CN108154164A - Signal of communication modulation classification system and method based on deep learning - Google Patents
Signal of communication modulation classification system and method based on deep learning Download PDFInfo
- Publication number
- CN108154164A CN108154164A CN201711131003.9A CN201711131003A CN108154164A CN 108154164 A CN108154164 A CN 108154164A CN 201711131003 A CN201711131003 A CN 201711131003A CN 108154164 A CN108154164 A CN 108154164A
- Authority
- CN
- China
- Prior art keywords
- signal
- data
- deep learning
- communication
- identification
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a kind of signal of communication modulation classification system and method based on deep learning, this method includes：The unified representation part of signal of communication, includes the following steps：Big data and depth are excavated and are applied in complex electromagnetic environment monitoring, make monitoring process more intelligent by step 1；Step 2 by the way that signal of communication unified representation is formed database form, makes environmental monitoring Modulation Identification frame change, breaking through previous increase monitoring signals can cause identification structure to change and change problem of identification structure etc.；The present invention has deep learning ability；Modulation recognition is represented independent of limited learning sample；Do not have to change Modulation Identification structure during signal kinds increase, autgmentability and versatility are good；In being monitored for radio signal, it can also be used to other to need to application fields such as signal modulation mode identification or parameter extractions.
Description
Technical field
The present invention relates to a kind of signal of communication modulation classification system and method, more particularly to a kind of based on deep learning
Signal of communication modulation classification system and method.
Background technology
What traditional spectrum monitoring faced is sparse signal environment, thus can according to the priori of signal carry out by
One analysis, and good result can be reached.With a large amount of increases using with signal kinds of wireless device, current frequency spectrum prison
Survey faces more complicated and unknown electromagnetic environment.The present invention is proposed by studying the latest theories of big data and deep learning
Radio signal characteristics unified representation new method is modulated identification using the expression, breaks through using limited learning sample to not
Know the key technology that radio signal is recognized, improve the knowledge acquisition to radio signal big data and value excavates energy
Power, to solve the problems, such as that current spectrum monitoring can not effectively recognize complex electromagnetic environment.
Invention content
The technical problems to be solved by the invention are to provide a kind of signal of communication modulation classification system based on deep learning
And method, with deep learning ability；Modulation recognition is represented independent of limited learning sample；During signal kinds increase not
With Modulation Identification structure is changed, autgmentability and versatility are good；In being monitored for radio signal, it can also be used to other to need to letter
The application fields such as number Modulation Mode Recognition or parameter extraction.
The present invention is to solve above-mentioned technical problem by following technical proposals：A kind of communication letter based on deep learning
Number modulation classification system, which is characterized in that it includes：
Original signal collection, the magnanimity actual signal being collected into electromagnetic space；
Deep learning engine is made inferences and learnt to input data using deep learning；
Knowledge base, be study as a result, using time domain, frequency domain and other latent structure signals unified representation, obtain letter
The recognition result of number modulation system.
The present invention also provides a kind of signal of communication modulation classification method based on deep learning, including：
The unified representation part of signal of communication, includes the following steps：
Big data and depth are excavated and are applied in complex electromagnetic environment monitoring, make monitoring process more intelligent by step 1
Change；
Step 2 by the way that signal of communication unified representation is formed database form, sends out environmental monitoring Modulation Identification frame
Changing, the problem for identifying that structure changes and change identification structure can be led to by breaking through previous increase monitoring signals；
Step 3 on existing Research foundation, utilizes the unified representation of time domain, frequency domain and other latent structure signals；
Step 4 completes signal debugging identification using the unified representation of signal, increases monitoring signals online updating classification gauge
Then, without changing identification structure offline；
Supervised learning and deep learning part, include the following steps：
Step 11, prepare data, data are pre-processed select again suitable data structure storage training data and
Test tuple；
Step 12, input mass data carry out unsupervised learning to first layer；
Step 13, clusters data by first layer, and similar data are divided into same class, are sentenced at random
It is disconnected；
Step 14 adjusts the threshold values of each node in the second layer with supervised learning, improves the input of the second layer data
Correctness；
Step 15 carries out unsupervised learning, and use unsupervised learning every time with a large amount of data to each layer network
Only one layer of training, using its training result as the input of its higher level；
Step 10 six is gone after input with supervised learning to adjust all layers.
Preferably, when signal debugging identification is carried out in the step 4, increase monitoring signals online updating classifying rules, nothing
Identification structure need to be changed offline.
The positive effect of the present invention is：The present invention has deep learning ability；Modulation recognition represent independent of
Limited learning sample；Do not have to change Modulation Identification structure during signal kinds increase, autgmentability and versatility are good；For radio
In signal monitoring, it can also be used to other to need to application fields such as signal modulation mode identification or parameter extractions.
Description of the drawings
Fig. 1 is the flow diagram of the unified representation part of signal of communication in the present invention.
Fig. 2 is supervised learning and the flow diagram of deep learning part in the present invention.
Fig. 3 is the functional block diagram of the signal of communication modulation classification system the present invention is based on deep learning.
Specific embodiment
Present pre-ferred embodiments are provided below in conjunction with the accompanying drawings, with the technical solution that the present invention will be described in detail.
Signal of communication modulation classification method the present invention is based on deep learning includes the unified representation part of signal of communication, prison
Educational inspector practises and deep learning part, wherein：
As shown in Figure 1, the unified representation part of signal of communication, includes the following steps：
Big data and depth are excavated and are applied in complex electromagnetic environment monitoring, make monitoring process more intelligent by step 1
Change；
Step 2 by the way that signal of communication unified representation is formed database form, sends out environmental monitoring Modulation Identification frame
Changing, the problem for identifying that structure changes and change identification structure can be led to by breaking through previous increase monitoring signals；
Step 3 on existing Research foundation, utilizes the unified representation of time domain, frequency domain and other latent structure signals；
Step 4 completes signal modulate using the unified representation of signal, increases monitoring signals online updating classification gauge
Then, without changing identification structure offline.
As shown in Fig. 2, supervised learning and deep learning part, include the following steps：
Step 11, prepare data, data are pre-processed select again suitable data structure storage training data and
Test tuple；
Step 12, input mass data carry out unsupervised learning to first layer；
Step 13, clusters data by first layer, and similar data are divided into same class, are sentenced at random
It is disconnected；
Step 14 adjusts the threshold values of each node in the second layer with supervised learning, improves the input of the second layer data
Correctness；
Step 15 carries out unsupervised learning, and use unsupervised learning every time with a large amount of data to each layer network
Only one layer of training, using its training result as the input of its higher level；
Step 10 six is gone after input with supervised learning to adjust all layers.
When signal debugging identification is carried out in the step 4, increase monitoring signals online updating classifying rules, without offline
Change identification structure.
As shown in figure 3, the signal of communication modulation classification system the present invention is based on deep learning includes：
Original signal collection 1, the magnanimity actual signal being collected into electromagnetic space；
Deep learning engine 2 is made inferences and learnt to input data using deep learning；
Knowledge base 3, be study as a result, using time domain, frequency domain and other latent structure signals unified representation, obtain letter
The recognition result of number modulation system.
Deep learning is derived from traditional artificial neural network, and its essence is pass through machine learning mould of the structure with more hidden layers
The training data of type and magnanimity learns more useful feature, so as to finally promote the accuracy of classification or prediction.Therefore, it is " deep
Spend model " it is means, " feature learning " is purpose.It is different from traditional shallow-layer study, the characteristics of deep learning is：1) it emphasizes
The depth of model structure, usually there is five, layer or even a hidden node of ten multilayers；2) the important of feature learning is clearly highlighted
Property, i.e., by successively eigentransformation, by sample luv space Feature Mapping to a new feature space, so as to make classification
Or prediction is more prone to.Compared with the method for artificial rule construct feature, using big data come learning characteristic, number can be more excavated
According to abundant and comprehensive internal information.The feature obtained according to deep learning can construct the unified representation of signal of communication.
The present invention, which excavates big data and depth, to be applied in complex electromagnetic environment monitoring, makes monitoring process more intelligent
Change；By by signal of communication unified representation formed database form, environmental monitoring Modulation Identification frame is made to change, break through with
Past increase monitoring signals can lead to the problem for identifying that structure changes and need change identification structure, in existing Research foundation
On, using the unified representation of time domain, frequency domain and other latent structure signals, it can complete signal debugging using the expression and know
Not, increase monitoring signals online updating classifying rules, without changing identification structure offline.The present invention monitors for radio signal
In, it can also be used to it is other to need to application fields, such as adaptive access such as signal modulation mode identification or parameter extractions.
By the existing analysis to signal of communication modulation type, fraction information known class marking signal can be obtained, and
It is more the signal of unknown class label in signal environment, therefore the signal largely without class label is added to limited has category
It trains to be learnt together in note signal, here it is the semi-supervised learning methods based on incomplete marking signal.Semi-supervised point
Training has the sample of class label with the help of the sample that class method passes through no class label, obtains than only being instructed with the sample for having class label
The classifier performance got more preferably grader, make up class label sample it is insufficient the defects of.The present invention carries out signal tune
During examination identification, increase monitoring signals online updating classifying rules, without changing identification structure offline.
Particular embodiments described above, the technical issues of to the solution of the present invention, technical solution and advantageous effect carry out
It is further described, it should be understood that the above is only a specific embodiment of the present invention, is not limited to
The present invention, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should be included in this
Within the protection domain of invention.
Claims (3)
1. a kind of signal of communication modulation classification system based on deep learning, which is characterized in that it includes：
Original signal collection, the magnanimity actual signal being collected into electromagnetic space；
Deep learning engine is made inferences and learnt to input data using deep learning；
Knowledge base, be study as a result, using time domain, frequency domain and other latent structure signals unified representation, obtain signal tune
The recognition result of mode processed.
A kind of 2. signal of communication modulation classification method based on deep learning, which is characterized in that it includes：
The unified representation part of signal of communication, includes the following steps：
Big data and depth are excavated and are applied in complex electromagnetic environment monitoring, make monitoring process more intelligent by step 1；
Step 2 by the way that signal of communication unified representation is formed database form, becomes environmental monitoring Modulation Identification frame
Change, the problem for identifying that structure changes and change identification structure can be led to by breaking through previous increase monitoring signals；
Step 3 on existing Research foundation, utilizes the unified representation of time domain, frequency domain and other latent structure signals；
Step 4 completes signal debugging identification using the unified representation of signal, increases monitoring signals online updating classifying rules, no
Structure is identified with offline change；
Supervised learning and deep learning part, include the following steps：
Step 11 prepares data, data is pre-processed and select suitable data structure storage training data and test again
Tuple；
Step 12, input mass data carry out unsupervised learning to first layer；
Step 13, clusters data by first layer, and similar data are divided into same class, are judged at random；
Step 14 adjusts the threshold values of each node in the second layer with supervised learning, improves the correct of the second layer data input
Property；
Step 15 carries out each layer network unsupervised learning, and only instructed with unsupervised learning every time with a large amount of data
Practice one layer, using its training result as the input of its higher level；
Step 10 six is gone after input with supervised learning to adjust all layers.
3. the signal of communication modulation classification method based on deep learning as claimed in claim 2, which is characterized in that the step
When signal debugging identification is carried out in four, increase monitoring signals online updating classifying rules, without changing identification structure offline.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711131003.9A CN108154164A (en) | 2017-11-15 | 2017-11-15 | Signal of communication modulation classification system and method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711131003.9A CN108154164A (en) | 2017-11-15 | 2017-11-15 | Signal of communication modulation classification system and method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108154164A true CN108154164A (en) | 2018-06-12 |
Family
ID=62468723
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711131003.9A Pending CN108154164A (en) | 2017-11-15 | 2017-11-15 | Signal of communication modulation classification system and method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108154164A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113518050A (en) * | 2021-06-24 | 2021-10-19 | 华东交通大学 | Modulation identification method, system, readable storage medium and device |
| CN113536955A (en) * | 2021-06-23 | 2021-10-22 | 中电科思仪科技股份有限公司 | Signal modulation type identification method capable of realizing continuous learning |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101667252A (en) * | 2009-10-15 | 2010-03-10 | 哈尔滨工业大学 | Classification and identification method for communication signal modulating mode based on ART2A-DWNN |
| CN102869064A (en) * | 2012-07-27 | 2013-01-09 | 南京邮电大学 | Cluster modulation identification method based on feature level and strategy level combined fusion |
| CN103793713A (en) * | 2012-10-31 | 2014-05-14 | 中国科学院沈阳自动化研究所 | Low-rank constraint online self-supervised learning scene classification method |
| US9042496B1 (en) * | 2013-02-19 | 2015-05-26 | The United States Of America, As Represented By The Secretary Of The Army | Signal modulation scheme determination through an at least fourth-order noise-insensitive cumulant |
| CN104935359A (en) * | 2015-05-12 | 2015-09-23 | 国网重庆市电力公司电力科学研究院 | Low voltage power line carrier communication signal modulation mode identification device and system |
| CN105320961A (en) * | 2015-10-16 | 2016-02-10 | 重庆邮电大学 | Handwriting numeral recognition method based on convolutional neural network and support vector machine |
| CN106780466A (en) * | 2016-12-21 | 2017-05-31 | 广西师范大学 | A kind of cervical cell image-recognizing method based on convolutional neural networks |
| CN106911603A (en) * | 2017-03-07 | 2017-06-30 | 北京工业大学 | A kind of broadband monitoring pattern Imitating signal modulation style real-time identification method |
| CN107016665A (en) * | 2017-02-16 | 2017-08-04 | 浙江大学 | A kind of CT pulmonary nodule detection methods based on depth convolutional neural networks |
| CN107122738A (en) * | 2017-04-26 | 2017-09-01 | 成都蓝色起源科技有限公司 | Automatic Communication Signals Recognition based on deep learning model and its realize system |
| CN107124381A (en) * | 2017-03-27 | 2017-09-01 | 华南理工大学 | A kind of Digital Communication Signal Modulation automatic identifying method |
-
2017
- 2017-11-15 CN CN201711131003.9A patent/CN108154164A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101667252A (en) * | 2009-10-15 | 2010-03-10 | 哈尔滨工业大学 | Classification and identification method for communication signal modulating mode based on ART2A-DWNN |
| CN102869064A (en) * | 2012-07-27 | 2013-01-09 | 南京邮电大学 | Cluster modulation identification method based on feature level and strategy level combined fusion |
| CN103793713A (en) * | 2012-10-31 | 2014-05-14 | 中国科学院沈阳自动化研究所 | Low-rank constraint online self-supervised learning scene classification method |
| US9042496B1 (en) * | 2013-02-19 | 2015-05-26 | The United States Of America, As Represented By The Secretary Of The Army | Signal modulation scheme determination through an at least fourth-order noise-insensitive cumulant |
| CN104935359A (en) * | 2015-05-12 | 2015-09-23 | 国网重庆市电力公司电力科学研究院 | Low voltage power line carrier communication signal modulation mode identification device and system |
| CN105320961A (en) * | 2015-10-16 | 2016-02-10 | 重庆邮电大学 | Handwriting numeral recognition method based on convolutional neural network and support vector machine |
| CN106780466A (en) * | 2016-12-21 | 2017-05-31 | 广西师范大学 | A kind of cervical cell image-recognizing method based on convolutional neural networks |
| CN107016665A (en) * | 2017-02-16 | 2017-08-04 | 浙江大学 | A kind of CT pulmonary nodule detection methods based on depth convolutional neural networks |
| CN106911603A (en) * | 2017-03-07 | 2017-06-30 | 北京工业大学 | A kind of broadband monitoring pattern Imitating signal modulation style real-time identification method |
| CN107124381A (en) * | 2017-03-27 | 2017-09-01 | 华南理工大学 | A kind of Digital Communication Signal Modulation automatic identifying method |
| CN107122738A (en) * | 2017-04-26 | 2017-09-01 | 成都蓝色起源科技有限公司 | Automatic Communication Signals Recognition based on deep learning model and its realize system |
Non-Patent Citations (4)
| Title |
|---|
| F. N. KHAN等: ""Joint modulation format/bit-rate classification and signal-to-noise ratio estimation in multipath fading channels using deep machine learning"", 《ELECTRONICS LETTERS》 * |
| 丁锡龙等: ""基于深度神经网络的数字信号调制类型自动识别方法"", 《湘潭大学自然科学学报》 * |
| 吴明等: ""基于高斯过程分类的调制识别方法"", 《计算机仿真》 * |
| 朱平飞等: ""基于流式感知数据的行为识别"", 《电子科学技术》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113536955A (en) * | 2021-06-23 | 2021-10-22 | 中电科思仪科技股份有限公司 | Signal modulation type identification method capable of realizing continuous learning |
| CN113536955B (en) * | 2021-06-23 | 2023-08-25 | 中电科思仪科技股份有限公司 | Signal modulation type identification method capable of realizing continuous learning |
| CN113518050A (en) * | 2021-06-24 | 2021-10-19 | 华东交通大学 | Modulation identification method, system, readable storage medium and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110162018B (en) | Incremental equipment fault diagnosis method based on knowledge distillation and hidden layer sharing | |
| CN104570083B (en) | Geologic body automatic identifying method based on multi-dimensional earthquake attribute | |
| CN109962909A (en) | A kind of network intrusions method for detecting abnormality based on machine learning | |
| CN112819059B (en) | Rolling bearing fault diagnosis method based on popular retention transfer learning | |
| CN103716204A (en) | Abnormal intrusion detection ensemble learning method and apparatus based on Wiener process | |
| Gai et al. | A Parameter‐Optimized DBN Using GOA and Its Application in Fault Diagnosis of Gearbox | |
| CN106250848A (en) | A kind of single class Acoustic Object recognition methods merged based on multi-model | |
| CN107895171A (en) | A kind of intrusion detection method based on K averages Yu depth confidence network | |
| CN109613623B (en) | Lithology prediction method based on residual error network | |
| CN112738092A (en) | Log data enhancement method, classification detection method and system | |
| CN115687925A (en) | Fault type identification method and device for unbalanced sample | |
| CN114124420A (en) | Network flow abnormity detection method based on deep neural network | |
| CN108154164A (en) | Signal of communication modulation classification system and method based on deep learning | |
| CN111626377A (en) | Lithofacies identification method, device, equipment and storage medium | |
| CN117475191A (en) | Bearing fault diagnosis method for feature alignment domain antagonistic neural network | |
| CN113837002B (en) | Small sample data fault diagnosis method based on improved TimeGan model | |
| CN110008705A (en) | A kind of recognition methods of malice domain name, device and electronic equipment based on deep learning | |
| CN117852550A (en) | Method, medium and system for electronic examination paper | |
| CN105809200A (en) | Biologically-inspired image meaning information autonomous extraction method and device | |
| Wang | Research on the fault diagnosis of mechanical equipment vibration system based on expert system | |
| CN116318925A (en) | Multi-CNN fusion intrusion detection method, system, medium, equipment and terminal | |
| CN116662899A (en) | Noise-containing data anomaly detection method based on self-adaptive strategy | |
| CN111079809A (en) | Intelligent system method for electric connector | |
| CN110490218A (en) | A kind of rolling bearing fault self-learning method based on two-stage DBN | |
| Zhang | The WSN intrusion detection method based on deep data mining |
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: 20180612 |