CN108282426A - Radio signal recognition recognition methods based on lightweight depth network - Google Patents
Radio signal recognition recognition methods based on lightweight depth network Download PDFInfo
- Publication number
- CN108282426A CN108282426A CN201711293537.1A CN201711293537A CN108282426A CN 108282426 A CN108282426 A CN 108282426A CN 201711293537 A CN201711293537 A CN 201711293537A CN 108282426 A CN108282426 A CN 108282426A
- Authority
- CN
- China
- Prior art keywords
- signal
- depth network
- modulation
- layer
- lightweight depth
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3912—Simulation models, e.g. distribution of spectral power density or received signal strength indicator [RSSI] for a given geographic region
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
- H04L1/004—Arrangements for detecting or preventing errors in the information received by using forward error control
- H04L1/0056—Systems characterized by the type of code used
- H04L1/0059—Convolutional codes
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The present invention discloses a kind of radio signal recognition recognition methods based on lightweight depth network, and implementation step is:(1) coded modulation allied signal is built;(2) training sample set and test sample collection are generated;(3) lightweight depth network is built;(4) parameter of lightweight depth network is set;(5) training lightweight depth network;(6) cognition recognition accuracy is obtained.The present invention has universality strong, one-dimensional wireless electric signal can directly be handled, manual features extraction and priori are not needed, the channel coding method and Modulation Identification mode of identification radio signal can be recognized simultaneously, complexity is low, model lightweight, the advantage that classification results are accurate, stable, can be used in radio signal recognition identification technology field.
Description
Technical field
The invention belongs to field of communication technology, a kind of lightweight depth in signal processing technology field is further related to
The radio signal recognition recognition methods of network.The present invention simulates the process of biological brain cognition identification, can be in complicated electricity
Under magnetic environment, the hierarchical semantic feature of all kinds of radio signals is automatically extracted, realizes the automatic channel coding of radio signal
Type combines cognition identification with modulation system type.Compared to existing deep learning model, the present invention, which not only has, calculates complexity
It spends low, parameter small scale, be easy to the characteristics of hardware realization, and can accurately recognize identification knot compared with acquisition under low signal-to-noise ratio
Fruit.
Background technology
The joint identification of radio signal coded modulation is fought in military electronic, is played the part of in hostile scouting and signal capture analysis
Key player, in the case where Given information extremely lacks, the coded modulation joint understanding of signal Zuo Wei not signal processing stream
Journey critical process plays decisive role to the final identification of information.Major scientific research institution and colleges and universities grind both at home and abroad at present
Study carefully and be all based on signal priori greatly, the channel coding method that radio signal is carried out using the method for artificial design features is identified.This
Traditional methods need the domain knowledge of a large amount of priori and professional system, and the feature obtained is in universality and robustness
On there are many restrictions.With the complication of communication environment, the interference of the electromagnetic environment where signal is also more and more, by
It is also more prominent the shortcomings that conventional method under the complex communication environment of severe jamming.On the other hand, although based on conventional method
The identification of radio signal channel coding type can reach relatively satisfactory discrimination on certain class signal, but join in coded modulation
It closes and needs further to develop and improve in the accuracy and validity of cognition identification.Therefore we simulate biological brain cognition identification
Process, establish a lightweight depth network model, automatically extract the hierarchical semantic feature of all kinds of radio signals, it is real
To Automatic Feature Extraction and coded modulation the joint cognition identification of radio signal under present complex electromagnetic environment.
In the patent document of its application, " the figure field communication signal modulation based on fractional lower-order Cyclic Spectrum is known for University of Electronic Science and Technology
It is disclosed in other method " (201710546645.9 application publication number CN of application number, 107135176 A) a kind of based on fractional lower-order
The figure field communication signal modulate method of Cyclic Spectrum.The step of this method is:It is followed using the three-dimensional fractional lower-order for receiving signal
Ring is composed, and will be transformed on figure domain by the modulated signal of α Stable distritation noise jammings, the sparse adjacent square that then can be indicated from figure
Feature of the extraction effective characteristic parameters line index arrangement set as modulation type in battle array, according to training signal and reception signal
Line index arrangement set Hamming distance, to realize under α Stable distritation noise jammings, the modulation of more stable more effective signal of communication
The identification of type.Although this method proposes a kind of figure field communication signal modulate method based on fractional lower-order Cyclic Spectrum,
But the shortcoming that this method still has is:This method needs to carry out the conversion of figure domain just to signal can be identified, mistake
It is extracted in dependent on manual features, model is complicated.And it is only capable of that the modulation system of signal is individually identified, signal can not be compiled
Code modulation system is identified.
In paper " a kind of channel coding using soft-decision identifies new algorithm " (electronic letters, vol that abundant east et al. is delivered at it
2 months the 2nd phases in 2013) in elaborate a kind of code identification new algorithm using soft-decision.The algorithm implementation method is as follows:It is based on
Containing wrong equation model, using log-likelihood ratio, using the probability that equation is set up as the measurement for weighing solution vector performance, to solve
Equation completes the channel coding identification of signal.Shortcoming existing for this method is:Although this method proposes a kind of channel volume
Code recognition methods, but it should be understood that a large amount of signal priori, is only capable of that the channel coding type of signal is individually identified, it can not be right
The modulation system of signal is identified, and needs complicated manual features extraction.
Invention content
The present invention in view of the above shortcomings of the prior art, proposes that a kind of radio signal recognition of lightweight depth network is known
Other method.
Realizing the concrete thought of the object of the invention is, radio signal recognition identification is carried out using lightweight depth network.
The algorithm can reach higher cognition discrimination in signal cognition identification, while can reduce conventional modulated recognition methods pair again
Manual features are extracted and the high dependency of priori, can appreciate that the channel coding method class for identifying a variety of radio signals
Type and modulation system type, and simplify identification step.To realize the joint cognition identification of radio signal coded modulation, and make
The joint cognition identification of radio signal coded modulation is more flexible, efficient.
Realize that the specific steps of the object of the invention include as follows:
(1) coded modulation allied signal is built:
Each information sequence in the information sequence set received is carried out four kinds of channel codings by (1a) successively, raw
At different encoded signals;
Each signal after coding is carried out six kinds of modulation by (1b) successively, obtains coded modulation allied signal;
(2) training sample set and test sample collection are generated:
Between (2a) is with 100 information points to all information points of each sample of signal in coded modulation allied signal
Every by 440 information points one sample of signal of composition of each continuous acquisition, by all sample of signal composition sample of signal collection;
(2b) concentrates the sample of signal for randomly selecting 80% to form training sample set, remaining sample of signal from sample of signal
Form test sample collection;
(3) lightweight depth network is built:
(3a) builds 16 layers of lightweight depth network for automatically extracting coded modulation allied signal feature;
(3b) setting lightweight depth network in loss function be cross entropy, optimization algorithm be Back Propagation Algorithm,
Activation primitive is set as correcting linear unit activating function;
(4) parameter of lightweight depth network is set:
It is 440 input neural units that input layer, which is arranged, in (4a);
The parameter that the different convolution kernels of different convolutional layers in lightweight depth network are arranged in (4b) is as follows:First convolutional layer is
64 convolution kernels, the matrix that each convolution kernel is 1 × 19;Second convolutional layer is 64 convolution kernels, and each convolution kernel is 1 × 21
Matrix;Third convolutional layer is 128 convolution kernels, the matrix that each convolution kernel is 1 × 19;Volume Four lamination is 128 convolution kernels,
The matrix that each convolution kernel is 1 × 21;5th convolutional layer is 256 convolution kernels, the matrix that each convolution kernel is 1 × 19;6th
Convolutional layer is 256 convolution kernels, the matrix that each convolution kernel is 1 × 21;
(4c) is by the first pond layer, the Chi Huafang of the second pond layer, third pond layer, the 4th pond layer and the 5th pond layer
Formula is set as maximum pond mode;Set grader layer to more classification function Softmax;
The neuron number of the first full articulamentum and the second full articulamentum is respectively in (4d) setting lightweight depth network
64 and 24;
(5) training lightweight depth network:
Training sample set is input to training 18 times in lightweight depth network, obtains trained lightweight depth net
Network;
(6) cognition recognition accuracy is obtained:
Test sample collection is input in trained lightweight depth network by (6a), obtains cognition recognition result;
(6b) compares the true classification for recognizing recognition result and test sample collection, statistics cognition recognition correct rate.
Compared with the prior art, the present invention has the following advantages:
First, since the present invention is when being arranged the parameter of lightweight depth network, set input layer to 440 input god
Through unit, for directly being handled original information sequence, overcomes and need to carry out the conversion of figure domain to signal in the prior art
Can just carry out cognition identification the problem of so that the present invention can utilize lightweight depth network model directly to one-dimensional signal into
Row cognition identification.
Second, since the present invention is when being arranged the parameter of lightweight depth network, it is arranged in different convolutional layers different
The convolution kernel of quantity and size, the profound feature for successively extracting signal, while the diversity of feature is increased, it overcomes
The prior art excessively depends on the shortcomings that manual features extraction, and lightweight depth network model in the present invention is allow to automatically process
The cognition of multiple types signal identifies.
Third, since the present invention builds 16 layers of lightweight depth net for automatically extracting coded modulation allied signal feature
Network carries out feature extraction and analysis to signal automatically using one-dimensional convolution, and overcome the prior art needs when carrying out cognition identification
It is to be understood that the shortcomings that a large amount of signal prioris, while reducing the number of parameters of model so that model of the invention is lighter
Quantization improves the efficiency of network signal cognition identification.
4th, since the present invention builds 16 layers of lightweight depth net for automatically extracting coded modulation allied signal feature
Network carries out coded modulation joint cognition identification for the coded modulation allied signal to structure, overcomes the prior art and be only capable of list
The modulation system of only identification signal or the problem of be only capable of that the channel coding type of signal is individually identified, makes lightweight in the present invention
Depth network model can carry out the modulating-coding joint cognition identification of signal, enhance the pervasive of lightweight depth network model
Property.
Description of the drawings
Fig. 1 is the flow chart of the present invention;
Fig. 2 is the analogous diagram for 24 kinds of coded modulation allied signals that the present invention is built;
Fig. 3 is the result figure of emulation experiment of the present invention.
Specific implementation mode
Invention is described further below in conjunction with the accompanying drawings.
With reference to attached drawing 1, the specific steps of the present invention are further described.
Step 1, coded modulation allied signal is built.
Each information sequence in the information sequence set received is carried out four kinds of channel codings by the first step successively,
Generate different encoded signals.
Four kinds of channel codings refer to, Hamming code channel coding, half code check 216 nonsystematic convolutional code channels
432 nonsystematic convolutional code channels of coding, 216 nonsystematic convolutional code channel codings of 2/3rds code checks, 3/4ths code checks
Coding.
Each signal after coding is carried out six kinds of modulation, obtains coded modulation allied signal by second step successively.
Six kinds of modulation refer to binary phase shift keying modulation, the modulation of quaternary phase-shift keying (PSK), octal system phase-shift keying (PSK)
Secondary modulation, the quaternary phase shift that modulation, binary digit frequency modulation(PFM), binary digit frequency modulation(PFM) are combined with frequency modulation(PFM)
The secondary modulation that keying is combined with frequency modulation(PFM).
Step 2, training sample set and test sample collection are generated.
The first step, all information points to each sample of signal in coded modulation allied signal are with 100 information points
440 information points of each continuous acquisition are formed a sample of signal, all sample of signal are formed sample of signal by interval
Collection.
Second step concentrates the sample of signal for randomly selecting 80% to form training sample set, remaining signal from sample of signal
Sample forms test sample collection.
Step 3, lightweight depth network is built.
The first step builds 16 layers of lightweight depth network for automatically extracting coded modulation allied signal feature.
The structure of 16 layers of lightweight depth network is:Convolutional layer → the first of the convolutional layer of input layer → first → second
The pond of the convolutional layer of the pond layer of pond layer → third convolutional layer → second → Volume Four lamination → third pond layer → the 5th → the 4th
The full articulamentum of full articulamentum → the second of pond layer → the first of layer → the 6th convolutional layer → the 5th → grader layer → output layer.
Second step, the loss function being arranged in lightweight depth network is cross entropy, optimization algorithm is that error Back-Propagation is calculated
Method, activation primitive are set as correcting linear unit activating function.
Step 4, the parameter of lightweight depth network is set.
The first step, setting input layer are 440 input neural units.
Second step, the parameter that the different convolution kernels of different convolutional layers in lightweight depth network are arranged are as follows:First convolution
Layer is 64 convolution kernels, the matrix that each convolution kernel is 1 × 19;Second convolutional layer be 64 convolution kernels, each convolution kernel be 1 ×
21 matrix;Third convolutional layer is 128 convolution kernels, the matrix that each convolution kernel is 1 × 19;Volume Four lamination is 128 volumes
Product core, the matrix that each convolution kernel is 1 × 21;5th convolutional layer is 256 convolution kernels, the matrix that each convolution kernel is 1 × 19;
6th convolutional layer is 256 convolution kernels, the matrix that each convolution kernel is 1 × 21.
Third walks, by the pond of the first pond layer, the second pond layer, third pond layer, the 4th pond layer and the 5th pond layer
Change mode is set as maximum pond mode;Set grader layer to more classification function Softmax.
4th step is arranged the neuron number of the first full articulamentum and the second full articulamentum in lightweight depth network and distinguishes
For 64 and 24.
Step 5, training lightweight depth network model.
Training sample set is input in lightweight depth network model and is trained 18 times, trained lightweight is obtained
Depth network model.
Step 6, cognition recognition accuracy is obtained.
Test sample collection is input in trained lightweight depth network by the first step, obtains cognition recognition result.
Second step compares the true classification for recognizing recognition result and test sample collection, and statistics cognition identification is correct
Rate.
1. simulated conditions:
The emulation experiment of the present invention is in Intel (R) E5-2630CPU 2GHz, GTX1080, Ubuntu16.04LTS systems
Under, on TensorFlow1.0.1 operation platforms, complete the present invention and structure coded modulation allied signal and lightweight depth net
The emulation experiment of the radio signal recognition identification of network.
2. emulation experiment content:
The oscillogram of 24 kinds of coded modulation allied signals used in the emulation experiment of the present invention is as shown in Fig. 2, Fig. 2 (a) institutes
It is shown as the allied signal oscillogram of the channel combined binary phase shift keying modulation of Hamming code, Fig. 2 (b) show half
The allied signal oscillogram of the channel combined binary phase shift keying modulation of 216 nonsystematic convolutional codes of code check, Fig. 2 (c) institutes
It is shown as the allied signal waveform of the channel combined binary phase shift keying modulation of 216 nonsystematic convolutional codes of 2/3rds code checks
Figure, Fig. 2 (d) show the channel combined binary phase shift keying modulation of 432 nonsystematic convolutional codes of 3/4ths code checks
Allied signal oscillogram, Fig. 2 (e) show the allied signal waveform of channel combined eight phase shift key modulation of Hamming code
Figure, Fig. 2 (f) show the connection of channel combined eight phase shift key modulation of 216 nonsystematic convolutional codes of half code check
Conjunction signal waveforms, Fig. 2 (g) show the channel combined eight phases phase shift key of 216 nonsystematic convolutional codes of 2/3rds code checks
432 nonsystematic convolutional codes that the allied signal oscillogram of control modulation, Fig. 2 (h) show 3/4ths code checks are channel combined
The allied signal oscillogram of eight phase shift key modulations, Fig. 2 (i) show the channel combined binary number word frequency of Hamming code
The allied signal oscillogram of modulation, the channel combined binary system of 216 nonsystematic convolutional codes that Fig. 2 (j) is half code check
The 216 nonsystematic convolutional code channels that the allied signal oscillogram of digital frequency modulation, Fig. 2 (k) show 2/3rds code checks are compiled
The warbled allied signal oscillogram of code joint binary digit, Fig. 2 (l) show 432 nonsystematics of 3/4ths code checks
The warbled allied signal oscillogram of the channel combined binary digit of convolutional code, Fig. 2 (m) show Hamming code channel volume
Code joint binary digit frequency modulation(PFM) show two points with the allied signal oscillogram of warbled secondary modulation, Fig. 2 (n)
One of the channel combined binary digit frequency modulation(PFM) of 216 nonsystematic convolutional codes and the warbled secondary modulation of code check
Allied signal oscillogram, Fig. 2 (o) show the channel combined binary number of 216 nonsystematic convolutional codes of 2/3rds code checks
Word frequency is modulated show the 432 of 3/4ths code checks with the allied signal oscillogram of warbled secondary modulation, Fig. 2 (p)
The allied signal waveform of nonsystematic convolutional code channel combined binary digit frequency modulation(PFM) and warbled secondary modulation
Figure, Fig. 2 (q) show the allied signal oscillogram of the channel combined quaternary digital frequency modulation of Hamming code, Fig. 2 (r) institutes
It is shown as the allied signal waveform of the channel combined quaternary digital frequency modulation of 216 nonsystematic convolutional codes of half code check
Figure, Fig. 2 (s) show the channel combined quaternary digital frequency modulation of 216 nonsystematic convolutional codes of 2/3rds code checks
Allied signal oscillogram, Fig. 2 (t) show the channel combined quaternary number of 432 nonsystematic convolutional codes of 3/4ths code checks
Word frequency modulation allied signal oscillogram, Fig. 2 (u) show the channel combined quaternary digital frequency modulation of Hamming code with
The allied signal oscillogram of warbled secondary modulation, Fig. 2 (v) show 216 nonsystematic convolutional codes of half code check
Channel combined quaternary digital frequency modulation and the allied signal oscillogram of warbled secondary modulation, Fig. 2 (w) are shown
For 2/3rds code checks the channel combined quaternary digital frequency modulation of 216 nonsystematic convolutional codes with it is warbled secondary
The allied signal oscillogram of modulation, Fig. 2 (x) show 432 nonsystematic convolutional codes channel combined four of 3/4ths code checks
The allied signal oscillogram of binary digits frequency modulation(PFM) and warbled secondary modulation.
3. the simulation experiment result is analyzed:
The simulation experiment result of the present invention is as shown in Figure 3.Horizontal axis in Fig. 3 represents training iterations, and the longitudinal axis corresponds to every
The loss function value train loss of secondary iteration.During to lightweight depth network training, each training result is counted
Loss function value, the training effect of the smaller representative model of loss function value is better.As seen from Figure 3, with the increasing of iterations
Add loss function value to successively decrease and finally converge to stabilization, illustrate the training effect of this emulation experiment with increasing for frequency of training and
It improves.
Test sample is inputted in trained lightweight depth network, the cognition for obtaining each signal in 24 kinds of signals is known
Not as a result, again comparing the true classification of the cognition recognition result of each signal and test sample collection, statistics cognition identification
As a result the number of correct test sample finds out the percentage of test sample shared by the cognition correct test sample of recognition result,
The cognition recognition accuracy for obtaining this emulation experiment is 94%.
It can be illustrated by above emulation experiment, combine cognition identification for radio signal coded modulation, the present invention can
To complete the cognition identification mission of different classes of radio signal, method is feasible.
Claims (4)
1. a kind of radio signal recognition recognition methods based on lightweight depth network, which is characterized in that include the following steps:
(1) coded modulation allied signal is built:
Each information sequence in the information sequence set received is carried out four kinds of channel codings by (1a) successively, is generated not
Same encoded signal;
Each signal after coding is carried out six kinds of modulation by (1b) successively, obtains coded modulation allied signal;
(2) training sample set and test sample collection are generated:
(2a) using 100 information points as interval, incites somebody to action all information points of each sample of signal in coded modulation allied signal
Each 440 information points of continuous acquisition form a sample of signal, and all sample of signal are formed sample of signal collection;
(2b) concentrates the sample of signal for randomly selecting 80% to form training sample set, remaining sample of signal composition from sample of signal
Test sample collection;
(3) lightweight depth network is built:
(3a) builds 16 layers of lightweight depth network for automatically extracting coded modulation allied signal feature;
Loss function in (3b) setting lightweight depth network is cross entropy, optimization algorithm is Back Propagation Algorithm, activation
Function setup is to correct linear unit activating function;
(4) parameter of lightweight depth network is set:
It is 440 input neural units that input layer, which is arranged, in (4a);
The parameter that the different convolution kernels of different convolutional layers in lightweight depth network are arranged in (4b) is as follows:First convolutional layer is 64
Convolution kernel, the matrix that each convolution kernel is 1 × 19;Second convolutional layer is 64 convolution kernels, the square that each convolution kernel is 1 × 21
Battle array;Third convolutional layer is 128 convolution kernels, the matrix that each convolution kernel is 1 × 19;Volume Four lamination is 128 convolution kernels, often
The matrix that a convolution kernel is 1 × 21;5th convolutional layer is 256 convolution kernels, the matrix that each convolution kernel is 1 × 19;Volume six
Lamination is 256 convolution kernels, the matrix that each convolution kernel is 1 × 21;
(4c) sets the pond mode of the first pond layer, the second pond layer, third pond layer, the 4th pond layer and the 5th pond layer
It is set to maximum pond mode;Set grader layer to more classification function Softmax;
It is respectively 64 Hes that the neuron number of the first full articulamentum and the second full articulamentum in lightweight depth network, which is arranged, in (4d)
24;
(5) training lightweight depth network:
Training sample set is input to training 18 times in lightweight depth network, obtains trained lightweight depth network;
(6) cognition recognition accuracy is obtained:
Test sample collection is input in trained lightweight depth network by (6a), obtains cognition recognition result;
(6b) compares the true classification for recognizing recognition result and test sample collection, statistics cognition recognition correct rate.
2. the radio signal recognition recognition methods according to claim 1 based on lightweight depth network, feature exist
In four kinds of channel codings described in step (1a) refer to Hamming code channel coding, 216 nonsystematic convolution of half code check
Code channel coding, 216 nonsystematic convolutional code channel codings of 2/3rds code checks, 432 nonsystematic convolution of 3/4ths code checks
Code channel coding.
3. the radio signal recognition recognition methods according to claim 1 based on lightweight depth network, feature exist
In six kinds of modulation described in step (1b) refer to binary phase shift keying modulation, the modulation of quaternary phase-shift keying (PSK), octal system phase
Move keying modulation, the secondary modulation that binary digit frequency modulation(PFM), binary digit frequency modulation(PFM) are combined with frequency modulation(PFM), four into
The secondary modulation that phase-shift keying (PSK) processed is combined with frequency modulation(PFM).
4. the radio signal recognition recognition methods according to claim 1 based on lightweight depth network, feature exist
In the structure of 16 layers of lightweight depth network described in step (3a) is:The convolutional layer of the convolutional layer of input layer → first → second
The pond layer of → the first pond layer → third convolutional layer → second → Volume Four lamination → five convolutional layer → the of third pond layer → the
The full articulamentum of full articulamentum → the second of pond layer → the first of the convolutional layer of four pond layers → the 6th → the 5th → grader layer → output
Layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711293537.1A CN108282426B (en) | 2017-12-08 | 2017-12-08 | Radio signal recognition recognition methods based on lightweight depth network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711293537.1A CN108282426B (en) | 2017-12-08 | 2017-12-08 | Radio signal recognition recognition methods based on lightweight depth network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108282426A true CN108282426A (en) | 2018-07-13 |
CN108282426B CN108282426B (en) | 2019-11-26 |
Family
ID=62801440
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711293537.1A Active CN108282426B (en) | 2017-12-08 | 2017-12-08 | Radio signal recognition recognition methods based on lightweight depth network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108282426B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109347601A (en) * | 2018-10-12 | 2019-02-15 | 哈尔滨工业大学 | The interpretation method of anti-tone interference LDPC code based on convolutional neural networks |
CN109495214A (en) * | 2018-11-26 | 2019-03-19 | 电子科技大学 | Channel coding type recognition methods based on one-dimensional Inception structure |
CN109525528A (en) * | 2018-09-29 | 2019-03-26 | 电子科技大学 | Figure domain signal recognition method towards MQAM modulated signal |
CN109787929A (en) * | 2019-02-20 | 2019-05-21 | 深圳市宝链人工智能科技有限公司 | Signal modulate method, electronic device and computer readable storage medium |
CN110647810A (en) * | 2019-08-16 | 2020-01-03 | 西北大学 | Method and device for constructing and identifying radio signal image identification model |
CN110853630A (en) * | 2019-10-30 | 2020-02-28 | 华南师范大学 | Lightweight speech recognition method facing edge calculation |
CN111079898A (en) * | 2019-11-28 | 2020-04-28 | 华侨大学 | Channel coding identification method based on TextCNN network |
CN111490853A (en) * | 2020-04-15 | 2020-08-04 | 成都海擎科技有限公司 | Channel coding parameter identification method based on deep convolutional neural network |
CN111795611A (en) * | 2020-05-20 | 2020-10-20 | 中南民族大学 | Low-complexity unmanned aerial vehicle modulation mode blind identification and countercheck method and system |
CN112737733A (en) * | 2020-12-28 | 2021-04-30 | 中国人民解放军国防科技大学 | Channel coding code pattern recognition method based on one-dimensional convolutional neural network |
CN113518049A (en) * | 2021-04-13 | 2021-10-19 | 江苏师范大学 | Modulation identification method based on fractional low-order polar coordinate and deep learning |
CN113542180A (en) * | 2021-06-30 | 2021-10-22 | 北京频谱视觉科技有限公司 | Frequency domain identification method of radio signal |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104408469A (en) * | 2014-11-28 | 2015-03-11 | 武汉大学 | Firework identification method and firework identification system based on deep learning of image |
CN106680775A (en) * | 2016-12-12 | 2017-05-17 | 清华大学 | Method and system for automatically identifying radar signal modulation modes |
CN107220606A (en) * | 2017-05-22 | 2017-09-29 | 西安电子科技大学 | The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks |
-
2017
- 2017-12-08 CN CN201711293537.1A patent/CN108282426B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104408469A (en) * | 2014-11-28 | 2015-03-11 | 武汉大学 | Firework identification method and firework identification system based on deep learning of image |
CN106680775A (en) * | 2016-12-12 | 2017-05-17 | 清华大学 | Method and system for automatically identifying radar signal modulation modes |
CN107220606A (en) * | 2017-05-22 | 2017-09-29 | 西安电子科技大学 | The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks |
Non-Patent Citations (2)
Title |
---|
GIHAN J. MENDIS,JIN WEI,ARJUNA MADANAYAKE: "Deep Learning-Based Automated Modulation Classification for Cognitive Radio", 《2016 IEEE INTERNATIONAL CONFERENCE ON COMMUNICATION SYSTEMS (ICCS)》 * |
XIAOYU LIU, DIYU YANG, ALY EL GAMAL: "Deep Neural Network Architectures for Modulation Classification", 《2017 51ST ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS, AND COMPUTERS》 * |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109525528A (en) * | 2018-09-29 | 2019-03-26 | 电子科技大学 | Figure domain signal recognition method towards MQAM modulated signal |
CN109525528B (en) * | 2018-09-29 | 2021-01-26 | 电子科技大学 | Image domain signal identification method facing MQAM modulation signal |
CN109347601A (en) * | 2018-10-12 | 2019-02-15 | 哈尔滨工业大学 | The interpretation method of anti-tone interference LDPC code based on convolutional neural networks |
CN109347601B (en) * | 2018-10-12 | 2021-03-16 | 哈尔滨工业大学 | Convolutional neural network-based decoding method of anti-tone-interference LDPC code |
CN109495214A (en) * | 2018-11-26 | 2019-03-19 | 电子科技大学 | Channel coding type recognition methods based on one-dimensional Inception structure |
CN109787929A (en) * | 2019-02-20 | 2019-05-21 | 深圳市宝链人工智能科技有限公司 | Signal modulate method, electronic device and computer readable storage medium |
