal to the number of times of empirical mode decomposition; in this step, a time-frequency analysis method needs to be screened in a targeted manner according to the analysis requirements of the signals and the characteristics of the signals;
(2) Performing Graham angle and/or difference field GASF/GADF conversion on the time domain, time domain and frequency domain characteristic results obtained in the step (1); converting the original time sequence with the length of n and the time-frequency domain characteristic sequence into an n multiplied by n characteristic coding matrix;
(3) Cascading I paths, Q paths, amplitude values and time-frequency domain feature coding matrixes of radiation source IQ signals in the step (2), selecting k features with high discrimination to form a matrix array with the size of kXnxn, and applying the matrix array to an individual identification task based on a deep neural network;
for different received signals, the cascaded feature matrix is selected in a targeted manner; more signals with a large number of sampling points select time-frequency domain characteristics to be cascaded, otherwise, time-domain characteristics are cascaded; when time domain and time-frequency domain features with different feature sequence lengths need to be cascaded simultaneously, selecting the size of a coding matrix with the largest feature dimension as a reference, and performing zero filling operation on a smaller coding matrix to enable the size of the coding matrix to be the same as the reference; under the condition that the number of samples is enough, the features with highest discrimination and best recognition effect are selected as the coding mode of the final signal as the features required by individual recognition through the combination of the transformation feature cascade.
2. The feature coding method facing individual identification of radiation sources according to claim 1, characterized in that:
based on the Gelam angular field GASF/GADF conversion, time sequence signals with n sampling points are recorded as a vector V, and V multiplied by V is utilized T Then obtaining a matrix with the size of n multiplied by n, wherein the number of required radiation source IQ signal sampling points is below 1E3 order of magnitude; for radiation source IQ signals with long sampling time and the number of sampling points higher than 1E3, the peak characteristics of the signals are further obtained through an envelope detection method, the obtained sequence length is reduced, and the characteristics and the processing method are applied to subsequent processing; in addition, the magnitude of the number order of the sampling points of the step can be controlled to be matched with the output length by directly analyzing the size of the window and the displacement length of the window by the time-frequency domain analysis method in the step (1); for sequence data with the length of n, adopting segmented aggregation approximation to reduce the length of the sequence and then converting the converted n multiplied by n matrix; segmenting the sequence, and then compressing the subsequences in each segment into a value by averaging; the applied radiation source IQ signal is being sampledThe sample rate absolute time parameter has no requirement, and only the relative time relation between different sampling points of the radiation source IQ signal is considered during processing.
3. The feature encoding method for individual identification of radiation sources according to claim 1, wherein: when the gram angle field transformation is used in the step (2), the following transformation requirements are met:
scaling the data range obtained by the analysis in the step (1) to [ -1,1]Or [0,1](ii) a Converting the scaled sequence data into a polar coordinate system, namely, taking the numerical value as a cosine value of an included angle and taking the time stamp as a radius; if the data scaling range is [ -1,1]Then the converted angle range is [0, π](ii) a If the zoom range is [0,1 ]]Then the converted angle range is
After zooming, an inner product-like operation is performed in a rectangular coordinate system, the zoomed one-dimensional sequence data is converted from the rectangular coordinate system to a polar coordinate system, and then the angle sum or the angle difference is determined by considering the angle sum GASF/angle difference GADF between different points to identify the time correlation at different time points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211155942.8A CN115563465A (en) | 2022-09-22 | 2022-09-22 | Characteristic coding method for individual identification of radiation source |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211155942.8A CN115563465A (en) | 2022-09-22 | 2022-09-22 | Characteristic coding method for individual identification of radiation source |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115563465A true CN115563465A (en) | 2023-01-03 |
Family
ID=84741065
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211155942.8A Pending CN115563465A (en) | 2022-09-22 | 2022-09-22 | Characteristic coding method for individual identification of radiation source |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115563465A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116166954A (en) * | 2023-04-20 | 2023-05-26 | 南京桂瑞得信息科技有限公司 | Radiation source individual identification method based on self-adaptive signal characteristic embedded knowledge graph |
-
2022
- 2022-09-22 CN CN202211155942.8A patent/CN115563465A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116166954A (en) * | 2023-04-20 | 2023-05-26 | 南京桂瑞得信息科技有限公司 | Radiation source individual identification method based on self-adaptive signal characteristic embedded knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111783558A (en) | Satellite navigation interference signal type intelligent identification method and system | |
| Sporring et al. | Information measures in scale-spaces | |
| Wei et al. | PRI modulation recognition based on squeeze-and-excitation networks | |
| CN109633588A (en) | Recognition Method of Radar Emitters based on depth convolutional neural networks | |
| Ghadimi et al. | Deep learning-based approach for low probability of intercept radar signal detection and classification | |
| CN110764067A (en) | Fourier transform LFM-BPSK composite modulation radar signal parameter estimation method | |
| CN110187313B (en) | Radar signal sorting and identifying method and device based on fractional order Fourier transform | |
| CN115563465A (en) | Characteristic coding method for individual identification of radiation source | |
| CN111680737A (en) | Radar radiation source individual identification method under differential signal-to-noise ratio condition | |
| CN113923042A (en) | Malicious software abuse DoH detection and identification system and method | |
| CN116797796A (en) | Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling | |
| Yang et al. | Radar emitter signal recognition based on time-frequency analysis | |
| CN113780521B (en) | Radiation source individual identification method based on deep learning | |
| CN113723244A (en) | Radar radiation source signal separation method based on improved variational modal decomposition | |
| CN114185030A (en) | Radar signal intra-pulse modulation identification method and system | |
| Si et al. | An efficient deep convolutional neural network with features fusion for radar signal recognition | |
| CN116842450A (en) | Depth measurement learning-based radar radiation source individual identification method | |
| CN109991578B (en) | Multi-component radar signal modulation recognition method based on blind compression kernel dictionary learning | |
| CN112462335B (en) | Multifunctional 3D radar transceiver and operation method | |
| CN112069987B (en) | Interference type automatic identification method based on statistical manifold optimization dimension reduction | |
| CN111308426B (en) | Low signal-to-noise ratio periodic frequency modulation signal detection and separation method suitable for single antenna receiver | |
| Tang et al. | Radar Signal Recognition Based on Dual-Channel Model With HOG Feature Extraction | |
| CN114866311B (en) | Radio frequency fingerprint extraction method based on time sequence characterization | |
| CN114755654B (en) | Damaged radar signal restoration method based on image mimicry technology | |
| CN117857270B (en) | Radio communication signal modulation identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |