CN107301381A - Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy - Google Patents
Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy Download PDFInfo
- Publication number
- CN107301381A CN107301381A CN201710404829.1A CN201710404829A CN107301381A CN 107301381 A CN107301381 A CN 107301381A CN 201710404829 A CN201710404829 A CN 201710404829A CN 107301381 A CN107301381 A CN 107301381A
- Authority
- CN
- China
- Prior art keywords
- mrow
- layer
- signal
- radar
- radar emitter
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/021—Auxiliary means for detecting or identifying radar signals or the like, e.g. radar jamming signals
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
Abstract
The invention discloses a kind of Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy, the problem of prior art recognition accuracy is low is mainly solved.Implementation step is：1. original radar emitter signal is subjected to data prediction；2. a pair pretreated radar emitter signal extracts envelope characteristic, the slice feature of ambiguity function feature 0, circulation spectrum signature and spectrum signature, and the value linear transformation of these features is saved as into image set to [0,255]；3. designing convolutional neural networks CNN, and introduce multi-task learning in CNN and inactivate strategy at random；4. respectively with four features training collection training convolutional neural networks CNN, recycling four convolutional neural networks CNN models trained to classify respectively to four characteristic test collection, recognizing radar radiation source result is exported.Recognition accuracy of the present invention is high, is scouted and threat radar warning system available for electronic intelligence reconnaissance, electronic support.
Description
Technical field
The present invention relates to signal processing technology field, more particularly to a kind of Recognition Method of Radar Emitters, available for electronics
Intelligence reconnaissance, electronic support are scouted and threat radar warning system.
Background technology
Radar emitter signal is, by feature extraction, its personal feature to be characterized by convolutional neural networks, in order to prevent
Network over-fitting, introduces multi-task learning and random inactivation strategy, so that unique recognize individual radiation source exactly.
Recognizing radar radiation source is electronic intelligence reconnaissance ELINT, electronic support scouts ESM and threat radar alerts RWR systems
Crucial processing procedure in system, is also premise and the basis of electronic interferences, and its identification level is to weigh radar countermeasure set technology
The important symbol of advanced degree.With the continuous improvement developed rapidly with Modern Electronic Countermeasure level of Radar Technology, radar spoke
Penetrate identifing source system and be faced with new challenge.The complexity of emitter Signals is one of wherein topmost challenge, is embodied in three
Individual aspect：One is that the ever-increasing covering frequence of radar emitter signal causes unknown radar signal type and quantity increasingly
It is many；Two be that, with the raising of Radar Technology level, large amount of complex radar starts to occur, the radar emitter signal shape of generation
Formula is complicated, and frequency is changeable；Three be due to the continuous broadening working frequency range of Radar emitter and increasingly complicated working system, difference
Mutual is overlapping in the frequency range and time domain of Radar emitter.
By receiving the signal that unknown Radar emitter is launched, its personal feature is analyzed, so as to uniquely identify individual spoke
Source is penetrated, recognizing radar radiation source process substantially divides three steps：
(1) radar signal is pre-processed.For example：Data filtering noise reduction, multipath signal differentiation, pulse normalization and alignment of data
Deng；
(2) signature analysis, extraction and optimization.For example：The sign of fingerprint characteristic and extraction, characteristic vector construction, feature are excellent
Change, feature database foundation and renewal etc.；
(3) classifier design method.For example：It is a variety of suitable for classifier design of engineer applied etc.；
However, when using conventional method to extract feature to radar emitter signal complicated and changeable, recycling SVM, ELM etc.
When classical recognizer is identified, same rear end recognizer is obvious to the performance difference of different tagsorts.Cause
The main cause of this problem is that complicated radar standard adds Radar emitter value volume and range of product, and then causes to distinguish between its class
Other property feature is difficult to extract, and traditional recognition method is gradually lost validity.
Radar emitter signal recognition methods is quickly grown in recent years, including the instantaneous autocorrelation haracter of construction signal
Levy, time-frequency atom sets up complete atom, construct low, high frequency detail wavelet coefficient Energy distribution entropy and blind source separating suppresses to intersect
The methods such as item.Wherein:
Based on signal transient autocorrelation method, it extracts the instantaneous frequency feature of signal first, and instantaneous frequency is carried out
Normalized is cascaded, characteristic of division vector is extracted, classification, the feature that this method is extracted finally are realized using Hierarchical Decision Making method
Vector has separation property between preferable class, and has the advantages that calculating speed is fast, is easy to Project Realization.But this method is in signal to noise ratio
When relatively low, recognition effect is relatively low.
Complete atom method is set up based on time-frequency atom, it set up complete atom by time-frequency atom method first
Storehouse, and radar emitter signal is made into Its Sparse Decomposition in atom, finally by match tracing decomposition result, to extract signal
In carrier frequency.But this method amount of calculation is larger.
Based on wavelet structure coefficient Energy distribution entropy method, after it is first by wavelet transformation, and small echo is approached with low frequency
The high frequency detail wavelet coefficient energy point of the Energy distribution entropy of coefficient and the reflection signal edge after the related denoising of yardstick is calculated
Cloth entropy constitutes the two-dimensional feature vector of radar emitter signal together.But this method is poor compared with noise immunity under high s/n ratio.
Based on the method for blind source separating suppressing crossterms, it extracts each independent element by blind source separating first, and through when
The Joint diagonalization suppressing crossterms of frequency distribution matrix, then by each twocomponent signal from item summation reconstruct Wigner-Ville distribution,
Each LFM compositions are recognized using Wigner-Hough.But this method robustness is poor.
The above method can not take into account the problems such as emitter Signals are various informative, frequency is changeable overlapping with time-frequency, model simultaneously
Generalization Capability it is poor, how further to extract trickleer robust features and improve system Generalization Capability turn into radar
The key of Radar recognition.
The content of the invention
For the deficiency of existing radar emitter signal recognition methods, the present invention proposes a kind of based on deep learning and many
The Recognition Method of Radar Emitters of tasking learning strategy, to improve the robustness of feature, more accurately Discrimination Radar radiation source
Signal.
The technical solution adopted in the present invention is：It is special by the envelope characteristic, the ambiguity function that extract radar emitter signal
0 slice feature, circulation spectrum signature and spectrum signature are levied, further being extracted it on this basis using convolutional neural networks CNN is had
The characteristic details of effect, and learning characteristic cataloged procedure, realize accurately identifying for Radar emitter.Implementation step includes as follows：
(1) pretreatment of noise reduction, normalization and alignment of data is carried out successively to original radar emitter signal；
(2) envelope characteristic, the slice feature A of ambiguity function feature 0 are extracted respectively to pretreated radar emitter signalx
(ξ, τ), circulation spectrum signatureWith spectrum signature Sk0；
(3) feature extracted in (2) is subjected to linear transformation between [0,255] respectively, and by the letter after linear transformation
Number envelope characteristic, ambiguity function feature 0 cut into slices, circulates spectrum signature and spectrum signature saves as image set a, b, c and d respectively；
(4) it regard 80% in image set a, b, c and d in step (3) as training set i, j, k and l, remaining 20% conduct
Test set m, n, o and p；
(5) according to the corresponding convolutional neural networks CNN of Data Structure Design of original radar emitter signal；
(6) it is each behind convolutional neural networks CNN two full articulamentums to add one layer Dropout layers, and at last
Longitudinally connected two softmax layers again behind Dropout layers of layer, to prevent network over-fitting；
(7) convolutional neural networks CNN is respectively trained with training set i, j, k and l, obtains four convolutional neural networks CNN moulds
Type e, f, g and h；
(8) test set m, n, o and p are classified respectively using four convolutional neural networks CNN models e, f, g and h, it is defeated
Go out recognizing radar radiation source result.
The present invention has advantages below：
1) present invention is due to the envelope characteristic in radar emitter signal, the slice feature of ambiguity function feature 0, Cyclic Spectrum spy
Seek peace and characteristic details are further extracted on the basis of spectrum signature, can more accurately Discrimination Radar emitter Signals.
2) present invention prevents net due to introducing random inactivation strategy and multi-task learning strategy in convolutional neural networks CNN
Network over-fitting, in addition multi-task learning strategy both extracted the architectural feature of data, learnt association between different pieces of information again special
Levy, therefore can be in the case where Radar emitter data volume is few, Discrimination Radar emitter Signals exactly.
Brief description of the drawings
Fig. 1 is the implementation process figure of the present invention；
Fig. 2 is the Threshold Denoising algorithm principle figure used in the present invention；
Fig. 3 is multi-task learning schematic diagram in the present invention；
Fig. 4 is the Dropout schematic diagrams used in the present invention；
Fig. 5 is the convolutional neural networks CNN used in the present invention network structure.
Embodiment
Technical scheme and effect are further described below in conjunction with accompanying drawing：
Reference picture 1, step is as follows for of the invention realizing：
Step 1, data prediction is carried out to radar emitter signal.
Radar emitter signal is often subject to the shadow of the factors such as environment, receiving device in propagation, collection and transfer process
Ring, signal is disturbed that situation is more serious, it is radar emitter signal identification that noise reduction is carried out to obtained radar emitter signal
An important step.Secondly as wavelet transformation has low entropy, multiresolution, decorrelation and the basic function spy such as flexibly
Point, realizes that the method for SNR estimation and compensation is obtained a wide range of applications, finally, in order that after noise reduction using wavelet transformation in wavelet field
The energy of radar emitter signal is on the same order of magnitude, it is necessary to carry out normalizing to it before feature extraction and pattern-recognition
Change and registration process, implement step as follows：
1.1) using Threshold Denoising algorithm to original radar emitter signal noise reduction：
The schematic diagram of Threshold Denoising algorithm shown in reference picture 2, this step is implemented as follows：
First, wavelet basis and number of transitions are determined, and according to wavelet basis and number of transitions to original radar emitter signal
Wavelet transformation is carried out, wavelet transform signal is obtained；
Secondly, threshold filter is carried out to wavelet transform signal, obtains threshold filter signal；
Finally, the radar emitter signal after inverse wavelet transform, output noise reduction is carried out to threshold filter signal；
1.2) radar emitter signal after noise reduction is normalized：
Due to the change of ambient noise and direction of arrival degree, radar receiver is when scouting reception signal, even phase
The adjacent radar emitter signal received, its energy is also differed, and signal to noise ratio differs greatly.In order that radar emitter signal
Energy is on the same order of magnitude, it is necessary to which the carrier frequency of the radar emitter signal after noise reduction is become before feature extraction and identification
Same frequency is changed to, and is zero by the frequency setting, the envelope characteristic of pretreated radar emitter signal is obtained.
1.3) alignment of data processing is carried out to the radar emitter signal after normalization
Alignment of data is carried out to radar emitter signal using pulse correlation, carried out as follows：
1.3a) choose reference pulse；
1.3b) calculate the cross correlation value c with reference pulsej(τ)；
1.3c) according to cross correlation value cj(τ) estimates time delay：
1.3d) according to time delay τkTo the signal alignment processing after noise reduction, the signal after registration process is obtained：
Step 2, feature extraction is carried out to the radar emitter signal after data prediction.
Radar personal feature is primarily present in the parameters such as the amplitude of radar emitter signal, frequency, phase, passes through analysis
The personal feature of radar signal should integrate time domain, ambiguity function 0 cut into slices, each feature such as Cyclic Spectrum, frequency spectrum, and excellent in design
Grader, it is possible to achieve effective identification of radar emitter signal, step is as follows：
2.1) envelope characteristic of the radar emitter signal after noise reduction is extracted：
It is different because the change on the front and rear edge of pulse envelope, pulsewidth, envelope peak position are all relevant with specific circuit
Pulse-modulator has different parameters, causes modulation every time to have stable and unique pulse envelope feature, these features can
For Discrimination Radar emitter Signals, therefore signal envelope is the driving feature of Discrimination Radar emitter Signals, after noise reduction
The carrier frequency of radar emitter signal transform to 0, can obtain the envelope characteristic of pretreated radar emitter signal.
2.2) slice feature of ambiguity function feature 0 of the radar emitter signal after noise reduction is extracted：
Due to using ambiguity function " section of nearly zero " frequency deviation remains thunder as the characteristic features of radar emitter signal
Up to the stability discrimination property feature of emitter Signals, therefore, the section of ambiguity function feature 0 can effectively describe Radar emitter
Signal, section A of the ambiguity function at any frequency deviation ξ is calculated according to equation belowx(ξ,τ)：
Ax(ξ, τ)=∫ X*(f)X*(f-ξ)e-j2πfτdf
In formula, X*(f) it is the Fourier transform of signal；F is signal frequency；τ is time delay；ξ is Doppler frequency.,
The computational methods of above-mentioned slice of ambiguity function are a kind of common methods in signal processing technology field, are seen：" it is based on
Calculation formula (20) in the Radar emitter individual identification of ambiguity function "；
2.3) the circulation spectrum signature of the radar emitter signal after noise reduction is extracted：
Due to carrying out that the influence of noise and interference can be completely inhibited in analysis theories to signal in circulation spectral domain, and for many
Signal environment, as long as each signal cycle frequency is not overlapping, it is possible to be respectively processed, extracts corresponding signal characteristic parameter,
Therefore the Cyclic Spectrum of radar emitter signal is the validity feature of Discrimination Radar emitter Signals, is implemented as follows：
For a zero-mean random signal x (t), circulation spectrum signature is extracted as follows
Wherein, f is signal frequency, and τ is time delay；CoefficientRepresent the circulation auto-correlation intensity that frequency is α：
In formula, x (t) is signal, and τ is time delay, and T is the signal period, and t is the time, and j represents plural number；
The computational methods of above-mentioned Cyclic Spectrum are a kind of common methods in signal processing technology field, are seen：It is " flat based on circulation
Calculation formula (2-3) in the Radar emitter feature extraction surely analyzed and fusion recognition "；
2.4) spectrum signature of radar emitter signal is extracted：
For be all with the relative frequency deviation and absolute frequency deviation of different frequency sources, signal carrier frequency it is different, because
This, can be implemented as follows using the otherness of carrier frequency feature with Discrimination Radar emitter Signals：
Spectrum signature is extracted as follows
In formula, x (t) is signal, fcFor signal carrier frequency, n is time point sequence number, and N is the signal period, and j represents plural number.
Step 3, the radar emitter signal feature to extraction carries out linear transformation.
For training convolutional neural networks, it is necessary to which each feature of the radar emitter signal of extraction is saved as into image,
Therefore need to carry out it linear transformation, this step is implemented as follows：
First, it is the envelope characteristic extracted in step 2, the slice feature of ambiguity function feature 0, circulation spectrum signature and frequency spectrum is special
The value levied distinguishes linear transformation between [0,255]；
Secondly, the matrix of each feature is readjusted as two-dimensional matrix, interpolation method is linear interpolation；
Finally, the radar emitter signal envelope characteristic after linear transformation, ambiguity function feature 0 are cut into slices, Cyclic Spectrum it is special
Spectrum signature of seeking peace saves as image set a, b, c and d respectively.
Step 4, construction radar emitter signal training set and test set.
Using 80% in image set a, b, c and d as convolutional neural networks CNN training set i, j, k and l, remaining 20%
It is used as test set m, n, o and p.
Step 5, design convolutional neural networks CNN.
Because the data volume of Radar emitter is fewer, a certain degree of over-fitting occurs in the training process,
In order to prevent network over-fitting, multi-task learning and Dropout strategies are introduced in the convolutional neural networks CNN of design, it sets
Count step as follows：
5.1) multi-task learning：
Multi-task learning is a kind of conclusion migration mechanism, and main target is to utilize the training number for lying in multiple inter-related tasks
Specific area information in improves generalization ability, while a task is learnt, passes through shared convolutional neural networks CNN
Model is obtained to be contacted with inter-related task, the schematic diagram of multi-task learning shown in reference picture 3, and implementing for this step is the
Two longitudinally connected two softmax layers after random deactivating layer Dropout2 layers；
5.2) random inactivation：
Dropout refers in the training process of deep learning network, for neutral net unit, according to certain probability
It is temporarily abandoned from network, Dropout schematic diagrams shown in reference picture 4, it is in convolutional neural networks CNN to implement
It is random that the first random deactivating layer Dropout1 and second is separately added into behind the first complete full articulamentum fc2 of articulamentum fc1 and second
Dropout2 layers of deactivating layer；
5.3) planned network structure：
Convolutional neural networks CNN structure charts shown in reference picture 5, this step is implemented as follows：
Structure design data Layer data 5.3a) is constituted according to radar emitter signal image set；
The first convolutional layer conv1 5.3b) is added after data Layer data, the convolution kernel number of the convolutional layer is 32, convolution
Core size is 3*3, and step-length is 1；
The first pond layer pooling1 5.3c) is added behind the first convolutional layer conv1, the down-sampling factor of pond layer is 2*
2, step-length is 1；
5.3d) the first correcting layer relu1 is added behind the layer pooling1 of the first pond；
The second convolutional layer conv2 5.3e) is added behind the first correcting layer relu1, the convolution kernel number of the convolutional layer is
32, convolution kernel size is 3*3, and step-length is 1；
The second correcting layer relu2 5.3f) is added behind the second convolutional layer conv2；
The second pond layer pooling2 5.3g) is added behind the second correcting layer relu2, the down-sampling factor of pond layer is 2*
2, step-length is 1；
The 3rd convolutional layer conv3 5.3h) is added behind the layer pooling2 of the second pond, the convolution kernel number of the convolutional layer is
64, convolution kernel size is 3*3, and step-length is 1；
The 3rd pond layer pooling3 5.3i) is added behind the 3rd convolutional layer conv3, the down-sampling factor of pond layer is 2*
2, step-length is 1；
5.3j) the 3rd correcting layer relu3 is added behind the layer pooling3 of the 3rd pond；
The first complete full articulamentum fc2 of articulamentum fc1 and second 5.3k) are sequentially added behind the 3rd correcting layer relu3, its
Neuron number is respectively 64 and 128.
5.3l) because training sample number is less, in order to prevent network over-fitting, the first full connection in step (k)
The first random random deactivating layer of deactivating layer Dropout1 and second is separately added into behind the full articulamentum fc2 of layer fc1 and second
Dropout2 layers, and longitudinally connected two softmax layers after second random deactivating layer Dropout2 layers.
Step 6, training convolutional neural networks CNN.
Signal envelope feature, feature 0 are cut into slices, spectrum signature is circulated and frequency domain character is separately input to convolutional neural networks
CNN, and it is 5000,10000,15000 to set gradually iterations, respectively obtains 4 of the convolutional neural networks CNN trained
Model A, B, C and D.
Step 7, the convolutional neural networks CNN models trained are used to classify to radar emitter signal feature.
This step is implemented as follows：
1) the first test set m is classified using convolutional neural networks CNN the first model A, exports convolutional Neural net
Recognition accuracies of the network CNN to radar emitter signal envelope characteristic；
2) the second test set n is classified using convolutional neural networks CNN the second Model B, exports convolutional Neural net
The recognition accuracy that network CNN cuts into slices to radar emitter signal ambiguity function feature 0；
3) the 3rd test set o is classified using convolutional neural networks CNN the 3rd MODEL C, exports convolutional Neural net
Network CNN circulates the recognition accuracy of spectrum signature to radar emitter signal；
4) the 4th test set p is classified using convolutional neural networks CNN the 4th model D, exports convolutional Neural net
Recognition accuracies of the network CNN to radar emitter signal spectrum signature.
The effect of the present invention can be further illustrated by following experiment：
1) experiment condition：
Experimental situation：Intel Core i7CPU 2.00Ghz, 16GB internal memories, Linux system, deep learning framework is
Caffe, GPU are GTX1070.
2) experiment content and result：
Experiment 1：CNN is tested to different Radar emitter feature recognition performances, as a result such as table 1.
Recognition correct rates of the CNN of table 1 to radar emitter signal different characteristic
In table 1When expression iterations is 5000, convolutional neural networks CNN is not converged.
As seen from Table 1, when iterations is 10000, obtained convolutional neural networks CNN models are believed Radar emitter
Number envelope characteristic, the slice feature of ambiguity function feature 0, circulation spectrum signature and spectrum signature recognition correct rate are without iterations
For 15000 when obtained convolutional neural networks CNN models it is high to the recognition accuracy of each feature, from convolutional neural networks CNN
Spectrum signature recognition accuracy highests of the CNN to Radar emitter is can be seen that to the recognition accuracy of each feature, add with
Convolutional neural networks CNN after machine inactivation and multi-task learning strategy cuts into slices to signal envelope, ambiguity function feature 0, Cyclic Spectrum
There is lifting with the recognition accuracy of frequency spectrum.
Experiment 2：With the present invention with existing method support vector machines and extreme learning machine ELM to each spy of Radar emitter
That levies is identified, its result such as table 2.
Recognition correct rate of the distinct methods of table 2 to radar emitter signal feature
As seen from Table 2, the signal envelope of Radar emitter and circulation spectrum signature can preferably describe Radar emitter letter
Number, in signal envelope, Cyclic Spectrum and spectrum signature, the present invention is to its recognition accuracy highest, and ELM is to ambiguity function feature 0
The recognition accuracy highest of section, it should be noted that ELM algorithms are relatively unstable to the recognition accuracy of feature.
To sum up, the present invention passes through convolutional neural networks CNN Discrimination Radar emitter Signals envelope characteristic, ambiguity function feature
0 slice feature, circulation spectrum signature and spectrum signature, and the random inactivation of introducing and multi-task learning strategy prevent network over-fitting,
The convolutional neural networks CNN models obtained by training, being capable of Discrimination Radar emitter Signals exactly.
Claims (4)
1. a kind of Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy, it is characterised in that including：
(1) pretreatment of noise reduction, normalization and alignment of data is carried out successively to original radar emitter signal；
(2) envelope characteristic, the slice feature A of ambiguity function feature 0 are extracted respectively to pretreated radar emitter signalx(ξ,
τ), spectrum signature is circulatedAnd spectrum signature
(3) feature extracted in (2) is subjected to linear transformation between [0,255] respectively, and by the signal bag after linear transformation
Network feature, ambiguity function feature 0 cut into slices, circulates spectrum signature and spectrum signature saves as image set a, b, c and d respectively；
(4) it regard 80% in image set a, b, c and d in step (3) as training set i, j, k and l, remaining 20% conduct test
Collect m, n, o and p；
(5) according to the corresponding convolutional neural networks CNN of Data Structure Design of original radar emitter signal；
(6) it is each behind convolutional neural networks CNN two full articulamentums to add one layer Dropout layers, and in last layer
Longitudinally connected two softmax layers again behind Dropout layers, to prevent network over-fitting；
(7) convolutional neural networks CNN is respectively trained with training set i, j, k and l, obtain four convolutional neural networks CNN models e,
F, g and h；
(8) test set m, n, o and p are classified respectively using four convolutional neural networks CNN models e, f, g and h, exports thunder
Up to Radar recognition result.
2. the method according to claim 1, wherein the step (1), is carried out as follows：
(1a) determines wavelet basis and number of transitions；
(1b) carries out wavelet transformation according to wavelet basis and number of transitions to original radar emitter signal, obtains wavelet transformation letter
Number；
Wavelet transform signal is carried out threshold filter by (1c), obtains threshold filter signal；
(1d) carries out the radar emitter signal after inverse wavelet transform, output noise reduction to threshold filter signal；
(1e) chooses reference pulse according to the radar emitter signal after noise reduction；
(1f) calculates the cross correlation value c with reference pulsej(τ)；
(1g) is according to cross correlation value cj(τ) estimates time delay：
(1h) is according to time delay τkTo the signal alignment processing after noise reduction, the radar emitter signal after registration process is obtained：
The carrier frequency of radar emitter signal after registration process is transformed to same frequency by (1i), and by the frequency setting
It is zero, obtains the envelope characteristic of pretreated radar emitter signal.
3. the method according to claim 1, wherein the step (2), is carried out as follows：
(2a) extracts the slice feature A of ambiguity function feature 0 as followsx(ξ,τ)：
Ax(ξ, τ)=∫ X*(f)X*(f-ξ)e-j2πfτdf
Wherein, X*(f) it is the Fourier transform of signal；F is frequency, and τ is time delay, and ξ is Doppler frequency；
(2b) extracts circulation spectrum signature as follows
<mrow>
<msubsup>
<mi>S</mi>
<mi>x</mi>
<mi>&alpha;</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>f</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msubsup>
<mo>&Integral;</mo>
<mrow>
<mo>-</mo>
<mi>&infin;</mi>
</mrow>
<mi>&infin;</mi>
</msubsup>
<msubsup>
<mi>R</mi>
<mi>x</mi>
<mi>&alpha;</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>&tau;</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>j</mi>
<mn>2</mn>
<mi>&pi;</mi>
<mi>f</mi>
<mi>&tau;</mi>
</mrow>
</msup>
<mi>d</mi>
<mi>&tau;</mi>
</mrow>
Wherein, f is signal frequency, and τ is time delay；CoefficientRepresent the circulation auto-correlation intensity that frequency is α：
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msubsup>
<mi>R</mi>
<mi>x</mi>
<mi>&alpha;</mi>
</msubsup>
<mrow>
<mo>(</mo>
<mi>&tau;</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munder>
<mrow>
<mi>l</mi>
<mi>i</mi>
<mi>m</mi>
</mrow>
<mrow>
<mi>T</mi>
<mo>&RightArrow;</mo>
<mi>&infin;</mi>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<msubsup>
<mo>&Integral;</mo>
<mrow>
<mo>-</mo>
<msub>
<mi>T</mi>
<mo>\</mo>
</msub>
<mo>/</mo>
<mn>2</mn>
</mrow>
<mrow>
<mi>T</mi>
<mo>/</mo>
<mn>2</mn>
</mrow>
</msubsup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>x</mi>
<mo>*</mo>
</msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>-</mo>
<mi>&tau;</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>j</mi>
<mn>2</mn>
<mi>&pi;</mi>
<mi>&alpha;</mi>
<mi>t</mi>
</mrow>
</msup>
<mi>d</mi>
<mi>t</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mo><</mo>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>+</mo>
<mi>&tau;</mi>
<mo>/</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
<msup>
<mi>x</mi>
<mo>*</mo>
</msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>-</mo>
<mi>&tau;</mi>
<mo>/</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>j</mi>
<mn>2</mn>
<mi>&pi;</mi>
<mi>&alpha;</mi>
<mi>t</mi>
</mrow>
</msup>
<msub>
<mo>></mo>
<mi>t</mi>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
In formula, τ is time delay, and T is the signal period；
(2c) extracts spectrum signature as follows
<mrow>
<msub>
<mi>S</mi>
<msub>
<mi>k</mi>
<mn>0</mn>
</msub>
</msub>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>n</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>N</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>j</mi>
<mn>2</mn>
<mi>&pi;</mi>
<mi>n</mi>
<mi>k</mi>
<mo>/</mo>
<mi>N</mi>
</mrow>
</msup>
<mo>,</mo>
</mrow>
In formula, x (t) is signal, fcFor signal carrier frequency, n is time point sequence number, and N is the signal period.
4. according to the data structure of original radar emitter signal in the method according to claim 1, wherein step (5)
Corresponding convolutional neural networks CNN is designed, is carried out as follows：
(5a) constitutes structure design data Layer data according to radar emitter signal image set；
(5b) adds the first convolutional layer conv1 after data Layer data, and the convolution kernel number of the convolutional layer is 32, convolution kernel size
For 3*3, step-length is 1；
(5c) adds the first pond layer pooling1 behind the first convolutional layer conv1, and the down-sampling factor of pond layer is 2*2, step
A length of 1；
(5d) adds the first correcting layer relu1 behind the layer pooling1 of the first pond；
(5e) adds the second convolutional layer conv2 behind the first correcting layer relu1, and the convolution kernel number of the convolutional layer is 32, volume
Product core size is 3*3, and step-length is 1；
(5f) adds the second correcting layer relu2 behind the second convolutional layer conv2；
(5g) adds the second pond layer pooling2 behind the second correcting layer relu2, and the down-sampling factor of pond layer is 2*2, step
A length of 1；
(5h) adds the 3rd convolutional layer conv3 behind the layer pooling2 of the second pond, and the convolution kernel number of the convolutional layer is 64, volume
Product core size is 3*3, and step-length is 1；
(5i) adds the 3rd pond layer pooling3 behind the 3rd convolutional layer conv3, and the down-sampling factor of pond layer is 2*2, step
A length of 1；
(5j) adds the 3rd correcting layer relu3 behind the layer pooling3 of the 3rd pond；
(5k) sequentially adds the first complete full articulamentum fc2 of articulamentum fc1 and second behind the 3rd correcting layer relu3, its nerve
First number is respectively 64 and 128；
(5l) is separately added into the first random deactivating layer Dropout1 behind the first complete full articulamentum fc2 of articulamentum fc1 and second
With second random deactivating layer Dropout2 layers, and longitudinally connected two softmax after second random deactivating layer Dropout2 layers
Layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710404829.1A CN107301381A (en) | 2017-06-01 | 2017-06-01 | Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710404829.1A CN107301381A (en) | 2017-06-01 | 2017-06-01 | Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107301381A true CN107301381A (en) | 2017-10-27 |
Family
ID=60137280
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710404829.1A Pending CN107301381A (en) | 2017-06-01 | 2017-06-01 | Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107301381A (en) |
Cited By (44)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832787A (en) * | 2017-10-31 | 2018-03-23 | 杭州电子科技大学 | Recognition Method of Radar Emitters based on bispectrum own coding feature |
| CN108090412A (en) * | 2017-11-17 | 2018-05-29 | 西北工业大学 | A kind of radar emission source category recognition methods based on deep learning |
| CN108197146A (en) * | 2017-11-29 | 2018-06-22 | 山东航天电子技术研究所 | The essence classification in-orbit generation system of Radar recognition parameter based on pulse flow data |
| CN108229404A (en) * | 2018-01-09 | 2018-06-29 | 东南大学 | A kind of radar echo signal target identification method based on deep learning |
| CN108303682A (en) * | 2018-01-17 | 2018-07-20 | 电子科技大学 | A kind of passive MIMO radar external sort algorithm selection method based on relative entropy criterion |
| CN108416290A (en) * | 2018-03-06 | 2018-08-17 | 中国船舶重工集团公司第七二四研究所 | Radar signal feature method based on residual error deep learning |
| CN109410143A (en) * | 2018-10-31 | 2019-03-01 | 泰康保险集团股份有限公司 | Image enchancing method, device, electronic equipment and computer-readable medium |
| CN109444831A (en) * | 2018-09-06 | 2019-03-08 | 中国人民解放军战略支援部队航天工程大学 | A kind of radar chaff decision-making technique based on transfer learning |
| CN109471074A (en) * | 2018-11-09 | 2019-03-15 | 西安电子科技大学 | Recognition Method of Radar Emitters based on singular value decomposition Yu one-dimensional CNN network |
| CN109597043A (en) * | 2018-11-16 | 2019-04-09 | 江苏科技大学 | Radar Signal Recognition method based on quantum particle swarm convolutional neural networks |
| CN109614930A (en) * | 2018-12-11 | 2019-04-12 | 湖南华诺星空电子技术有限公司 | A kind of unmanned plane spectrum detection method based on deep learning |
| CN109818892A (en) * | 2019-01-18 | 2019-05-28 | 华中科技大学 | Construct Cyclic Spectrum characteristic parameter extraction model and signal modulation mode recognition methods |
| CN109919015A (en) * | 2019-01-28 | 2019-06-21 | 东南大学 | A kind of radio-frequency fingerprint extraction and recognition methods based on more sampling convolutional neural networks |
| CN109978028A (en) * | 2019-03-12 | 2019-07-05 | 北京理工大学 | A kind of Specific Emitter Identification method based on integrated deep learning |
| CN110109060A (en) * | 2019-03-27 | 2019-08-09 | 西安电子科技大学 | A kind of radar emitter signal method for separating and system based on deep learning network |
| CN110146855A (en) * | 2019-06-11 | 2019-08-20 | 北京无线电测量研究所 | Radar Intermittent AF panel thresholding calculation method and device |
| CN110163130A (en) * | 2019-05-08 | 2019-08-23 | 清华大学 | A kind of random forest grader and classification method of the feature pre-align for gesture identification |
| CN110991378A (en) * | 2019-12-11 | 2020-04-10 | 中国人民解放军空军工程大学 | Power amplifier individual identification method and device |
| CN110988803A (en) * | 2019-11-11 | 2020-04-10 | 浙江大学 | Radar radiation source individual identification system with high confidence degree dynamic adjustment |
| CN111025100A (en) * | 2019-12-10 | 2020-04-17 | 中国电力科学研究院有限公司 | Transformer ultrahigh frequency partial discharge signal mode identification method and device |
| CN111060878A (en) * | 2019-12-30 | 2020-04-24 | 南京东科优信网络安全技术研究院有限公司 | LFM radar working mode real-time classification method and device suitable for single pulse |
| CN111126134A (en) * | 2019-11-11 | 2020-05-08 | 浙江大学 | Radar radiation source deep learning identification method based on non-fingerprint signal eliminator |
| CN111239692A (en) * | 2020-01-10 | 2020-06-05 | 西安电子科技大学 | PRI (pulse repetition index) combined intra-pulse information radiation source signal identification method based on deep learning |
| CN111310719A (en) * | 2020-03-10 | 2020-06-19 | 电子科技大学 | Unknown radiation source individual identification and detection method |
| CN111310680A (en) * | 2020-02-24 | 2020-06-19 | 电子科技大学 | Radiation source individual identification method based on deep learning |
| CN111368648A (en) * | 2020-02-17 | 2020-07-03 | 西安电子科技大学 | Radar radiation source individual identification method and device, electronic equipment and storage medium thereof |
| CN111582320A (en) * | 2020-04-17 | 2020-08-25 | 电子科技大学 | Dynamic individual identification method based on semi-supervised learning |
| CN111582373A (en) * | 2020-05-09 | 2020-08-25 | 电子科技大学 | Radiation source identification method based on weighted migration extreme learning machine algorithm |
| CN111680737A (en) * | 2020-06-03 | 2020-09-18 | 西安电子科技大学 | Radar radiation source individual identification method under differential signal-to-noise ratio condition |
| CN111767848A (en) * | 2020-06-29 | 2020-10-13 | 哈尔滨工程大学 | Radiation source individual identification method based on multi-domain feature fusion |
| CN111913156A (en) * | 2020-08-05 | 2020-11-10 | 电子科技大学 | Radar radiation source individual identification method based on deep learning model and feature combination |
| CN112034434A (en) * | 2020-09-04 | 2020-12-04 | 中国船舶重工集团公司第七二四研究所 | Radar radiation source identification method based on sparse time-frequency detection convolutional neural network |
| CN112068070A (en) * | 2020-07-22 | 2020-12-11 | 中国人民解放军国防科技大学 | Robustness design method based on deep learning direction finding |
| CN112166304A (en) * | 2018-04-03 | 2021-01-01 | 祖克斯有限公司 | Error detection of sensor data |
| CN112464713A (en) * | 2020-10-21 | 2021-03-09 | 安徽农业大学 | Communication radiation source radio frequency fingerprint identification method based on deep learning |
| CN113158886A (en) * | 2021-04-19 | 2021-07-23 | 中国人民解放军63892部队 | Waveform agility radar radiation source identification method based on deep reinforcement learning |
| CN113283527A (en) * | 2021-06-07 | 2021-08-20 | 哈尔滨工程大学 | Radar threat assessment method based on level indexes |
| CN113343802A (en) * | 2021-05-26 | 2021-09-03 | 电子科技大学 | Multi-wavelet-based radio frequency fingerprint image domain identification method |
| CN113359091A (en) * | 2021-05-28 | 2021-09-07 | 太原理工大学 | Deep learning-based multi-kernel function aliasing radar radiation source identification method and system |
| CN113406589A (en) * | 2021-05-22 | 2021-09-17 | 西北工业大学 | Multi-target action identification method based on SIMO Doppler radar |
| CN113687326A (en) * | 2021-07-13 | 2021-11-23 | 广州杰赛科技股份有限公司 | Vehicle-mounted radar echo noise reduction method, device, equipment and medium |
| CN113899948A (en) * | 2021-12-08 | 2022-01-07 | 成都中星世通电子科技有限公司 | System and method for quickly extracting electromagnetic spectrum target characteristic data |
| CN116819477A (en) * | 2023-06-29 | 2023-09-29 | 中国船舶集团有限公司第七二三研究所 | Inter-pulse fingerprint feature combined radar radiation source individual identification method |
| CN117216519A (en) * | 2023-09-12 | 2023-12-12 | 中国电子科技集团公司第三十研究所 | Time-frequency aliasing signal identification method and system based on cyclic spectrum characteristics |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7830297B1 (en) * | 2007-12-20 | 2010-11-09 | Itt Manufacturing Enterprises, Inc. | System and method for detecting and de-interleaving radar emitters using snapshots of varying lengths |
| CN101893704A (en) * | 2010-07-20 | 2010-11-24 | 哈尔滨工业大学 | Rough set-based radar radiation source signal identification method |
| CN102279390A (en) * | 2011-05-06 | 2011-12-14 | 西南交通大学 | Intra-pulse modulation and recognition method of low signal-to-noise radar radiation source signal |
| US20120280848A1 (en) * | 2011-05-03 | 2012-11-08 | Src, Inc. | Method, system and program product for deinterleaving and classifying arbitrary radar pulse patterns using non-determinsitic finite state automata |
| CN103412287A (en) * | 2013-09-01 | 2013-11-27 | 西安电子科技大学 | Linear frequency modulation signal parameter evaluation method based on LVD (Lv's distribution) |
| CN103500359A (en) * | 2013-10-18 | 2014-01-08 | 无锡科技职业学院 | Radar radiation source identification method based on structure equivalence type fuzzy neural network |
| CN103968933A (en) * | 2014-04-09 | 2014-08-06 | 西安电子科技大学 | Fuzzy domain characteristics based optical fiber vibration signal identifying method |
| CN104459668A (en) * | 2014-12-03 | 2015-03-25 | 西安电子科技大学 | Radar target recognition method based on deep learning network |
| CN105426832A (en) * | 2015-11-12 | 2016-03-23 | 上海电机学院 | Communication radar radiation source identification method in presence of unsteady SNR (Signal Noise Ratio) |
| CN106093873A (en) * | 2016-06-04 | 2016-11-09 | 陈昌孝 | A kind of radar emitter signal recognition methods based on multisensor |
| CN106203395A (en) * | 2016-07-26 | 2016-12-07 | 厦门大学 | Face character recognition methods based on the study of the multitask degree of depth |
| CN106529478A (en) * | 2016-11-08 | 2017-03-22 | 中国人民解放军空军工程大学 | Radar radiation source signal identification method according to three-dimensional entropy characteristic |
-
2017
- 2017-06-01 CN CN201710404829.1A patent/CN107301381A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7830297B1 (en) * | 2007-12-20 | 2010-11-09 | Itt Manufacturing Enterprises, Inc. | System and method for detecting and de-interleaving radar emitters using snapshots of varying lengths |
| CN101893704A (en) * | 2010-07-20 | 2010-11-24 | 哈尔滨工业大学 | Rough set-based radar radiation source signal identification method |
| US20120280848A1 (en) * | 2011-05-03 | 2012-11-08 | Src, Inc. | Method, system and program product for deinterleaving and classifying arbitrary radar pulse patterns using non-determinsitic finite state automata |
| CN102279390A (en) * | 2011-05-06 | 2011-12-14 | 西南交通大学 | Intra-pulse modulation and recognition method of low signal-to-noise radar radiation source signal |
| CN103412287A (en) * | 2013-09-01 | 2013-11-27 | 西安电子科技大学 | Linear frequency modulation signal parameter evaluation method based on LVD (Lv's distribution) |
| CN103500359A (en) * | 2013-10-18 | 2014-01-08 | 无锡科技职业学院 | Radar radiation source identification method based on structure equivalence type fuzzy neural network |
| CN103968933A (en) * | 2014-04-09 | 2014-08-06 | 西安电子科技大学 | Fuzzy domain characteristics based optical fiber vibration signal identifying method |
| CN104459668A (en) * | 2014-12-03 | 2015-03-25 | 西安电子科技大学 | Radar target recognition method based on deep learning network |
| CN105426832A (en) * | 2015-11-12 | 2016-03-23 | 上海电机学院 | Communication radar radiation source identification method in presence of unsteady SNR (Signal Noise Ratio) |
| CN106093873A (en) * | 2016-06-04 | 2016-11-09 | 陈昌孝 | A kind of radar emitter signal recognition methods based on multisensor |
| CN106203395A (en) * | 2016-07-26 | 2016-12-07 | 厦门大学 | Face character recognition methods based on the study of the multitask degree of depth |
| CN106529478A (en) * | 2016-11-08 | 2017-03-22 | 中国人民解放军空军工程大学 | Radar radiation source signal identification method according to three-dimensional entropy characteristic |
Non-Patent Citations (4)
| Title |
|---|
| LIN LI 等: "Specific Radar Emitter Recognition Based on Wavelet Packet Transform and Probabilistic SVM", 《PROCEEDINGS OF THE 2009 IEEE INTERNATIONAL CONFERENCE ON INFORMATION AND AUTOMATION》 * |
| 付小利: "并行化深度学习算法及其应用研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑（月刊）》 * |
| 史亚 等: "多核融合框架下的雷达辐射源个体识别", 《电子与信息学报》 * |
| 苏欣 等著: "《Android手机应用网络流量分析与恶意行为检测研究》", 31 October 2016, 长沙：湖南大学出版社 * |
Cited By (71)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832787B (en) * | 2017-10-31 | 2020-09-22 | 杭州电子科技大学 | Radar radiation source identification method based on bispectrum self-coding characteristics |
| CN107832787A (en) * | 2017-10-31 | 2018-03-23 | 杭州电子科技大学 | Recognition Method of Radar Emitters based on bispectrum own coding feature |
| CN108090412A (en) * | 2017-11-17 | 2018-05-29 | 西北工业大学 | A kind of radar emission source category recognition methods based on deep learning |
| CN108197146A (en) * | 2017-11-29 | 2018-06-22 | 山东航天电子技术研究所 | The essence classification in-orbit generation system of Radar recognition parameter based on pulse flow data |
| CN108229404B (en) * | 2018-01-09 | 2022-03-08 | 东南大学 | Radar echo signal target identification method based on deep learning |
| CN108229404A (en) * | 2018-01-09 | 2018-06-29 | 东南大学 | A kind of radar echo signal target identification method based on deep learning |
| CN108303682A (en) * | 2018-01-17 | 2018-07-20 | 电子科技大学 | A kind of passive MIMO radar external sort algorithm selection method based on relative entropy criterion |
| CN108303682B (en) * | 2018-01-17 | 2021-03-30 | 电子科技大学 | Passive MIMO radar external radiation source selection method based on relative entropy criterion |
| CN108416290A (en) * | 2018-03-06 | 2018-08-17 | 中国船舶重工集团公司第七二四研究所 | Radar signal feature method based on residual error deep learning |
| CN112166304A (en) * | 2018-04-03 | 2021-01-01 | 祖克斯有限公司 | Error detection of sensor data |
| CN109444831A (en) * | 2018-09-06 | 2019-03-08 | 中国人民解放军战略支援部队航天工程大学 | A kind of radar chaff decision-making technique based on transfer learning |
| CN109444831B (en) * | 2018-09-06 | 2023-03-10 | 中国人民解放军战略支援部队航天工程大学 | Radar interference decision method based on transfer learning |
| CN109410143A (en) * | 2018-10-31 | 2019-03-01 | 泰康保险集团股份有限公司 | Image enchancing method, device, electronic equipment and computer-readable medium |
| CN109471074A (en) * | 2018-11-09 | 2019-03-15 | 西安电子科技大学 | Recognition Method of Radar Emitters based on singular value decomposition Yu one-dimensional CNN network |
| CN109471074B (en) * | 2018-11-09 | 2023-04-21 | 西安电子科技大学 | Radar radiation source identification method based on singular value decomposition and one-dimensional CNN network |
| CN109597043A (en) * | 2018-11-16 | 2019-04-09 | 江苏科技大学 | Radar Signal Recognition method based on quantum particle swarm convolutional neural networks |
| CN109597043B (en) * | 2018-11-16 | 2023-05-26 | 江苏科技大学 | Radar signal identification method based on quantum particle swarm convolutional neural network |
| CN109614930A (en) * | 2018-12-11 | 2019-04-12 | 湖南华诺星空电子技术有限公司 | A kind of unmanned plane spectrum detection method based on deep learning |
| CN109818892A (en) * | 2019-01-18 | 2019-05-28 | 华中科技大学 | Construct Cyclic Spectrum characteristic parameter extraction model and signal modulation mode recognition methods |
| CN109919015A (en) * | 2019-01-28 | 2019-06-21 | 东南大学 | A kind of radio-frequency fingerprint extraction and recognition methods based on more sampling convolutional neural networks |
| CN109978028A (en) * | 2019-03-12 | 2019-07-05 | 北京理工大学 | A kind of Specific Emitter Identification method based on integrated deep learning |
| CN110109060A (en) * | 2019-03-27 | 2019-08-09 | 西安电子科技大学 | A kind of radar emitter signal method for separating and system based on deep learning network |
| CN110109060B (en) * | 2019-03-27 | 2022-11-22 | 西安电子科技大学 | Radar radiation source signal sorting method and system based on deep learning network |
| CN110163130B (en) * | 2019-05-08 | 2021-05-28 | 清华大学 | Feature pre-alignment random forest classification system and method for gesture recognition |
| CN110163130A (en) * | 2019-05-08 | 2019-08-23 | 清华大学 | A kind of random forest grader and classification method of the feature pre-align for gesture identification |
| CN110146855A (en) * | 2019-06-11 | 2019-08-20 | 北京无线电测量研究所 | Radar Intermittent AF panel thresholding calculation method and device |
| CN110146855B (en) * | 2019-06-11 | 2020-10-23 | 北京无线电测量研究所 | Radar intermittent interference suppression threshold calculation method and device |
| CN111126134B (en) * | 2019-11-11 | 2022-03-22 | 浙江大学 | Radar radiation source deep learning identification method based on non-fingerprint signal eliminator |
| CN111126134A (en) * | 2019-11-11 | 2020-05-08 | 浙江大学 | Radar radiation source deep learning identification method based on non-fingerprint signal eliminator |
| CN110988803A (en) * | 2019-11-11 | 2020-04-10 | 浙江大学 | Radar radiation source individual identification system with high confidence degree dynamic adjustment |
| CN111025100A (en) * | 2019-12-10 | 2020-04-17 | 中国电力科学研究院有限公司 | Transformer ultrahigh frequency partial discharge signal mode identification method and device |
| CN110991378A (en) * | 2019-12-11 | 2020-04-10 | 中国人民解放军空军工程大学 | Power amplifier individual identification method and device |
| CN110991378B (en) * | 2019-12-11 | 2023-06-02 | 中国人民解放军空军工程大学 | Individual identification method and device for power amplifier |
| CN111060878B (en) * | 2019-12-30 | 2021-08-24 | 南京东科优信网络安全技术研究院有限公司 | LFM radar working mode real-time classification method and device suitable for single pulse |
| CN111060878A (en) * | 2019-12-30 | 2020-04-24 | 南京东科优信网络安全技术研究院有限公司 | LFM radar working mode real-time classification method and device suitable for single pulse |
| CN111239692A (en) * | 2020-01-10 | 2020-06-05 | 西安电子科技大学 | PRI (pulse repetition index) combined intra-pulse information radiation source signal identification method based on deep learning |
| CN111239692B (en) * | 2020-01-10 | 2023-01-24 | 西安电子科技大学 | PRI (pulse repetition index) combined intra-pulse information radiation source signal identification method based on deep learning |
| CN111368648A (en) * | 2020-02-17 | 2020-07-03 | 西安电子科技大学 | Radar radiation source individual identification method and device, electronic equipment and storage medium thereof |
| CN111310680A (en) * | 2020-02-24 | 2020-06-19 | 电子科技大学 | Radiation source individual identification method based on deep learning |
| CN111310680B (en) * | 2020-02-24 | 2022-11-04 | 电子科技大学 | Radiation source individual identification method based on deep learning |
| CN111310719A (en) * | 2020-03-10 | 2020-06-19 | 电子科技大学 | Unknown radiation source individual identification and detection method |
| CN111582320A (en) * | 2020-04-17 | 2020-08-25 | 电子科技大学 | Dynamic individual identification method based on semi-supervised learning |
| CN111582320B (en) * | 2020-04-17 | 2022-10-14 | 电子科技大学 | Dynamic individual identification method based on semi-supervised learning |
| CN111582373A (en) * | 2020-05-09 | 2020-08-25 | 电子科技大学 | Radiation source identification method based on weighted migration extreme learning machine algorithm |
| CN111680737B (en) * | 2020-06-03 | 2023-03-24 | 西安电子科技大学 | Radar radiation source individual identification method under differential signal-to-noise ratio condition |
| CN111680737A (en) * | 2020-06-03 | 2020-09-18 | 西安电子科技大学 | Radar radiation source individual identification method under differential signal-to-noise ratio condition |
| CN111767848A (en) * | 2020-06-29 | 2020-10-13 | 哈尔滨工程大学 | Radiation source individual identification method based on multi-domain feature fusion |
| CN111767848B (en) * | 2020-06-29 | 2022-04-12 | 哈尔滨工程大学 | Radiation source individual identification method based on multi-domain feature fusion |
| CN112068070A (en) * | 2020-07-22 | 2020-12-11 | 中国人民解放军国防科技大学 | Robustness design method based on deep learning direction finding |
| CN112068070B (en) * | 2020-07-22 | 2022-08-09 | 中国人民解放军国防科技大学 | Robustness design method based on deep learning direction finding |
| CN111913156A (en) * | 2020-08-05 | 2020-11-10 | 电子科技大学 | Radar radiation source individual identification method based on deep learning model and feature combination |
| CN112034434A (en) * | 2020-09-04 | 2020-12-04 | 中国船舶重工集团公司第七二四研究所 | Radar radiation source identification method based on sparse time-frequency detection convolutional neural network |
| CN112034434B (en) * | 2020-09-04 | 2022-05-20 | 中国船舶重工集团公司第七二四研究所 | Radar radiation source identification method based on sparse time-frequency detection convolutional neural network |
| CN112464713A (en) * | 2020-10-21 | 2021-03-09 | 安徽农业大学 | Communication radiation source radio frequency fingerprint identification method based on deep learning |
| CN112464713B (en) * | 2020-10-21 | 2022-11-08 | 安徽农业大学 | Communication radiation source radio frequency fingerprint identification method based on deep learning |
| CN113158886B (en) * | 2021-04-19 | 2023-04-18 | 中国人民解放军63892部队 | Waveform agility radar radiation source identification method based on deep reinforcement learning |
| CN113158886A (en) * | 2021-04-19 | 2021-07-23 | 中国人民解放军63892部队 | Waveform agility radar radiation source identification method based on deep reinforcement learning |
| CN113406589A (en) * | 2021-05-22 | 2021-09-17 | 西北工业大学 | Multi-target action identification method based on SIMO Doppler radar |
| CN113343802B (en) * | 2021-05-26 | 2022-05-03 | 电子科技大学 | Multi-wavelet-based radio frequency fingerprint image domain identification method |
| CN113343802A (en) * | 2021-05-26 | 2021-09-03 | 电子科技大学 | Multi-wavelet-based radio frequency fingerprint image domain identification method |
| CN113359091A (en) * | 2021-05-28 | 2021-09-07 | 太原理工大学 | Deep learning-based multi-kernel function aliasing radar radiation source identification method and system |
| CN113359091B (en) * | 2021-05-28 | 2022-05-17 | 太原理工大学 | Deep learning-based multi-kernel function aliasing radar radiation source identification method and system |
| CN113283527A (en) * | 2021-06-07 | 2021-08-20 | 哈尔滨工程大学 | Radar threat assessment method based on level indexes |
| CN113283527B (en) * | 2021-06-07 | 2022-04-29 | 哈尔滨工程大学 | Radar threat assessment method based on level indexes |
| CN113687326A (en) * | 2021-07-13 | 2021-11-23 | 广州杰赛科技股份有限公司 | Vehicle-mounted radar echo noise reduction method, device, equipment and medium |
| CN113687326B (en) * | 2021-07-13 | 2024-01-05 | 广州杰赛科技股份有限公司 | Vehicle-mounted radar echo noise reduction method, device, equipment and medium |
| CN113899948A (en) * | 2021-12-08 | 2022-01-07 | 成都中星世通电子科技有限公司 | System and method for quickly extracting electromagnetic spectrum target characteristic data |
| CN116819477A (en) * | 2023-06-29 | 2023-09-29 | 中国船舶集团有限公司第七二三研究所 | Inter-pulse fingerprint feature combined radar radiation source individual identification method |
| CN116819477B (en) * | 2023-06-29 | 2024-02-13 | 中国船舶集团有限公司第七二三研究所 | Inter-pulse fingerprint feature combined radar radiation source individual identification method |
| CN117216519A (en) * | 2023-09-12 | 2023-12-12 | 中国电子科技集团公司第三十研究所 | Time-frequency aliasing signal identification method and system based on cyclic spectrum characteristics |
| CN117216519B (en) * | 2023-09-12 | 2024-04-16 | 中国电子科技集团公司第三十研究所 | Time-frequency aliasing signal identification method and system based on cyclic spectrum characteristics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107301381A (en) | Recognition Method of Radar Emitters based on deep learning and multi-task learning strategy | |
| Wang et al. | Deep learning-based UAV detection in pulse-Doppler radar | |
| CN102279390B (en) | Intra-pulse modulation and recognition method of low signal-to-noise radar radiation source signal | |
| CN108416290A (en) | Radar signal feature method based on residual error deep learning | |
| CN110197209B (en) | Radiation source identification method based on multi-feature fusion | |
| Liu et al. | LPI radar signal detection based on radial integration of Choi-Williams time-frequency image | |
| CN109829509B (en) | Radar gesture recognition method based on fusion neural network | |
| Liu et al. | Deep learning and recognition of radar jamming based on CNN | |
| CN113156391B (en) | Radar signal multi-dimensional feature intelligent sorting method | |
| CN103093244B (en) | A kind of Radar Signal Recognition method based on Its Sparse Decomposition | |
| CN102831436B (en) | High spectrum image classifying method based on reference diagram mutual information wave band selection and EMD (Empirical Mode Decomposition) | |
| CN104237883A (en) | Airborne radar space time self-adaptation processing method with sparse representation | |
| Huynh-The et al. | Accurate LPI radar waveform recognition with CWD-TFA for deep convolutional network | |
| CN108052863A (en) | Electrical energy power quality disturbance recognition methods based on the maximum variance method of development | |
| Wang et al. | Automatic modulation classification based on joint feature map and convolutional neural network | |
| CN104268510A (en) | SAR image target recognition method based on non-negative matrix factorization of sparse constraint | |
| Li et al. | Automatic modulation classification based on bispectrum and CNN | |
| CN105160353A (en) | Polarimetric SAR data ground object classification method based on multiple feature sets | |
| CN108090462A (en) | A kind of Emitter Fingerprint feature extracting method based on box counting dimension | |
| Nuhoglu et al. | Image segmentation for radar signal deinterleaving using deep learning | |
| Sun et al. | GPS Spoofing Detection Based on Decision Fusion with a K-out-of-N Rule. | |
| CN113191185A (en) | Method for classifying targets of unmanned aerial vehicle by radar detection through Dense2Net | |
| CN104408472A (en) | Wishart and SVM (support vector machine)-based polarimetric SAR (synthetic aperture radar) image classification method | |
| Zhang et al. | Learning modulation filter networks for weak signal detection in noise | |
| Cunxiang et al. | Quasi-LFM radar waveform recognition based on fractional Fourier transform and time-frequency analysis |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171027 |