CN110647810A (en) * | 2019-08-16 | 2020-01-03 | 西北大学 | Method and device for constructing and identifying radio signal image identification model |
CN110853630B (en) * | 2019-10-30 | 2022-02-18 | 华南师范大学 | Lightweight speech recognition method facing edge calculation |
CN110853630A (en) * | 2019-10-30 | 2020-02-28 | 华南师范大学 | Lightweight speech recognition method facing edge calculation |
CN111079898A (en) * | 2019-11-28 | 2020-04-28 | 华侨大学 | Channel coding identification method based on TextCNN network |
CN111079898B (en) * | 2019-11-28 | 2023-04-07 | 华侨大学 | Channel coding identification method based on TextCNN network |
CN111490853A (en) * | 2020-04-15 | 2020-08-04 | 成都海擎科技有限公司 | Channel coding parameter identification method based on deep convolutional neural network |
CN111795611A (en) * | 2020-05-20 | 2020-10-20 | 中南民族大学 | Low-complexity unmanned aerial vehicle modulation mode blind identification and countercheck method and system |
CN111795611B (en) * | 2020-05-20 | 2021-02-02 | 中南民族大学 | Low-complexity unmanned aerial vehicle modulation mode blind identification and countercheck method and system |
CN112737733A (en) * | 2020-12-28 | 2021-04-30 | 中国人民解放军国防科技大学 | Channel coding code pattern recognition method based on one-dimensional convolutional neural network |
CN112737733B (en) * | 2020-12-28 | 2024-06-07 | 中国人民解放军国防科技大学 | Channel coding pattern recognition method based on one-dimensional convolutional neural network |
CN113518049A (en) * | 2021-04-13 | 2021-10-19 | 江苏师范大学 | Modulation identification method based on fractional low-order polar coordinate and deep learning |
CN113518049B (en) * | 2021-04-13 | 2024-04-26 | 江苏师范大学 | Modulation identification method based on fractional low-order polar coordinates and deep learning |
CN113542180A (en) * | 2021-06-30 | 2021-10-22 | 北京频谱视觉科技有限公司 | Frequency domain identification method of radio signal |
Also Published As
Publication number | Publication date |
---|---|
CN108282426B (en) | 2019-11-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108282426B (en) | Radio signal recognition recognition methods based on lightweight depth network | |
CN107979554B (en) | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks | |
CN108282263B (en) | Coded modulation joint recognition methods based on one-dimensional depth residual error light weight network | |
CN108243129A (en) | Radio signal characteristics extracting method based on attention depth network | |
CN108282262B (en) | Intelligent clock signal classification method based on gating cycle unit depth network | |
CN111062250A (en) | Multi-subject motor imagery electroencephalogram signal identification method based on depth feature learning | |
CN107194426A (en) | A kind of image-recognizing method based on Spiking neutral nets | |
CN105160678A (en) | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method | |
CN109116312A (en) | Radar signal classification method based on QMFB and convolutional neural networks | |
CN109495214A (en) | Channel coding type recognition methods based on one-dimensional Inception structure | |
CN110020715B (en) | Neural network identification method and device using mixed coding of fluctuation and pulse signals | |
CN109525369A (en) | A kind of channel coding type blind-identification method based on Recognition with Recurrent Neural Network | |
CN113838107B (en) | Automatic heterogeneous image registration method based on dense connection | |
Alnujaim et al. | Generative adversarial networks to augment micro-doppler signatures for the classification of human activity | |
CN109543819A (en) | A kind of scene matching aided navigation localization method based on improved confinement convolutional neural networks | |
CN109728824A (en) | A kind of LDPC code iterative decoding method based on deep learning | |
CN101667292B (en) | SAR image segmentation system and segmentation method based on immune clone and projection pursuit | |
CN116797796A (en) | Signal identification method based on time-frequency analysis and deep learning under DRFM intermittent sampling | |
CN110363163A (en) | A kind of SAR target image generation method that azimuth is controllable | |
CN110752894B (en) | CNN-based LDPC code blind channel decoding method and decoder | |
CN112905828A (en) | Image retriever, database and retrieval method combined with significant features | |
CN114417722A (en) | Lightning electromagnetic pulse waveform compression method and system based on artificial intelligence | |
Bartunik et al. | Using deep learning to demodulate transmissions in molecular communication | |
CN108282427B (en) | Radio signal recognition recognition methods based on multiple dimensioned light weight network | |
CN114371474A (en) | Intelligent radar signal sorting method and system based on convolution noise reduction self-encoder |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |