CN112580598B - Radio signal classification method based on multichannel Diffpool - Google Patents
Radio signal classification method based on multichannel Diffpool Download PDFInfo
- Publication number
- CN112580598B CN112580598B CN202011605253.3A CN202011605253A CN112580598B CN 112580598 B CN112580598 B CN 112580598B CN 202011605253 A CN202011605253 A CN 202011605253A CN 112580598 B CN112580598 B CN 112580598B
- Authority
- CN
- China
- Prior art keywords
- channel
- diffpool
- data
- signal
- network
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000010586 diagram Methods 0.000 claims abstract description 43
- 239000013598 vector Substances 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000013145 classification model Methods 0.000 claims abstract description 14
- 230000004913 activation Effects 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 2
- 230000000007 visual effect Effects 0.000 description 10
- 238000013507 mapping Methods 0.000 description 9
- 230000006870 function Effects 0.000 description 7
- 230000006855 networking Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000007476 Maximum Likelihood Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Mobile Radio Communication Systems (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
A multi-channel Diffpool based radio signal classification method, comprising: acquiring and preprocessing a radio modulation signal data set, and splitting each signal sample into data of two channels with equal length; constructing two one-dimensional convolution layers with the same parameter setting but different weights, respectively removing the data of the two channels obtained after pretreatment to obtain the convolved channel data of the two channels, processing the convolved channel data by using a nonlinear activation function to obtain the activated channel data of the two channels, and obtaining two network diagrams according to the two activated channel data; modifying a graph classification model Diffpool in the field of network graphs, and respectively processing the two network graphs obtained by using modified models with different initialization weights in two defined channels to obtain two corresponding one-dimensional feature vectors; and splicing the two one-dimensional feature vectors into one-dimensional feature vector, and processing the obtained feature vector by using the two full-connection layers to obtain a final classification result.
Description
Technical Field
The invention relates to data mining, network science and data analysis technology, in particular to a radio signal classification method based on multichannel Diffpool.
Background
The essence of the radio signal classification is to distinguish the modulation patterns to which the different signals belong, so that the radio modulated signal is demodulated to obtain the original signal with the useful information. Modulation identification of radio signals is an important component in the communication field, and in order to realize rapid and safe propagation of messages by means of channels under the condition of fully utilizing channel capacity, baseband signals are generally not directly transmitted, but are required to be modulated to obtain modulated signals, so that the modulated signals have more effective spectrum utilization and higher communication rate. With the rapid development of radio communication technology at present, the electromagnetic environment becomes more complex, and it is more important to correctly and efficiently identify the modulation type to which the modulated signal belongs, which is critical in civil and military aspects, such as spectrum management, electronic countermeasure, and communication reconnaissance.
On the radio signal classification problem, conventional classification methods include a modulation recognition method based on feature extraction and a maximum likelihood hypothesis test method. The modulation recognition method based on feature extraction requires manual calculation and extraction of a large number of features, such as histogram features, statistical moment features, transform domain features and the like, and classification is performed by utilizing a classifier in the machine learning field. The other maximum likelihood hypothesis test rule is to process likelihood functions of the radio signals to obtain statistics for comparison with a threshold, and finally realize classification of the modulation signals.
With the development of deep learning, convolutional neural networks are currently used for classifying modulation signals, so that good effects are achieved. In addition, some visual networking methods are provided for effectively combining signals and network diagrams, and signal data is mapped into the network diagrams, so that the classification and identification of radio signals are realized through the diagram classification technology in the field of the network diagrams, but the mapping technology through the visual networking method is inflexible at present, the obtained network diagrams can not be ensured to retain important information in original signals, and more time is required to be consumed in the mapping process.
At present, a method for mapping signals into network diagrams and then classifying the network diagrams by a visual networking method exists, but most of the methods have complex processes, the mapping process is dead, and the network diagrams obtained by mapping cannot be guaranteed to better represent original radio signals. For example, a radar antenna scanning mode identification method based on limited crossing visual view is disclosed in patent application number CN202010566006.0, the method converts signals into a network diagram through a visual networking method LPVG, then network characteristic parameters are extracted, and classification is carried out through a support vector machine SVM. In contrast, the method provided by the invention does not need to artificially convert signals or manually calculate some characteristic parameters of the network diagram, can realize an end-to-end classification framework by splitting the original radio signal into two channels of data, can simultaneously and autonomously train the processes of mapping to obtain the network diagram, extracting the characteristic vector of the network diagram and classifying the characteristic vector, simplifies the classification process, does not require the support of corresponding expertise, and can improve the classification precision.
Disclosure of Invention
The existing visual networking method generally maps signals into network diagrams according to fixed rules, and then classifies the network diagrams to realize signal classification. The invention provides an end-to-end multi-channel Diffpool-based radio signal classification method, which is used for resolving original radio signal data into data of two channels I and Q, respectively processing the data of the channels I and Q by using two different one-dimensional convolution layers and ReLU activation functions to obtain two corresponding feature matrixes, thereby obtaining two network graphs G I And G Q Respectively processing the network graph G by using two network graph classification models Diffpool with the last full-connection layer removed I And G Q The corresponding one-dimensional feature vector is obtained, and is spliced into one-dimensional feature vector and then classified by using two full-connection layers, so that the classification of radio signal data is completed, signals can be flexibly and efficiently mapped into a network diagram through autonomous training, the classification process is simplified, and the classification precision can be improved.
In order to achieve the above object, the present invention provides the following solutions: the invention provides a radio signal classification method based on multichannel Diffpool, which comprises the following steps:
s1, acquiring a radio modulation signal data set and preprocessing the radio modulation signal data set; the pretreatment process comprises the following steps: each signal sample is split into data for two channels of equal length.
S2, constructing two one-dimensional convolution layers with the same parameter setting and different weights, respectively removing the data of the two channels obtained after pretreatment to obtain the convolved channel data of the two channels, processing the convolved channel data by using a nonlinear activation function to obtain the activated channel data of the two channels, and obtaining two network diagrams according to the two activated channel data;
s2, the number of input channels of the two one-dimensional convolution layers is 1, the number of output channels is 128, the convolution kernel size is 3, the step length is 1, and the padding value is 1;
the nonlinear activation function used is a ReLU.
S3, modifying a graph classification model Diffpool in the field of network graphs, and respectively processing the two network graphs obtained in the step S2 by using modified models with different initialization weights in two constructed channels to obtain two corresponding one-dimensional feature vectors;
the specific process of modifying the network map classification model Diffpool in step S3 is as follows: and removing the last full-connection layer used for classification in the network diagram classification model Diffpool, and reserving other network layers such as a diagram pooling layer and the like to obtain a modified model.
S4, splicing the two one-dimensional feature vectors into a one-dimensional feature vector, and processing the obtained feature vector by using the two full-connection layers to obtain a final classification result.
Preferably, the specific way of selecting and preprocessing the radio modulated signal data set in step S1 is:
using the disclosed radio modulated signal data set rml2016.10a, which has a total of 11 modulation types, CPFSK, PAW4, 8PSK, AM-DSB, AM-SSB, BPSK, GFSK, QAM, QAM64, QPSK, and WBFM, each modulation type has 20 signal-to-noise ratios, each signal-to-noise ratio has 1000 samples, wherein each signal sample has a shape of 2x128, i.e., L is 128, and each signal sample is composed of I-channel data and Q-channel data, from which I-channel and Q-channel data are extracted, respectively denoted as i= { I 1 ，I 2 ，…，I n ，…，I 128 Sum q= { Q 1 ，Q 2 ，…，Q n ，…，Q 128 }。
Preferably, the one-dimensional convolution layer Conv1D described in step S2 I And Conv1D Q Is 1, the number of output channels is 128, the convolution kernel size is 3, the step size is 1, the padding value is 1, and Conv1D I And Conv1D Q The non-linear activation function selected is a ReLU, which has different initialization weights.
Preferably, the specific procedure of modifying the network map classification model Diffpool in step S3 is as follows: and removing the last full-connection layer used for classification in the network diagram classification model Diffpool, and reserving other network layers such as a diagram pooling layer.
Preferably, the entire radio classification framework is end-to-end, and the convolutional layer, modified Diffpool, and full-connectivity layers used in the process can be trained together.
The invention has the following advantages:
the classification method fully utilizes the self-learning capability of the neural network, enables the signals to be mapped into the proper network diagram in a learning way through the one-dimensional convolution layer, fully reserves the implicit characteristic information in the original radio signal in the mapping process, then modifies the classical model Diffpool used for classifying the network diagram in the network diagram field into a model capable of classifying the multi-channel network diagram, finally forms an end-to-end classification frame, avoids the complex operation of manually extracting the characteristics, and has high efficiency, flexible mapping method, reduced operation complexity and improved classification precision compared with the visual network construction method for converting the signals into the network diagram according to fixed rules.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of the present invention;
fig. 2 is a schematic diagram of a process for classifying radio signals by using a one-dimensional convolution layer and a modified multi-channel Diffpool model of the method of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the present invention provides a radio signal classification method based on multi-channel Diffpool, comprising the steps of:
s1, preprocessing a radio modulation signal with each shape of 2xL, wherein each signal sample comprises data of two channels I and Q, and splitting the data into data of the channels I and Q with each shape of 1 xL.
The present embodiment uses the disclosed radio modulated signal data set rml2016.10a, which has a total of 11 modulation types, CPFSK, PAW4, 8PSK, AM-DSB, AM-SSB, BPSK, GFSK, QAM, QAM64, QPSK, and WBFM, each of which has 20 signal-to-noise ratios, each of which has 1000 samples, where each signal sample has a shape of 2x128, i.e., L is 128.
Let each signal sample be represented as:
S＝{(I 1 ，Q 1 )，(I 2 ，Q 2 )，…，(I n ，Q n )，…，(I 128 ，Q 128 )} (1)
where 128 denotes the length of the signal, i.e. the number of time points, I n And Q n Respectively represent the I channel and the Q channel at the nth timeThe signal values corresponding to the intermediate points are extracted to obtain data of an I channel and a Q channel, which are respectively expressed as I= { I 1 ，I 2 ，…，I n ，…，I 128 Sum q= { Q 1 ，Q 2 ，…，Q n ，…，Q 128 }。
S2, constructing two different one-dimensional convolution layers Conv1D I And Conv1D Q Processing the I channel data and the Q channel data respectively to obtain convolution processed channel data I 'and Q' with the corresponding shapes LxL, processing the I 'channel data and the Q' channel data by using a nonlinear activation function to obtain activated channel data I 'and Q' with the corresponding shapes LxL, and taking the activated channel data I 'and Q' as feature matrixes capable of representing the network diagram to obtain a corresponding network diagram G I And G Q 。
As shown in FIG. 2, this embodiment constructs a one-dimensional convolution layer Conv1D I Processing I channel data with the shape of 1x128, wherein the number of input channels of the one-dimensional convolution layer is 1, the number of output channels is 128, the convolution kernel size is 3, the step length is 1, the padding value is 1, a matrix I' with the shape of 128x128 can be obtained, and when Q channel data are processed, a one-dimensional convolution layer Conv1D with the same parameter setting and different weight values is also constructed Q A matrix Q' of 128x128 shape is obtained.
Processing the matrix I 'and the matrix Q' with a nonlinear activation function ReLU to obtain feature matrices I 'and Q' each having a shape of 128x128, and obtaining a corresponding network graph G based on the feature matrices I 'and Q' I And G Q 。
S3, modifying a diagram classification model Diffpool in the network diagram field, constructing two channels, and processing a network diagram G in the first channel by using the diagram classification model Diffpool in the network diagram field I Processing network graph G using Diffpool of different parameters in the second channel Q Respectively extracting to obtain network diagram G I And G Q Feature vector F of (1) I And F Q The shapes are all 1xK.
When the network diagram classification model Diffpool is modified, the last full connection used for classification in the network diagram classification model Diffpool is removedThe network graph G is then processed by the remainder I Obtaining a one-dimensional characteristic vector F with the length of K I Processing the network map G in the same way by using Diffpool with different parameters Q Obtaining a one-dimensional characteristic vector F with the length of K Q 。
S4, as shown in FIG. 2, the feature vector F is spliced I And F Q The feature vector F with the shape of 1x2K is obtained, the feature vector F is processed by two full connection layers for classification, the whole radio classification framework is end-to-end, and a convolution layer, a Diffpool layer and the full connection layer used in the process can be trained together.
The classification accuracy of the method provided by the invention can reach 56.57% in the disclosed radio signal data set RML2016.10a, and compared with the method for establishing the network by the limited-crossing visual view provided in the paper 'time sequence network model based on the limited-crossing visual view', the classification accuracy obtained by the method for establishing the network by the visual view according to the fixed conversion rule is superior to that of the same data set. The invention has high efficiency, flexible mapping method, reduced operation complexity and improved classification accuracy of radio signal data.
The above embodiments are only illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope defined by the claims of the present invention without departing from the design spirit of the present invention.
Claims (5)
1. A multi-channel Diffpool based radio signal classification method comprising the steps of:
s1, selecting a radio modulation signal data set as a sample, preprocessing a radio modulation signal, wherein the shape of each signal sample is 2xL, and the signal sample comprises data of two channels I and Q, and is split into data of an I channel and data of a Q channel, wherein the shape of the data of the I channel and the data of the Q channel are both 1 xL;
let each signal sample be represented as:
S＝{(I 1 ,Q 1 ),(I 2 ,Q 2 ),…,(I n ,Q n ),…,(I L ,Q L )} (1)
wherein L represents the length of the signal, i.e. the number of time points, I n And Q n Respectively representing signal values corresponding to the I channel and the Q channel at the nth time point, extracting data of the I channel and the Q channel from the signal values, and respectively representing the data as I= { I 1 ,I 2 ,…,I n ,…,I L Sum q= { Q 1 ,Q 2 ,…,Q n ,…,Q L }；
S2, using two different one-dimensional convolution layers Conv1D I And Conv1D Q Processing the I channel data and the Q channel data respectively to obtain convolution processed channel data I 'and Q' with the corresponding shapes LxL, processing the I 'channel data and the Q' channel data by using a nonlinear activation function to obtain activated channel data I 'and Q' with the corresponding shapes LxL, and taking the activated channel data I 'and Q' as feature matrixes capable of representing the network diagram to obtain a corresponding network diagram G I And G Q ；
Constructing a one-dimensional convolutional layer Conv1D for processing I-channel data I The number of input channels of the one-dimensional convolution layer is 1, the number of output channels is L, the convolution kernel size is 3, the step length is 1, the padding value is 1, a matrix I' with the shape of LxL is obtained, and when Q channel data are processed, a one-dimensional convolution layer Conv1D with the same parameter setting and different weight values is also constructed Q Obtaining a matrix Q' with the shape of LxL;
processing the matrix I 'and the matrix Q' by using a nonlinear activation function to obtain feature matrices I 'and Q' with the shapes of LxL, and obtaining a corresponding network graph G according to the feature matrices I 'and Q' I And G Q ；
S3, modifying a diagram classification model Diffpool in the field of network diagrams, constructing two channels, and processing a network diagram G in the first channel by using the modified Diffpool I Processing the network graph G in a second channel using modified Diffpool of different parameters Q Respectively extracting network graphs G I And G Q Feature vector F of (1) I And F Q The shapes are all 1xK;
modifying the Diffpool model of the network graph classification modelThe last full connection layer in Diffpool is removed for classification and then the rest is used to process the network graph G I Obtaining a one-dimensional characteristic vector F with the length of K I Processing the network map G in the same manner with modified Diffpool of different parameters Q Obtaining a one-dimensional characteristic vector F with the length of K Q ；
S4, splicing the feature vectors F I And F Q And obtaining a feature vector F with the shape of 1x2K, processing the feature vector F by using two full-connection layers to classify, and training a convolution layer, a Diffpool and the full-connection layers used in the process together.
2. The multi-channel Diffpool based radio signal classification method of claim 1, wherein: the specific way of selecting and preprocessing the radio modulated signal data set in step S1 is:
using the disclosed radio modulated signal data set rml2016.10a, which has a total of 11 modulation types, CPFSK, PAW4, 8PSK, AM-DSB, AM-SSB, BPSK, GFSK, QAM, QAM64, QPSK, and WBFM, each modulation type has 20 signal-to-noise ratios, each signal-to-noise ratio has 1000 samples, wherein each signal sample has a shape of 2x128, i.e., L is 128, and each signal sample is composed of I-channel data and Q-channel data, from which I-channel and Q-channel data are extracted, respectively denoted as i= { I 1 ,I 2 ,…,I n ,…,I 128 Sum q= { Q 1 ,Q 2 ,…,Q n ,…,Q 128 }。
3. The multi-channel Diffpool based radio signal classification method according to claim 1, wherein:
step S2 describes a one-dimensional convolution layer Conv1D I And Conv1D Q Is 1, the number of output channels is 128, the convolution kernel size is 3, the step size is 1, the padding value is 1, and Conv1D I And Conv1D Q The non-linear activation function selected is a ReLU, which has different initialization weights.
4. The multi-channel Diffpool based radio signal classification method according to claim 1, wherein:
the specific process of modifying the network map classification model Diffpool in step S3 is as follows: and removing the last full-connection layer used for classification in the network diagram classification model Diffpool, and reserving other network layers.
5. The multi-channel Diffpool based radio signal classification method according to claim 1, wherein:
the whole radio classification framework is end-to-end, and the convolutional layer, modified Diffpool and full-connectivity layer used in the process can be trained together.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011605253.3A CN112580598B (en) | 2020-12-30 | 2020-12-30 | Radio signal classification method based on multichannel Diffpool |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011605253.3A CN112580598B (en) | 2020-12-30 | 2020-12-30 | Radio signal classification method based on multichannel Diffpool |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112580598A CN112580598A (en) | 2021-03-30 |
| CN112580598B true CN112580598B (en) | 2024-02-13 |
Family
ID=75144227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011605253.3A Active CN112580598B (en) | 2020-12-30 | 2020-12-30 | Radio signal classification method based on multichannel Diffpool |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112580598B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113392731B (en) * | 2021-05-31 | 2023-06-23 | 浙江工业大学 | Modulation signal classification method and system based on graph neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
| CN112069883A (en) * | 2020-07-28 | 2020-12-11 | 浙江工业大学 | Deep learning signal classification method fusing one-dimensional and two-dimensional convolutional neural network |
-
2020
- 2020-12-30 CN CN202011605253.3A patent/CN112580598B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
| CN112069883A (en) * | 2020-07-28 | 2020-12-11 | 浙江工业大学 | Deep learning signal classification method fusing one-dimensional and two-dimensional convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于卷积神经网络的通信信号调制识别研究;杨洁;夏卉;;计算机测量与控制(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112580598A (en) | 2021-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107979554B (en) | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks | |
| CN107124381B (en) | Automatic identification method for digital communication signal modulation mode | |
| CN112702294B (en) | Modulation recognition method for multi-level feature extraction based on deep learning | |
| CN111314257B (en) | Modulation mode identification method based on complex value neural network | |
| CN110113288B (en) | Design and demodulation method of OFDM demodulator based on machine learning | |
| CN110569752A (en) | convolutional neural network-based radar signal category determination method | |
| CN112069883B (en) | Deep learning signal classification method integrating one-dimensional two-dimensional convolutional neural network | |
| CN113014524B (en) | Digital signal modulation identification method based on deep learning | |
| CN114157539B (en) | Data-aware dual-drive modulation intelligent identification method | |
| CN110659684A (en) | Convolutional neural network-based STBC signal identification method | |
| CN114553648A (en) | Wireless communication modulation mode identification method based on space-time diagram convolutional neural network | |
| CN112580598B (en) | Radio signal classification method based on multichannel Diffpool | |
| CN115186712A (en) | Modulated signal identification method and system | |
| Lin et al. | A real-time modulation recognition system based on software-defined radio and multi-skip residual neural network | |
| CN114925720A (en) | Small sample modulation signal identification method based on space-time mixed feature extraction network | |
| CN114595729A (en) | Communication signal modulation identification method based on residual error neural network and meta-learning fusion | |
| CN114615118A (en) | Modulation identification method based on multi-terminal convolution neural network | |
| CN114398931A (en) | Modulation recognition method and system based on numerical characteristic and image characteristic fusion | |
| CN115955375A (en) | Modulated signal identification method and system based on CNN-GRU and CA-VGG feature fusion | |
| CN115086123A (en) | Modulation identification method and system based on fusion of time-frequency graph and constellation diagram | |
| CN113902095A (en) | Automatic modulation identification method, device and system for wireless communication | |
| CN114584441A (en) | Digital signal modulation identification method based on deep learning | |
| CN114419341A (en) | Convolutional neural network image identification method based on transfer learning improvement | |
| Xue et al. | Radio signal modulation recognition algorithm based on convolutional neural network | |
| Zhang et al. | Modulation format recognition based on CNN in satellite communication system |
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 |