CN109684995A - Specific Emitter Identification method and device based on depth residual error network - Google Patents
Specific Emitter Identification method and device based on depth residual error network Download PDFInfo
- Publication number
- CN109684995A CN109684995A CN201811575557.2A CN201811575557A CN109684995A CN 109684995 A CN109684995 A CN 109684995A CN 201811575557 A CN201811575557 A CN 201811575557A CN 109684995 A CN109684995 A CN 109684995A
- Authority
- CN
- China
- Prior art keywords
- residual error
- error network
- depth residual
- signal
- 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
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
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Collating Specific Patterns (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to Studies on Emitters ID field, in particular to a kind of Specific Emitter Identification method and device based on depth residual error network, this method includes: carrying out time frequency analysis to signal is received, and converts gray level image for obtained Hilbert time-frequency spectrum;It is input with gray level image, extracts the radio-frequency fingerprint feature of reflection in the picture using depth residual error network, obtain the recognition result of radiation source.The present invention is directed to signal of communication non-stationary, nonlinear characteristic, and using the gray level image of Hilbert time-frequency spectrum as the form of expression of signal, the radio-frequency fingerprint feature of radiation source is extracted using depth residual error network, completes Classification and Identification;Deep learning is applied to signal of communication process field, gives full play to its powerful self-learning capability, artificial understanding limitation is overcome, improves treatment effeciency;And very strong robustness is all had by recognition effect of emulation experiment verifying under the conditions of complex communication system and Complex Channel, there is great importance for the development of emitter Signals Recognition technology.
Description
Technical field
The invention belongs to Studies on Emitters ID field, in particular to a kind of specific source of radiation based on depth residual error network
Recognition methods and device.
Background technique
Specific Emitter Identification (SEI, Specific Emitter Identification), i.e., by extracting radio frequency letter
The fine feature of radiation source individual difference can be embodied on number, to realize the identification to specific objective.Due to radio-frequency fingerprint spy
Sign does not depend on Content of Communication, and is difficult to forge;Therefore, SEI technology wireless network secure and communication reconnaissance fight etc. it is civilian and
Military field has important application value.The core technology of SEI is to find and extract accurately and effectively radio-frequency fingerprint feature.
According to source difference, existing feature can be mainly divided into following two major classes: predetermined (predefined) feature and supposition
(inferred) feature.Predetermined characteristic obtains as its name suggests according to known characteristic mechanisms are pre-defined.This category feature
The understanding for needing that it is certain to have radio-frequency fingerprint feature, feature itself have specific physical significance, should be readily appreciated that.For example,
Brik et al. devises passive radiation device-identifying system (PARADIS, passive radiometric device
Identification system), it is super to the discrimination of 138 wireless devices that 6 kinds of modulation field characteristics are extracted from planisphere
Cross 99%.In addition, power amplifier coefficient and oscillator phase index are also, respectively, used as radio-frequency fingerprint feature, obtain
Good recognition effect.Compared with predetermined characteristic, thus it is speculated that feature does not have specific physical significance, is the mistake in Classification and Identification
Speculated in journey according to individual difference, is mainly obtained using the method for some mathematic(al) manipulations.For example, using Martin Hilb
Spy-Huang method, the statistic composed using a variety of Hilbert identify signal as feature;In addition, to third moment into
The methods of bispectrum that row Fourier transformation obtains and wavelet transformation, which are also used to speculate, obtains radio-frequency fingerprint feature.However, nothing
By be predetermined characteristic or be speculate feature, be all limited to people priori understanding, this has just been doomed the limitation of existing feature.
Specifically, to be primarily limited to people insufficient to the understanding of feature mechanism of production and the form of expression for predetermined characteristic;And speculate special
Sign is then highly dependent on people to the limited analysis and processing method of signal.
In recent years, deep learning obtains numerous breakthrough achievements in the application such as image recognition, man-machine game.It is roused by it
Dance, deep neural network are also gradually generalized to radar, signal of communication process field, and believe in Modulation Identification and specific radar
Number identification in achieve many research achievements.Since the difference between communication radiation source individual is very small, at this stage, depth is utilized
Degree study carries out the research of SEI still in its infancy.Kevin Merchant makees the waveform of time domain complex base band residual signals
For input, 7 ZigBee equipment are carried out using convolutional neural networks (CNN, convolutional neural network)
Identification, discrimination reach 92.29%.It should be pointed out that the network architecture of CNN has stronger specific aim to image and is applicable in
Property, but in the presence of the huge advantage that cannot give full play to CNN depth self study image information, identification effect is affected to a certain extent
Fruit;It selects to extract feature using the compressed bispectrum image of dimensionality reduction as input using CNN and identified, recognition effect is obvious
Better than conventional method.However, the existing feature published by comparing, it is possible to find the recognition effect of bispectrum feature is difficult to always
It is satisfactory, it indicates not being a kind of ideal selection using bispectrum image as the image for receiving signal;In addition, to bispectrum image
Compression dimensionality reduction while removing redundancy, will also result in the loss of Partial Feature details, further affect identification
Effect.
Summary of the invention
For this purpose, the present invention provides a kind of Specific Emitter Identification method and device based on depth residual error network, people is overcome
For the limitation of understanding, treatment effeciency is improved, there is very strong future in engineering applications.
According to design scheme provided by the present invention, a kind of Specific Emitter Identification method based on depth residual error network,
Include following content:
Time frequency analysis is carried out to signal is received, and converts gray level image for obtained Hilbert time-frequency spectrum;
It is input with gray level image, extracts the radio-frequency fingerprint feature of reflection in the picture using depth residual error network, obtain
The recognition result of radiation source.
Above-mentioned, time frequency analysis is carried out to signal is received by Hilbert-Huang transform.
Preferably, time frequency analysis includes following content:
Empirical mode decomposition, the intrinsic mode function and the residual components after decomposition of every layer of acquisition are carried out to signal is received;
Hilbert transform is carried out to intrinsic mode function one by one, obtains the instantaneous amplitude and wink of every layer of intrinsic mode function
When frequency, obtain receive signal Hilbert spectrum.
Above-mentioned, its matrix is obtained according to Hilbert spectrum, according to gray level image quantization digit, obtains the matrix pair
The gray level image answered.
Above-mentioned, depth residual error network model is using the network architecture trained, the network model knot trained
Structure is by being trained study to network model parameter using sample data.
Above-mentioned, by depth residual error network model training learning network mode input and desired residual error mapping is exported,
Realize the Classification and Identification to specific source of radiation.
Above-mentioned, depth residual error network architecture uses M residual unit, and M is the integer greater than 1.
Above-mentioned, depth residual error network model includes input layer, first convolutional layer, maximum pond layer, multiple convolutional layers, is averaged
Pond layer and output layer, wherein multiple convolutional layers include sequentially connected 8 convolutional layers, in each convolutional layer and first convolutional layer
Activation primitive be all made of ReLU activation primitive, output layer uses Softmax function.
A kind of Specific Emitter Identification device based on depth residual error network includes: conversion module and identification module,
In,
Conversion module for carrying out time frequency analysis to reception signal, and converts gray scale for obtained Hilbert time-frequency spectrum
Image;
Identification module extracts the radio frequency of reflection in the picture using depth residual error network for being input with gray level image
Fingerprint characteristic obtains the recognition result of radiation source.
Beneficial effects of the present invention:
The present invention is directed to emitter Signals non-stationary, nonlinear characteristic, converts ash for obtained Hilbert spectrum first
Image is spent, using gray level image as input, the vision of radio-frequency fingerprint feature in the picture is extracted using depth residual error network and embodies,
Complete Classification and Identification;Deep learning is applied to signal of communication field, gives full play to its powerful self-learning capability, is overcome artificial
Recognize limitation, improves treatment effeciency;And it is identified under the conditions of complex communication system and Complex Channel by emulation experiment verifying
Effect has very strong robustness to complex environment, all has great importance for the development of emitter Signals Recognition technology.
Detailed description of the invention:
Fig. 1 is Specific Emitter Identification method flow schematic diagram in embodiment;
Fig. 2 is communication system schematic diagram in embodiment;
Fig. 3 is transmitter Hilbert time-frequency spectrum schematic diagram in embodiment;
Fig. 4 is residual unit basic framework schematic diagram in embodiment;
Fig. 5 is depth residual error schematic network structure in embodiment;
Fig. 6 is Specific Emitter Identification schematic device in embodiment;
Fig. 7 is influence schematic diagram of the residual unit number to recognition performance in embodiment;
Fig. 8 is the recognition performance schematic diagram in embodiment under awgn channel;
Fig. 9 is recognition performance schematic diagram under slow fading channel in embodiment.
Specific embodiment:
To make the object, technical solutions and advantages of the present invention clearer, understand, with reference to the accompanying drawing with technical solution pair
The present invention is described in further detail.
It is shown in Figure 1 in the embodiment of the present invention for emitter Signals non-stationary, nonlinear specific, one kind is provided
Include following content based on the Specific Emitter Identification method of depth residual error network:
S101, time frequency analysis is carried out to reception signal, and converts gray level image for obtained Hilbert time-frequency spectrum;
S102, with gray level image it is input, extracts reflection radio-frequency fingerprint feature in the picture using depth residual error network,
Obtain the recognition result of radiation source.
Specific Emitter Identification belongs to classification problem, and key is to extract accurately and reliably fine feature.Artificial understanding
Limitation often lead to complex characteristic is abstracted and be extracted.In view of the non-stationary feature of signal of communication, and
Radio-frequency fingerprint feature has nonlinear characteristic.In another embodiment of the present invention, selection is adopted in further embodiment of the present invention
Time frequency analysis is carried out to signal is received with Hilbert-Huang transformation.Hilbert-Huang transformation is mainly including empirical modal point
It solves (EMD) and Hilbert converts two processes.In another embodiment of the present invention, it can be obtained firstly, decomposing and receiving signal r (t)
It arrives
Wherein, N is total number of plies that EMD is decomposed, ci(t) intrinsic mode function (IMF) for being i-th layer, e (t) are signal point
Residual components after solution.Residual components e (t) is not considered, to ci(t) Hilbert transformation is carried out one by one, to obtain every layer of IMF
Instantaneous amplitude ai(t) and instantaneous frequency ωi(t), therefore, receiving signal can be further represented as
The time-frequency distributions that formula (2) provides are known as Hilbert time-frequency spectrum, are denoted as H (ω, t).At this point, we obtain
Hilbert time-frequency spectrum matrix H.If Bi,jFor the gray value of (i, j) a pixel, ζ is the digit of the quantization of gray level image, then
The corresponding gray level image of Hilbert time-frequency spectrum matrix H can be expressed as
Wherein,It indicates to be rounded downwards.
Shown in Figure 3, (a), (b) are respectively the Hilbert time-frequency spectrum of two transmitters, wherein (b) figure transmitter is abnormal
Change degree ratio (a) figure it is big, it can be seen from the figure that receive signal main energetic be concentrated mainly on high frequency section (i.e. signal
Carrier frequency), and middle low frequency part then embodies a large amount of local feature, this is also exactly the main difference place of different transmitters.It is right
For power amplifier distortion, the nonlinear degree the high just to have much stronger frequency component and is leaked to middle low frequency part,
The time-frequency distributions of Hilbert spectrum will be more mixed and disorderly, as shown on the right.And left figure medium-high frequency/intermediate frequency/low frequency distributed area is clearly demarcated, then
Illustrate that its power amplifier nonlinear degree is lower.EM2Algorithm is extracted the Energy-Entropy for receiving the Hilbert time-frequency spectrum H (ω, t) of signal
With the mean μ and standard deviation of gray level imageAs radio-frequency fingerprint feature;Further, by extracting smoothness, brightness and the rolling composed
Feature drops.Features described above is all that a variety of different statistics are extracted on Hilbert time-frequency spectrum.It should be noted, however, that
The process for extracting various statistics is also the process of information loss.Different statistics only goes to describe from different sides
The information contained in Hilbert time-frequency spectrum;The type for increasing statistic only fully portrays Hilbert time-frequency spectrum as far as possible
The feature reflected on figure, and its overall picture can not be had a guide look of always.
Deep layer convolutional neural networks can automatically Learning Integration low-dimensional/middle dimension/higher-dimension feature, be hopeful to overcome artificial
The limitation of understanding.Generally, it is considered that network is deeper, training difficulty is bigger, suffers from the puzzlement of gradient disperse problem.In order to
Weaken the puzzlement, uses depth residual error network to extract radio-frequency fingerprint feature in the embodiment of the present invention.Fig. 4 gives a residual error list
The basic framework of member.Assuming that the input of certain section of neural network is x, output is contemplated to be H (x), directly using input pass to output as
When initial results, it is then F (x)=H (x)-x, i.e. residual error that this section of network, which needs the target learnt,.In the embodiment of the present invention, depth
In residual error network model, the convolutional network of each stacking is not directed through to learn an ideal potential mapping, but is passed through
Network goes to learn ideal residual error mapping, to improve the generalization ability of training effectiveness and network.By the Hilbert of formula (3)
Time-frequency spectrum gray level image carries out the identification of specific source of radiation using depth residual error network as input.The network architecture and specific ginseng
Number is as shown in Fig. 5 and table 1.In planned network, multiple residual units are analyzed and used, totally 10 layers of network, need to train altogether
Multiple parameters.It is used in addition to output layer except Softmax function, other layers of activation primitive is all made of ReLU.Using cross entropy
Cost function is measured, and cost function can be optimized using learning rate for 0.05 gradient descent method.Trained batch size every time
It can be 300.
The network parameter of 1 depth residual error network of table
Based on above-mentioned method, the embodiment of the present invention also provides a kind of Specific Emitter Identification based on depth residual error network
Device, it is shown in Figure 6, include: conversion module 101 and identification module 102, wherein
Conversion module 101 for carrying out time frequency analysis to reception signal, and converts obtained Hilbert time-frequency spectrum to
Gray level image;
Identification module 102 utilizes depth residual error network extraction reflection penetrating in the picture for being input with gray level image
Frequency fingerprint characteristic obtains the recognition result of radiation source.
For the validity for further verifying the embodiment of the present invention, below by specific emulation experiment to technical solution of the present invention
It is further explained explanation:
Verifying analysis is carried out to technical solution of the present invention using emulation signal.Radio-frequency fingerprint feature is tended to originate from transmitter
Analog circuit it is imperfect.The non-linear of RF power amplification will lead to amplitude/amplitude compression phenomenon and amplitude/phase transfer characteristic.
For this purpose, portraying the nonlinear distortion of RF power amplification using Taylor series model in further embodiment of the present invention.Assuming that s0
(t) baseband modulation signal, f are indicatedcFor carrier frequency, for inputting the rf modulated signal of power amplifier
The output signal for carrying power amplifier distortion can indicate are as follows:
Wherein, Λ (t) is the system response function of power amplifier, and K is the polynomial order of Taylor, { λ1,λ2,L,λKIt is function
The nonlinear factor put, usual λ1=1.Shown in Figure 2, (a), (b) are respectively the communication system under single-hop and repeater mode.
For single-hop mode, emits signal by channel and directly reach receiver.Consider non-frequency-selective channel, then
Receiving signal can be expressed as
Wherein, α is the fading factor of channel, and ν (t) is white Gaussian noise.For awgn channel, α=1;Rayleigh is believed
Road, α Rayleigh distributed.
For repeater mode, emit signal before arriving at the receiver, needs first to be amplified and forwarded by relaying.With formula
(5) similar, the signal of relay reception can be expressed as
Wherein, α and ν (t) be respectively transmitter to the fading factor of channel between relaying and noise,It indicates
The coefficient of power amplifier of transmitter.After the amplification and forwarding of relaying, the signal for reaching receiver can be expressed as
Wherein, β and υ (t) be respectively be relayed to the fading factor of channel and interchannel noise between receiver,For the coefficient of power amplifier of relaying.Clearly for repeater mode, the reception signal in formula (7) is not only carried respectively
The fingerprint characteristic of transmitter, and it has been mixed into the nonlinear characteristic of relaying.Which increase the difficulties of feature extraction and target identification
Degree.For the gauge signal quality under different channels, average signal-to-noise ratio is defined first
Wherein, E { α2Be channel fading factor-alpha square average value.Particularly, for awgn channel, E { α2}=1.
The condition setting of emulation experiment are as follows: qpsk modulation signal, observation symbol lengths are 100, and character rate is
200Kbps, carrier frequency 420KHz, sample rate 1MHz, base band pulse molding use rolloff-factor for 0.35 raised cosine pulse.
For awgn channel, α=1;For slow fading channel, α Rayleigh distributed.Radiation source number K=5 to be identified.For not
Same transmitter receives signal according to formula (4) and emulates generation, and the setting of power amplifier nonlinear factor is as shown in table 2.In each letter
It makes an uproar than under, each radiation source includes 5000 sample of signal, wherein it randomly selects 3000 samples and is used to train, residue 2000
A sample is for testing.The gray level image quantified using 8bit, dimension of picture size are 300 × 300.
The power amplifier nonlinear factor of 2 transmitter of table
Using NVIDIA Titan Xp GPU training and test depth residual error network in experiment, MATLAB R2015a is carried out
The processing work of all data, the depth residual error network realized on platform TensorFlow 1.3.0 using Python.
Firstly, discussing influence of the residual unit number to recognition performance.In an experiment, using awgn channel and formula (5)
Single-hop mode, residual unit number are set to 2,4,6,8 and 10.Fig. 7 gives the corresponding identification of different residual unit numbers
Effect.From figure it can be found that with residual unit number increase, recognition correct rate is substantially to be gradually increased.Work as residual error
When unit number is 2, recognition correct rate is minimum, this illustrates that shallower network can not completely extract the personal feature in image.
When residual unit number be 8 when, recognition correct rate highest, even better than residual unit number be 10 the case where, this is residual error list
When first number is 10, the difficulty of network training is too big, it is not easy to which the state being optimal, the accuracy of identification receive certain
Fluctuation.In addition, the discrimination between heterogeneous networks is very close, this illustrates residual error when residual error number is more than or equal to 4
The neural network that number is 4 has been enough the needs of competency extraction.It is considered that with the intensification of network, of network parameter
Several and training time gradually increases, and therefore, in subsequent experimental, can be used uniformly the network that residual error number is 4 and be identified
The test and analysis of performance.
Then, by the comparison with conventional method, the recognition performance under additive white Gaussian noise channel is discussed, this hair is investigated
Recognition performance of the technical solution under awgn channel in bright embodiment.Comparison scheme uses EM2Algorithm extracts radio-frequency fingerprint feature,
And utilize support vector machines (SVM, support based on radial base (RBF, radial basis function) kernel function
Vector machine) carry out Classification and Identification.The comparison of the recognition performance under single-hop and relay scene is set forth in Fig. 8.From
It can be found that the recognition correct rate of technical solution is equal in the embodiment of the present invention under the mode that either single-hop still relays in figure
It is substantially better than EM2Algorithm, this is because EM2Algorithm is only extracted Energy-Entropy and single order, second moment from Hilbert time-frequency spectrum
It is identified as feature, hence it is evident that be not enough to cover the fine feature contained in Hilbert time-frequency spectrum comprehensively, this is by artificial
What the limitation of understanding determined.Hilbert time-frequency spectrum is converted into image in technical solution of the present invention, using neural network from complete
Feature is extracted in whole image, efficiently avoids the loss of information, to significantly improve recognition effect.In addition, with tradition
Method is compared, and in neural metwork training, is indicated since network may learn image of the fine feature under different signal-to-noise ratio,
To further improve the robustness to interchannel noise.
Furthermore the recognition performance under slow fading channel is discussed, further investigates technical solution of the present invention under Complex Channel
Recognition performance.The recognition performance under single-hop and relay scene is set forth in Fig. 9, and comparison scheme is EM2Algorithm.From figure
It can be found that and EM2Algorithm is compared, technical solution of the present invention to single-hop and relaying both of which recognition correct rate there are about
30% promotion, this is considerable.In addition, compared with awgn channel in Fig. 8, EM2Algorithm deteriorates under slow fading channel
Obviously, and technical solution of the present invention then still keeps preferable recognition correct rate.Further illustrate technical solution of the present invention to complexity
Channel has stronger robustness.
In the present invention, time frequency analysis is carried out to signal is received using Hilbert-Huang transformation, Hilbert is composed and is converted
For gray level image.Using gray level image as input, the vision of radio-frequency fingerprint feature in the picture is extracted using depth residual error network
It embodies, to complete Classification and Identification.It is indicated using the image of Hilbert spectrum gray level image as signal, effectively by radiation source
Nuance is converted into the visual signature of image, lays a good foundation to extract time-frequency fingerprint characteristic using convolutional neural networks;It will
The method of deep learning is applied to Specific Emitter Identification, and ladder when network training is effectively improved using depth residual error network
Disperse problem is spent, trained efficiency is improved;Recognition effect under complicated communication system and complicated channel condition, experiment
The result shows that technical solution of the present invention has very strong robustness to complex environment.
Unless specifically stated otherwise, the opposite step of the component and step that otherwise illustrate in these embodiments, digital table
It is not limit the scope of the invention up to formula and numerical value.
Based on above-mentioned method, the embodiment of the present invention also provides a kind of server, comprising: one or more processors;It deposits
Storage device, for storing one or more programs, when one or more of programs are executed by one or more of processors,
So that one or more of processors realize above-mentioned method.
Based on above-mentioned method, the embodiment of the present invention also provides a kind of computer-readable medium, is stored thereon with computer
Program, wherein the program realizes above-mentioned method when being executed by processor.
The technical effect and preceding method embodiment phase of device provided by the embodiment of the present invention, realization principle and generation
Together, to briefly describe, Installation practice part does not refer to place, can refer to corresponding contents in preceding method embodiment.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description
It with the specific work process of device, can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
In all examples being illustrated and described herein, any occurrence should be construed as merely illustratively, without
It is as limitation, therefore, other examples of exemplary embodiment can have different values.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.
The flow chart and block diagram in the drawings show the system of multiple embodiments according to the present invention, method and computer journeys
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, section or code of table, a part of the module, section or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually base
Originally it is performed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that
It is the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, can uses and execute rule
The dedicated hardware based system of fixed function or movement is realized, or can use the group of specialized hardware and computer instruction
It closes to realize.
In several embodiments provided herein, it should be understood that disclosed systems, devices and methods, it can be with
It realizes by another way.The apparatus embodiments described above are merely exemplary, for example, the division of the unit,
Only a kind of logical function partition, there may be another division manner in actual implementation, in another example, multiple units or components can
To combine or be desirably integrated into another system, or some features can be ignored or not executed.Another point, it is shown or beg for
The mutual coupling, direct-coupling or communication connection of opinion can be through some communication interfaces, device or unit it is indirect
Coupling or communication connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in the executable non-volatile computer-readable storage medium of a processor.Based on this understanding, of the invention
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the present invention
State all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can be with
Store the medium of program code.
Finally, it should be noted that embodiment described above, only a specific embodiment of the invention, to illustrate the present invention
Technical solution, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to this hair
It is bright to be described in detail, those skilled in the art should understand that: anyone skilled in the art
In the technical scope disclosed by the present invention, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features;And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection of the invention
Within the scope of.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (9)
1. a kind of Specific Emitter Identification method based on depth residual error network, which is characterized in that include following content:
Time frequency analysis is carried out to signal is received, and converts gray level image for obtained Hilbert time-frequency spectrum;
It is input with gray level image, extracts the radio-frequency fingerprint feature of reflection in the picture using depth residual error network, obtain radiation
The recognition result in source.
2. the Specific Emitter Identification method according to claim 1 based on depth residual error network, which is characterized in that pass through
Hilbert-Huang transform carries out time frequency analysis to signal is received.
3. the Specific Emitter Identification method according to claim 2 based on depth residual error network, which is characterized in that time-frequency
Analysis includes following content:
Empirical mode decomposition, the intrinsic mode function and the residual components after decomposition of every layer of acquisition are carried out to signal is received;
Hilbert transform is carried out to intrinsic mode function one by one, obtains the instantaneous amplitude and instantaneous frequency of every layer of intrinsic mode function
Rate obtains the Hilbert spectrum for receiving signal.
4. the Specific Emitter Identification method according to claim 1 or 3 based on depth residual error network, which is characterized in that
Its matrix is obtained according to Hilbert spectrum, according to gray level image quantization digit, obtains the corresponding gray level image of the matrix.
5. the Specific Emitter Identification method according to claim 1 based on depth residual error network, which is characterized in that depth
Residual error network model is using the network architecture trained, and the network architecture trained is by utilizing sample data pair
Network model parameter is trained study.
6. the Specific Emitter Identification method according to claim 1 based on depth residual error network, which is characterized in that pass through
Depth residual error network model training learning network mode input and the desired residual error mapping of output, realization divide specific source of radiation
Class identification.
7. the Specific Emitter Identification method according to claim 1 based on depth residual error network, which is characterized in that depth
Residual error network architecture uses M residual unit, and M is the integer greater than 1.
8. the Specific Emitter Identification method according to claim 1 based on depth residual error network, which is characterized in that depth
Residual error network model includes input layer, first convolutional layer, maximum pond layer, multiple convolutional layers, average pond layer and output layer,
In, multiple convolutional layers include sequentially connected 8 convolutional layers, and the activation primitive in each convolutional layer and first convolutional layer is all made of
ReLU activation primitive, output layer use Softmax function.
9. a kind of Specific Emitter Identification device based on depth residual error network is, characterized by comprising: conversion module and identification
Module, wherein
Conversion module for carrying out time frequency analysis to reception signal, and converts grayscale image for obtained Hilbert time-frequency spectrum
Picture;
Identification module extracts the radio-frequency fingerprint of reflection in the picture using depth residual error network for being input with gray level image
Feature obtains the recognition result of radiation source.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811575557.2A CN109684995A (en) | 2018-12-22 | 2018-12-22 | Specific Emitter Identification method and device based on depth residual error network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811575557.2A CN109684995A (en) | 2018-12-22 | 2018-12-22 | Specific Emitter Identification method and device based on depth residual error network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109684995A true CN109684995A (en) | 2019-04-26 |
Family
ID=66188102
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811575557.2A Pending CN109684995A (en) | 2018-12-22 | 2018-12-22 | Specific Emitter Identification method and device based on depth residual error network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109684995A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110133577A (en) * | 2019-05-23 | 2019-08-16 | 成都理工大学 | The relatively prime MIMO array DOA algorithm for estimating of single base based on time-frequency residual error network |
CN110263810A (en) * | 2019-05-17 | 2019-09-20 | 西北大学 | A kind of LoRa signal source discrimination |
CN110365639A (en) * | 2019-05-29 | 2019-10-22 | 中国科学院信息工程研究所 | A kind of malicious traffic stream detection method and system based on depth residual error network |
CN110458051A (en) * | 2019-07-25 | 2019-11-15 | 中移(杭州)信息技术有限公司 | A kind of method, apparatus, server and the readable storage medium storing program for executing of equipment control |
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 |
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 |
CN112087774A (en) * | 2020-09-14 | 2020-12-15 | 桂林电子科技大学 | Communication radiation source individual identification method based on residual error neural network |
CN112507855A (en) * | 2020-12-04 | 2021-03-16 | 国网浙江省电力有限公司武义县供电公司 | Communication radiation source individual identification method based on instantaneous envelope equipotential sphere diagram |
CN112733613A (en) * | 2020-12-18 | 2021-04-30 | 国网浙江省电力有限公司武义县供电公司 | Radiation source identification method based on Hilbert transform and Helbert coefficient characteristics |
CN112749633A (en) * | 2020-12-25 | 2021-05-04 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Separate and reconstructed individual radiation source identification method |
CN113343868A (en) * | 2021-06-15 | 2021-09-03 | 四川九洲电器集团有限责任公司 | Radiation source individual identification method and device, terminal and storage medium |
CN113537053A (en) * | 2021-07-15 | 2021-10-22 | 四川九洲电器集团有限责任公司 | Method for constructing radio frequency fingerprint identification model in civil aviation field |
CN113905383A (en) * | 2021-08-26 | 2022-01-07 | 湖南艾科诺维科技有限公司 | IFF signal identification method, device and medium based on radio frequency fingerprint |
CN115510924A (en) * | 2022-11-17 | 2022-12-23 | 中铁第一勘察设计院集团有限公司 | Radio frequency fingerprint identification method based on improved variational modal decomposition |
CN115865579A (en) * | 2023-03-02 | 2023-03-28 | 武汉能钠智能装备技术股份有限公司 | Blind estimation method for roll-off coefficient of digital modulation signal based on deep learning |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103245937A (en) * | 2013-05-27 | 2013-08-14 | 四川大学 | Micro moving target feature extracting method based on micro Doppler effect |
CN106778610A (en) * | 2016-12-16 | 2017-05-31 | 哈尔滨工程大学 | A kind of intra-pulse modulation recognition methods based on time-frequency image feature |
CN107578775A (en) * | 2017-09-07 | 2018-01-12 | 四川大学 | A kind of multitask method of speech classification based on deep neural network |
CN107886050A (en) * | 2017-10-16 | 2018-04-06 | 电子科技大学 | Utilize time-frequency characteristics and the Underwater targets recognition of random forest |
CN108282263A (en) * | 2017-12-15 | 2018-07-13 | 西安电子科技大学 | Coded modulation joint recognition methods based on one-dimensional depth residual error light weight network |
CN108764077A (en) * | 2018-05-15 | 2018-11-06 | 重庆邮电大学 | A kind of digital signal modulated sorting technique based on convolutional neural networks |
-
2018
- 2018-12-22 CN CN201811575557.2A patent/CN109684995A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103245937A (en) * | 2013-05-27 | 2013-08-14 | 四川大学 | Micro moving target feature extracting method based on micro Doppler effect |
CN106778610A (en) * | 2016-12-16 | 2017-05-31 | 哈尔滨工程大学 | A kind of intra-pulse modulation recognition methods based on time-frequency image feature |
CN107578775A (en) * | 2017-09-07 | 2018-01-12 | 四川大学 | A kind of multitask method of speech classification based on deep neural network |
CN107886050A (en) * | 2017-10-16 | 2018-04-06 | 电子科技大学 | Utilize time-frequency characteristics and the Underwater targets recognition of random forest |
CN108282263A (en) * | 2017-12-15 | 2018-07-13 | 西安电子科技大学 | Coded modulation joint recognition methods based on one-dimensional depth residual error light weight network |
CN108764077A (en) * | 2018-05-15 | 2018-11-06 | 重庆邮电大学 | A kind of digital signal modulated sorting technique based on convolutional neural networks |
Non-Patent Citations (2)
Title |
---|
JINGWEN ZHANG 等: "Specific Emitter Identification via Hilbert–Huang Transform in Single-Hop and Relaying Scenarios", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》 * |
任东方 等: "基于 ITD 与纹理分析的特定辐射源识别方法", 《通信学报》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110263810A (en) * | 2019-05-17 | 2019-09-20 | 西北大学 | A kind of LoRa signal source discrimination |
CN110133577A (en) * | 2019-05-23 | 2019-08-16 | 成都理工大学 | The relatively prime MIMO array DOA algorithm for estimating of single base based on time-frequency residual error network |
CN110365639A (en) * | 2019-05-29 | 2019-10-22 | 中国科学院信息工程研究所 | A kind of malicious traffic stream detection method and system based on depth residual error network |
CN110458051A (en) * | 2019-07-25 | 2019-11-15 | 中移(杭州)信息技术有限公司 | A kind of method, apparatus, server and the readable storage medium storing program for executing of equipment control |
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 |
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 |
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 |
CN112087774A (en) * | 2020-09-14 | 2020-12-15 | 桂林电子科技大学 | Communication radiation source individual identification method based on residual error neural network |
CN112087774B (en) * | 2020-09-14 | 2023-04-18 | 桂林电子科技大学 | Communication radiation source individual identification method based on residual error neural network |
CN112507855A (en) * | 2020-12-04 | 2021-03-16 | 国网浙江省电力有限公司武义县供电公司 | Communication radiation source individual identification method based on instantaneous envelope equipotential sphere diagram |
CN112733613A (en) * | 2020-12-18 | 2021-04-30 | 国网浙江省电力有限公司武义县供电公司 | Radiation source identification method based on Hilbert transform and Helbert coefficient characteristics |
CN112749633A (en) * | 2020-12-25 | 2021-05-04 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Separate and reconstructed individual radiation source identification method |
CN112749633B (en) * | 2020-12-25 | 2022-05-17 | 西南电子技术研究所(中国电子科技集团公司第十研究所) | Separate and reconstructed individual radiation source identification method |
CN113343868A (en) * | 2021-06-15 | 2021-09-03 | 四川九洲电器集团有限责任公司 | Radiation source individual identification method and device, terminal and storage medium |
CN113537053A (en) * | 2021-07-15 | 2021-10-22 | 四川九洲电器集团有限责任公司 | Method for constructing radio frequency fingerprint identification model in civil aviation field |
CN113537053B (en) * | 2021-07-15 | 2023-08-22 | 四川九洲电器集团有限责任公司 | Method for constructing radio frequency fingerprint identification model in civil aviation field |
CN113905383A (en) * | 2021-08-26 | 2022-01-07 | 湖南艾科诺维科技有限公司 | IFF signal identification method, device and medium based on radio frequency fingerprint |
CN113905383B (en) * | 2021-08-26 | 2024-02-06 | 湖南艾科诺维科技有限公司 | IFF signal identification method, device and medium based on radio frequency fingerprint |
CN115510924A (en) * | 2022-11-17 | 2022-12-23 | 中铁第一勘察设计院集团有限公司 | Radio frequency fingerprint identification method based on improved variational modal decomposition |
CN115510924B (en) * | 2022-11-17 | 2023-04-07 | 中铁第一勘察设计院集团有限公司 | Radio frequency fingerprint identification method based on improved variational modal decomposition |
CN115865579A (en) * | 2023-03-02 | 2023-03-28 | 武汉能钠智能装备技术股份有限公司 | Blind estimation method for roll-off coefficient of digital modulation signal based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109684995A (en) | Specific Emitter Identification method and device based on depth residual error network | |
Jagannath et al. | A comprehensive survey on radio frequency (RF) fingerprinting: Traditional approaches, deep learning, and open challenges | |
Zhang et al. | An efficient deep learning model for automatic modulation recognition based on parameter estimation and transformation | |
Ozturk et al. | RF-based low-SNR classification of UAVs using convolutional neural networks | |
CN108616470A (en) | Modulation Signals Recognition method based on convolutional neural networks | |
Chen et al. | Deep STFT-CNN for spectrum sensing in cognitive radio | |
CN113156376B (en) | SACNN-based radar radiation source signal identification method | |
Tan et al. | Semi-supervised specific emitter identification based on bispectrum feature extraction CGAN in multiple communication scenarios | |
Yao et al. | Few-shot specific emitter identification using asymmetric masked auto-encoder | |
Zhou et al. | Specific emitter identification via bispectrum‐radon transform and hybrid deep model | |
CN113205140B (en) | Semi-supervised specific radiation source individual identification method based on generative countermeasure network | |
Teng et al. | Modulation classification based on spectral correlation and SVM | |
Liang et al. | A radio signal recognition approach based on complex-valued CNN and self-attention mechanism | |
Cai et al. | Jamming pattern recognition using spectrum waterfall: A deep learning method | |
Wang et al. | Adversarial unsupervised domain adaptation for cross scenario waveform recognition | |
Zha et al. | LT-SEI: Long-tailed specific emitter identification based on decoupled representation learning in low-resource scenarios | |
CN115114958A (en) | Electromagnetic signal open set identification method based on supervised contrast learning | |
Zhang et al. | Turning wingbeat sounds into spectrum images for acoustic insect classification | |
CN115392285A (en) | Deep learning signal individual recognition model defense method based on multiple modes | |
Cheng et al. | Radar signal recognition exploiting information geometry and support vector machine | |
Dey et al. | Morse code datasets for machine learning | |
Khan et al. | A hybrid convolutional neural network with fusion of handcrafted and deep features for FHSS signals classification | |
Yin et al. | Towards fully domain‐independent gesture recognition using COTS WiFi device | |
Guo et al. | Radar signal recognition based on FCBF and Adaboost algorithm | |
CN113905383A (en) | IFF signal identification method, device and medium based on radio frequency fingerprint |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190426 |