CN114266312B - Method for identifying radio frequency fingerprint and communication protocol based on multitask learning - Google Patents
Method for identifying radio frequency fingerprint and communication protocol based on multitask learning Download PDFInfo
- Publication number
- CN114266312B CN114266312B CN202111582991.5A CN202111582991A CN114266312B CN 114266312 B CN114266312 B CN 114266312B CN 202111582991 A CN202111582991 A CN 202111582991A CN 114266312 B CN114266312 B CN 114266312B
- Authority
- CN
- China
- Prior art keywords
- task
- communication protocol
- differential
- signal
- wireless communication
- 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
- 238000004891 communication Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000010586 diagram Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 230000005855 radiation Effects 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 239000003086 colorant Substances 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention discloses a recognition method of radio frequency fingerprints and communication protocols based on multitask learning, which comprises the following steps: 1) Acquiring a receiver-side signal; 2) Preprocessing signals; 3) Performing differential processing on the I/Q signals to obtain a differential constellation track diagram: 4) Making an input data set of the neural network; 5) Multitasking neural network training; 6) And carrying out radio frequency fingerprint identification and wireless communication protocol identification. The method can improve the recognition accuracy and shorten the training time on the basis of a single task, and the differential constellation locus diagram can eliminate the condition of constellation diagram rotation caused by frequency offset.
Description
Technical Field
The invention relates to a wireless communication physical layer security technology, in particular to a method for identifying radio frequency fingerprints and communication protocols based on multitask learning.
Background
The wireless local area network technology is widely applied in the communication field, and with the popularization of wireless communication, potential safety hazards such as illegal interception and information modification exist in wireless signals, and great influence is easily generated for users. Therefore, in order to secure information security, it is very important to effectively identify the radio frequency fingerprint of the wireless signal and the physical layer network information protocol.
With the development of technology, a deep learning mode is more and more popular, and compared with a traditional manual feature classification mode, the deep learning method improves classification accuracy and saves manpower and material resources. Compared with the traditional single-task method, the multi-task deep learning method has higher recognition precision, shortens training time and improves efficiency.
Different radiation source devices have different internal hardware during production, so that each device has unique radio frequency fingerprints, and the identification of the radio frequency fingerprints can be realized by analyzing the received wireless signals and extracting fine characteristics. The physical layer of the wireless local area network takes IEEE802.11 series standards as wireless communication protocol specifications, and the modulation techniques of different wireless communication protocols are different. In the 2.4GHz working frequency band, the IEEE802.11b protocol mainly adopts a Complementary Code Keying (CCK) modulation mode, and the IEEE802.11g protocol mainly adopts an OFDM technology. The constellation of the signal includes the characteristic of the steady state part, but the constellation obtained directly on the complex plane can rotate due to carrier frequency offset, and the rotation angles are different at different moments. The constellation diagram after differential processing can make up for frequency offset to obtain a stable constellation diagram, has the characteristics of radio frequency fingerprints and wireless communication protocols, and can effectively carry out multi-task identification and classification.
Disclosure of Invention
The invention aims at overcoming the defects of the prior art and provides a method for identifying radio frequency fingerprints and communication protocols based on multitask learning. The method adopts a differential constellation track diagram and a multi-task neural network structure, can effectively identify the radio frequency fingerprint and the wireless communication protocol, can improve the identification precision and shorten the training time, and can be applied to the fields of non-cooperative reception such as information security of wireless signals.
The technical scheme for realizing the aim of the invention is as follows:
the recognition method of the radio frequency fingerprint and the communication protocol based on the multitask learning is different from the prior art in that the recognition method comprises the following steps:
1) Acquiring a receiver-side signal: taking a Wifi signal generated by a router as a transmitting end signal, and taking USRP as a receiving end to perform oversampling reception;
2) Signal pretreatment: preprocessing the received I/Q signal, screening out useful data by Matlab, namely assuming that the received data is a sequence with length L, randomly selecting n segments of data and calculating average power P as shown in formula (1) n Selecting a proper threshold according to the signal waveform, and if the average power of the n sections of signals is greater than the threshold, marking the signals as useful signals X= { X (0), X (1), … and X (n-1) }; if the signal is not greater than the threshold value, continuing screening in a sliding window counter mode for useless signals:
P n ＝|f(t)| 2 (1)；
3) Drawing a differential constellation trace diagram of a complex sequence { x (0), x (1), …, x (L-1) }: the process is as follows:
3-1) carrying out energy normalization processing on the signals, so as to eliminate the influence generated by different powers of the transmitting end equipment;
3-2) assume that the baseband signal transmitted by the transmitter is x (t), and the transmitter frequency is noted as f c Then the expression of the transmitted signal is as shown in equation (2):
s(t)＝x(t)e -j2πfct (2)，
3-3) assume the carrier frequency of the receiver is f R ，For receiving the phase error of the resulting signal, Δf=f due to errors in the carrier frequencies of the transmitter and receiver R -f c The baseband signal obtained after the receiver performs down-conversion processing is denoted as y (t), and the expression of y (t) is shown in formula (3):
3-4) performing differential processing on y (t) by using a formula (4), wherein the differential interval is m:
W(t)＝y(t)·y*(t+m)＝x(t)·x(t+m)e -j2πΔfm (4)；
3-5) drawing the signals subjected to differential processing on a complex plane by using a scattepplot function of Matlab;
3-6) finally, carrying out scattered point density treatment: according to the fact that the received signals are affected by different wireless communication protocols and equipment differences, the shapes of drawn differential constellation track diagrams and the offset angles of differential result gathering areas are different, and the scattered point density diagrams show different colors according to gathered points.
4) An input dataset for a neural network, comprising:
4-1) preparing 7200 differential constellation track diagrams obtained by Matlab into a data set, wherein 5400 differential constellation track diagrams are training sets, and 1800 differential constellation track diagrams are test sets;
4-2) performing multi-label processing on the data set: marking each picture in the data set obtained in the step 4-1) as two attribute labels, namely a wireless communication protocol label and a radiation source equipment label;
4-3) performing label binarization processing: namely, binarizing the data set label into a [ 01 0 … 0] form and modifying the size of the picture into 64 multiplied by 64;
5) Multitasking neural network training: the whole multitasking neural network training adopts 'relu' as an activation function, which comprises the following steps:
5-1) constructing a multi-task convolutional neural network in a hard sharing mode, and training the multi-task convolutional neural network in a multi-label mode, namely, in a mode of one input and multiple outputs;
5-2) construction with reference to the VGG16 model: the hard sharing mode described in the step 5-1) has 3 sharing blocks, and the first block consists of 2 layers of 64×3×3 convolution layers; the second block consists of 2 layers of 128 x 3 convolutional layers; the third block consists of 3 layers of 256×3×3 convolutional layers and employs a maximum pooling layer of size 2×2;
5-3) in step 5-2), a BN layer and a dropout layer are adopted after each conv layer to prevent overfitting, the dropout parameter is 0.4, and the whole process adopts 'relu' as an activation function;
5-4) after the two tasks in the step 5-2) share network structures, parameters and characteristics, each task is identified by adopting an independent network layer, wherein the wireless communication protocol identification task adopts two full-connection layers, the parameters of the first full-connection layer are 1024, and the parameters of the second layer are 2, namely the corresponding identified types;
5-5) the radiation source identification task is different from the network information identification task in the step 5-4), and the parameters of the second layer full-connection layer of the radiation source identification task are 6, namely the radiation sources corresponding to 6 types;
5-6) saving the trained model weights;
5-6) the wireless communication protocol identification task and the radiation source identification task employ a softmax classifier and a cross entropy loss function, i.e., as shown in equation (5):
wherein q is a predictive tag, and p is a real tag;
5-7) the size of the parameter Batchsize is set to 20, the optimizer selects Adam optimizer, and the learning rate is set to 0.0001;
6) The weight coefficients of the wireless communication protocol identification task and the radio frequency fingerprint identification task are set to be 1.0, the total loss of the network is the sum of the loss of the radiation source identification task and the loss of the wireless communication protocol identification, and the total loss of the network corresponds to the formula (6):
where k is the number of tasks.
7) And storing the trained multi-task neural network and carrying out radio frequency fingerprint identification and wireless communication protocol identification.
The technical scheme fully utilizes the advantages of multi-task deep learning: the multiple tasks share the characteristics, so that the recognition precision is improved on the basis of a single task, and the multiple tasks are trained simultaneously, so that the training time can be effectively shortened, and the efficiency is improved; the differential constellation diagram of the signal is adopted as the characteristic, so that the condition of constellation diagram rotation caused by carrier frequency deviation is eliminated theoretically; after differential processing, the offset angles of the differential result gathering areas of different devices are different due to the difference of hardware between the different devices, and the radiation source can be identified by utilizing the fine characteristics; the points after the difference of different wireless communication protocols correspond to a limited number of positions on a complex plane, and as the different wireless communication protocols contain different modulation modes, the positions corresponding to the difference points of each wireless communication protocol are different, so that the identification of the wireless communication protocol can be performed, and if a new radiation source or the wireless communication protocol appears, the neural network can be trained again.
The method adopts a differential constellation track diagram and a multi-task neural network structure, can effectively identify the radio frequency fingerprint and the wireless communication protocol, can improve the identification precision and shorten the training time, and can be applied to the fields of non-cooperative reception such as information security of wireless signals.
Drawings
FIG. 1 is a schematic flow chart of a method of an embodiment;
fig. 2 is a diagram of differential constellation traces of the same device ieee802.11b and ieee802.11g protocols in an embodiment;
FIG. 3 is a diagram of differential constellation traces of different radiation sources for the same communication protocol in an embodiment;
FIG. 4 is a diagram of a hard-shared multi-tasking neural network in an embodiment;
fig. 5 is a schematic diagram of a VGG16 module structure modified in the embodiment.
Detailed Description
The present invention will now be further illustrated with reference to the drawings and examples, but is not limited thereto.
Examples:
referring to fig. 1, a method for identifying a radio frequency fingerprint and a communication protocol based on multitasking learning includes the following steps:
1) Acquiring a receiver-side signal: taking a Wifi signal generated by a router as a transmitting end signal, and taking USRP as a receiving end to perform oversampling reception;
2) Signal pretreatment: preprocessing the received I/Q signal, screening out useful data by Matlab, namely assuming that the received data is a sequence with length L, randomly selecting n sections of data and calculating average power P as shown in formula (1) n Selecting a proper threshold according to the signal waveform, and if the average power of the n sections of signals is greater than the threshold, marking the signals as useful signals X= { X (0), X (1), … and X (L-1) }; if the signal is not greater than the threshold value, continuing screening in a sliding window counter mode for useless signals:
P l ＝|f(t)| 2 (1)；
3) Drawing a differential constellation trace diagram of a complex sequence { x (0), x (1), …, x (L-1) }: the differential constellation track diagrams shown in fig. 2 and 3 are drawn, and the process is as follows:
3-1) carrying out energy normalization processing on the signals, so as to eliminate the influence generated by different powers of the transmitting end equipment;
3-2) assume that the baseband signal transmitted by the transmitter is x (t), and the transmitter frequency is noted as f c Then the expression of the transmitted signal is as shown in equation (2):
s(t)＝x(t)e -j2πfct (2)，
3-3) assume the carrier frequency of the receiver is f R ，For receiving the phase error of the resulting signal, Δf=f due to errors in the carrier frequencies of the transmitter and receiver R -f c The signal obtained after the receiver performs the down-conversion process is denoted as y (t), and the expression of y (t) is shown in formula (3):
y(t)＝x(t)e j2πΔft+ (3)；
3-4) performing differential processing on y (t) by using a formula (4), wherein the differential interval is m:
W(t)＝y(t)·y*(t+m)＝x(t)·x(t+m)e -j2πΔfm (4)；
3-5) drawing the signals subjected to differential processing on a complex plane by using a scattepplot function of Matlab;
3-6) finally, carrying out scattered point density treatment: according to the fact that the received signals are affected by different wireless communication protocols and equipment differences, the drawn differential constellation diagram shape and the differential result gathering area offset angle are different, and the scattered point density diagram reflects different colors according to the gathered points.
4) An input dataset for a neural network, comprising:
4-1) preparing 7200 differential constellation track diagrams obtained by Matlab into a data set, wherein 5400 differential constellation track diagrams are training sets, and 1800 differential constellation track diagrams are test sets;
4-2) performing multi-label processing on the data set: marking each picture in the data set obtained in the step 4-1) as two attribute labels, namely a wireless communication protocol label and a radiation source label;
4-3) performing label binarization processing: namely, binarizing the data set label into a [ 01 0 … 0] form and modifying the size of the picture into 64 multiplied by 64;
5) Multitasking neural network training: the hard sharing mode is adopted, which comprises the following steps:
5-1) as shown in fig. 4, constructing a multi-task convolutional neural network in a hard sharing mode, and training the multi-task neural network in a multi-label mode, namely, in a form of one input and multiple outputs;
5-2) construction with reference to the VGG16 model: the improved VGG16 is shown in FIG. 5, wherein 3 shared blocks are shared in the hard sharing mode in the step 5-1), and the first block consists of 2 layers of 64×3×3 convolution layers; the second block consists of 2 layers of 128 x 3 convolutional layers; the third block consists of 3 layers of 256×3×3 convolutional layers and employs a maximum pooling layer of size 2×2;
5-3) in step 5-2), a BN layer and a dropout layer are adopted after each conv layer to prevent overfitting, the dropout parameter is 0.4, and the whole process adopts 'relu' as an activation function;
5-4) after the two tasks in the step 5-2) share network structures, parameters and characteristics, each task is identified by adopting an independent network layer, wherein the wireless communication protocol identification task adopts two full-connection layers, the parameters of the first full-connection layer are 1024, and the parameters of the second layer are 2, namely the corresponding identified types;
5-5) the radiation source identification task is different from the network information identification task in the step 5-4), and the parameters of the second layer full-connection layer of the radiation source identification task are 6, namely the radiation sources corresponding to 6 types;
5-6) saving the trained model weights;
5-6) the wireless communication protocol identification task and the radiation source identification task employ a softmax classifier and a cross entropy loss function, i.e., as shown in equation (5):
wherein q is a predictive tag, and p is a real tag;
5-7) the size of the parameter Batchsize is set to 20, the optimizer selects Adam optimizer, and the learning rate is set to 0.0001;
6) The weight coefficients of the wireless communication protocol identification task and the radio frequency fingerprint identification task are set to be 1.0, the total loss of the network is the sum of the loss of the radiation source identification task and the loss of the wireless communication protocol identification, and the total loss of the network corresponds to the formula (6):
where k is the number of tasks.
7) And storing the trained multi-task neural network and carrying out radio frequency fingerprint identification and wireless communication protocol identification.
Claims (1)
1. The recognition method of the radio frequency fingerprint and the communication protocol based on the multitask learning is characterized by comprising the following steps:
1) Acquiring a receiver-side signal: taking a Wifi signal generated by a router as a transmitting end signal, and taking USRP as a receiving end to perform oversampling reception;
2) Signal pretreatment: preprocessing the received I/Q signal, screening out useful data by Matlab, namely assuming that the received data is a sequence with length L, randomly selecting n sections of data and calculating average power P as shown in formula (1) n Selecting a threshold according to the signal waveform, and if the average power of the n sections of signals is greater than the threshold, marking the signals as useful signals X= { X (0), X (1), … and X (L-1) }; if the signal is not greater than the threshold value, continuing screening in a sliding window counter mode for useless signals:
P l ＝|f(t)| 2 (1)；
3) Drawing a differential constellation trace diagram of a complex sequence { x (0), x (1), …, x (L-1) }: the process is as follows:
3-1) carrying out energy normalization processing on the signals, so as to eliminate the influence generated by different powers of the transmitting end equipment;
3-2) assume that the baseband signal transmitted by the transmitter is x (t), and the transmitter frequency is noted as f c Then the expression of the transmitted signal is as shown in equation (2):
s(t)＝x(t)e -j2πfct (2)，
3-3) assume the carrier frequency of the receiver is f R ，To receive the phase error of the resulting signal, let Δf=f R -f c The signal obtained after the receiver performs the down-conversion process is denoted as y (t), and the expression of y (t) is shown in formula (3):
3-4) performing differential processing on y (t) by using a formula (4), wherein the differential interval is m:
W(t)＝y(t)·y * (t+m)＝x(t)·x(t+m)e -j2πΔfm (4)；
3-5) drawing the signals subjected to differential processing on a complex plane by using a scattepplot function of Matlab;
3-6) carrying out scattered point density treatment: according to the fact that received signals are affected by different wireless communication protocols and equipment differences, the drawn differential constellation diagram shape and the difference of differential result gathering area offset angles are different, and the scattered point density diagram reflects different colors according to gathered points;
4) An input dataset for a neural network, comprising:
4-1) preparing 7200 differential constellation track diagrams obtained by Matlab into a data set, wherein 5400 differential constellation track diagrams are training sets, and 1800 differential constellation track diagrams are test sets;
4-2) performing multi-label processing on the data set: marking each picture in the data set obtained in the step 4-1) as two attribute labels, namely a wireless communication protocol label and a radiation source label;
4-3) performing label binarization processing: namely, binarizing the data set label into a [ 01 0 … 0] form and modifying the size of the picture into 64 multiplied by 64;
5) Multitasking neural network training: the hard sharing mode is adopted, which comprises the following steps:
5-1) constructing a multi-task convolutional neural network in a hard sharing mode, and training the multi-task convolutional neural network in a multi-label mode, namely, in a mode of one input and multiple outputs;
5-2) construction with reference to the VGG16 model: the hard sharing mode described in the step 5-1) has 3 sharing blocks, and the first block consists of 2 layers of 64×3×3 convolution layers; the second block consists of 2 layers of 128 x 3 convolutional layers; the third block consists of 3 layers of 256×3×3 convolutional layers and employs a maximum pooling layer of size 2×2;
5-3) in the step 5-2), a BN layer and a dropout layer are adopted after each conv layer, the dropout parameter is 0.4, and the whole multi-task neural network training adopts 'relu' as an activation function;
5-4) after the two tasks in the step 5-2) share network structures, parameters and characteristics, each task is identified by adopting an independent network layer, wherein the wireless communication protocol identification task adopts two full-connection layers, the parameters of the first full-connection layer are 1024, and the parameters of the second layer are 2, namely the corresponding identified types;
5-5) the radiation source identification task is different from the network information identification task in the step 5-4), and the parameters of the second layer full-connection layer of the radiation source identification task are 6, namely the radiation sources corresponding to 6 types;
5-6) saving the trained model weights;
5-6) the wireless communication protocol identification task and the radiation source identification task employ a softmax classifier and a cross entropy loss function, i.e., as shown in equation (5):
wherein q is a predictive tag, and p is a real tag;
5-7) the size of the parameter Batchsize is set to 20, the optimizer selects Adam optimizer, and the learning rate is set to 0.0001;
6) The weight coefficients of the wireless communication protocol identification task and the radiation source identification task are set to be 1.0, the total loss of the network is the sum of the losses of the radio frequency fingerprint identification task and the wireless communication protocol identification, and the total loss of the network corresponds to the formula (6):
wherein k is the number of tasks;
7) And storing the trained multi-task neural network and carrying out radio frequency fingerprint identification and wireless communication protocol identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111582991.5A CN114266312B (en) | 2021-12-22 | 2021-12-22 | Method for identifying radio frequency fingerprint and communication protocol based on multitask learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111582991.5A CN114266312B (en) | 2021-12-22 | 2021-12-22 | Method for identifying radio frequency fingerprint and communication protocol based on multitask learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114266312A CN114266312A (en) | 2022-04-01 |
| CN114266312B true CN114266312B (en) | 2024-03-22 |
Family
ID=80829529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111582991.5A Active CN114266312B (en) | 2021-12-22 | 2021-12-22 | Method for identifying radio frequency fingerprint and communication protocol based on multitask learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114266312B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111163460A (en) * | 2019-12-19 | 2020-05-15 | 北京交通大学 | Radio frequency fingerprint extraction method based on multiple interval difference constellation trajectory diagram |
| WO2020244134A1 (en) * | 2019-06-05 | 2020-12-10 | 华南理工大学 | Multi-task feature sharing neural network-based intelligent fault diagnosis method |
| CN112398553A (en) * | 2020-11-03 | 2021-02-23 | 上海电机学院 | Communication radiation source individual identification method based on differential equipotential sphere diagram |
-
2021
- 2021-12-22 CN CN202111582991.5A patent/CN114266312B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020244134A1 (en) * | 2019-06-05 | 2020-12-10 | 华南理工大学 | Multi-task feature sharing neural network-based intelligent fault diagnosis method |
| CN111163460A (en) * | 2019-12-19 | 2020-05-15 | 北京交通大学 | Radio frequency fingerprint extraction method based on multiple interval difference constellation trajectory diagram |
| CN112398553A (en) * | 2020-11-03 | 2021-02-23 | 上海电机学院 | Communication radiation source individual identification method based on differential equipotential sphere diagram |
Non-Patent Citations (1)
| Title |
|---|
| 彭超然 ; 刁伟鹤 ; 杜振宇 ; .基于深度卷积神经网络的数字调制方式识别.计算机测量与控制.2018,(第08期),第222-226页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114266312A (en) | 2022-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ya et al. | Large-scale real-world radio signal recognition with deep learning | |
| US10985955B2 (en) | Method for automatically identifying modulation mode for digital communication signal | |
| Riyaz et al. | Deep learning convolutional neural networks for radio identification | |
| CN105678273B (en) | The starting point detection algorithm of radio-frequency fingerprint identification technology transient signal | |
| CN111832417B (en) | Signal modulation pattern recognition method based on CNN-LSTM model and transfer learning | |
| CN110197127A (en) | Wireless signal detection and electromagnetic interference categorizing system and method based on deep learning | |
| CN108764077A (en) | A kind of digital signal modulated sorting technique based on convolutional neural networks | |
| Xiao et al. | Motion-Fi: Recognizing and counting repetitive motions with passive wireless backscattering | |
| Vagollari et al. | Joint detection and classification of RF signals using deep learning | |
| CN111585922A (en) | Modulation mode identification method based on convolutional neural network | |
| Gong et al. | A generative adversarial network based framework for specific emitter characterization and identification | |
| Lin et al. | GLR-SEI: green and low resource specific emitter identification based on complex networks and fisher pruning | |
| Li et al. | Differential contour stellar-based radio frequency fingerprint identification for Internet of Things | |
| CN114266312B (en) | Method for identifying radio frequency fingerprint and communication protocol based on multitask learning | |
| Yang et al. | Modulation recognition using wavelet transform based on AlexNet | |
| CN115866615B (en) | Wireless network communication relation discovery method based on electromagnetic spectrum characteristics | |
| Yin et al. | Short-time modulation classification of complex wireless communication signal based on deep neural network | |
| CN113242201A (en) | Wireless signal enhanced demodulation method and system based on generation classification network | |
| Zheng et al. | Deep learning for cooperative radio signal classification | |
| CN111917674A (en) | Modulation identification method based on deep learning | |
| Huang et al. | Automatic Recognition of Communication Signal Modulation Based on the Multiple‐Parallel Complex Convolutional Neural Network | |
| Kurmantayev et al. | RiSi: Spectro-temporal RAN-agnostic Modulation Identification for OFDMA Signals | |
| Zhang et al. | Open Set Domain Adaptation for Automatic Modulation Classification in Dynamic Communication Environments | |
| Wen et al. | RF transmitter identification and classification based on deep residual shrinkage network | |
| Guo et al. | Specific emitter identification for WiFi devices via Bezier curve fitting |
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 |