CN114598582A - Ultra-high order signal modulation format rapid identification method based on transfer learning - Google Patents
Ultra-high order signal modulation format rapid identification method based on transfer learning Download PDFInfo
- Publication number
- CN114598582A CN114598582A CN202210145229.9A CN202210145229A CN114598582A CN 114598582 A CN114598582 A CN 114598582A CN 202210145229 A CN202210145229 A CN 202210145229A CN 114598582 A CN114598582 A CN 114598582A
- Authority
- CN
- China
- Prior art keywords
- signal
- neural network
- convolutional neural
- network model
- domain
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000013526 transfer learning Methods 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 28
- 230000003044 adaptive effect Effects 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000007493 shaping process Methods 0.000 claims description 18
- 238000011084 recovery Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 239000006185 dispersion Substances 0.000 claims description 4
- 239000000835 fiber Substances 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 abstract description 12
- 238000004891 communication Methods 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 12
- 230000003287 optical effect Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 6
- 239000013307 optical fiber Substances 0.000 description 5
- 238000005457 optimization Methods 0.000 description 5
- 230000001427 coherent effect Effects 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000010287 polarization Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention discloses a method for quickly identifying a modulation format of an ultra-high order signal based on transfer learning, which comprises the following steps: taking a low-order signal with a label as a source domain and taking an ultra-high-order signal without the label as a target domain; constructing a convolutional neural network model; training the constructed convolutional neural network model by using a domain adaptive method to obtain a pre-trained convolutional neural network model; migrating the knowledge learned from the source domain and the pre-trained convolutional neural network model into a target domain, and finely adjusting the parameters of the pre-trained convolutional neural network model by using the target domain to obtain a convolutional neural network model with the recognition capability; and at a signal receiving end, preprocessing the received signal and then sending the preprocessed signal into a convolutional neural network model with identification capability to obtain a signal modulation format identification result. By introducing transfer learning, the invention can accurately recover the original signal on the basis of extremely limited training time and training sample number, thereby realizing reliable and efficient transmission of the communication system.
Description
Technical Field
The invention relates to a method for quickly identifying an ultra-high order signal modulation format based on transfer learning, and belongs to the technical field of optical transmission analysis methods and machine learning.
Background
With the continuous improvement of transmission capacity and transmission efficiency of optical fiber communication systems, optical fiber communication networks are developing towards more complicated and dynamic trends. To meet the increasing demand for information exchange, more and more communication signal modulation formats are available. The future optical fiber communication system supports a mixed form of multiple transmission rates and multiple modulation formats, and can adaptively adjust parameters such as signal rate, modulation format, wavelength and the like according to the dynamic channel condition, system resources and user service requirements so as to realize the maximum utilization of resources. Therefore, monitoring and identification of system parameters is one of the core technologies of future intelligent optical networks. The identification of the modulation format of the unknown received signal by the receiving end is a very light problem.
At present, a plurality of modulation format identification technologies based on a constellation diagram and a machine learning algorithm are proposed, and these methods usually utilize the constellation diagram of a signal received by a receiving end as a signal classification feature, and extract a high-dimensional abstract feature and give an identification result by means of some simple machine learning algorithms or neural network models. But modulation format identification for very high order modulation formats such as 16384QAM remains to be addressed. This is especially true for very large constellation because the modulation order is increased and the recovery at the receiving end is difficult. Although there is conventional identification based on constellation modulation format, the problem of identification of large size and ultra-high order is not solved.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method is characterized in that a domain self-adaptive theory in transfer learning is utilized, and a source domain of a signal modulation format recognition problem based on geometric forming and probability forming is transferred to a target domain of a high-order modulation format signal recognition problem based on an oversized constellation diagram. The method can avoid the problem of difficult identification caused by the fact that the constellation diagram of the ultrahigh-order modulation signal in the elastic optical network is too disordered and the signal characteristics are not obvious.
The invention adopts the following technical scheme for solving the technical problems:
a method for quickly identifying a modulation format of an ultra-high-order signal based on transfer learning comprises the following steps:
step 1, using a low-order signal with a label as a source domain and an ultra-high-order signal without the label as a target domain;
step 2, constructing a convolutional neural network model, wherein the model comprises a feature extractor, a label classifier and a domain classifier;
step 3, dividing the source domain into a training set, a verification set and a test set, and training the convolutional neural network model constructed in the step 2 by using the training set, the verification set and the test set and applying a field adaptive method to obtain a pre-trained convolutional neural network model;
step 4, migrating the knowledge learned from the source domain and the pre-trained convolutional neural network model into a target domain, and finely adjusting the parameters of the pre-trained convolutional neural network model by using the target domain to obtain the convolutional neural network model with the recognition capability;
and 5, at a signal receiving end, carrying out DSP (digital signal processor) pretreatment on the received signal and then sending the signal into a convolutional neural network model with identification capability to obtain a signal modulation format identification result.
As a preferred scheme of the present invention, in step 1, the low-order signal with a tag includes a PS-16QAM probability shaping constellation and a GS-32QAM geometric shaping constellation; the ultra high order signal without the tag includes a probability shaped 1024QAM constellation.
As a preferred embodiment of the present invention, in step 3, the training of the convolutional neural network model constructed in step 2 by using a domain adaptive method includes: (i) optimizing parameters of the label classifier and the domain classifier to minimize errors thereof on the source domain training set; (ii) optimizing parameters of the underlying deep feature map to minimize loss of the tag classifier; (iii) parameters of the underlying deep feature map are optimized to maximize the loss of the domain classifier.
As a preferable aspect of the present invention, the signal features extracted by the feature extractor include: amplitude, phase of the signal, and combined distribution of the IQ two-way signal.
As a preferred aspect of the present invention, the DSP preprocessing includes: analog-to-digital conversion, dispersion compensation of the fiber channel, clock recovery of the signal, and blind equalization.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects:
the invention adopts a transfer learning method to carry out data recovery on the signals of the receiving end. A common constellation forming optimization method is loaded as a pre-training model, a domain adaptive theory is adopted to analyze the characteristic relation of data between a source domain and a target domain, and a data recovery model of a receiving end is quickly constructed, so that diversified constellation forming optimization methods including high-order probability forming, geometric forming and the like of the transmitting end are responded, and high-precision and high-accuracy identification of signals of the receiving end is realized.
Drawings
FIG. 1 is a schematic diagram of a constellation shaping identification method based on transfer learning;
FIG. 2 is a diagram of a receiving end network model construction based on transfer learning;
fig. 3 is a constellation distribution form, in which, (a) is a probability-shaped 16QAM constellation, (b) is a geometry-shaped 16QAM constellation, and (c) is a probability-shaped 1024QAM constellation;
FIG. 4 is a schematic diagram of a domain adaptive migration;
FIG. 5 is a system model diagram;
fig. 6 is a simulation result diagram, in which (a) is an unidentified constellation point distribution diagram, and (b) is a constellation point distribution diagram after the convolutional network identification.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
The flow of the constellation forming rapid identification method based on transfer learning is shown in figure 1.
According to the method, common constellation shaping such as PS-16QAM probability shaping and GS-32QAM geometric shaping is extracted as source domain characteristics according to a field self-adaptive theory in transfer learning, and a pre-training model is loaded to realize rapid identification of the ultrahigh-order modulation format signal at a receiving end. Aiming at the problems that the optimization algorithm of a transmitting end is unknown, training data is insufficient and original signal data is difficult to accurately recover at an optical network receiving end, the invention can accurately recover the original signal by introducing transfer learning on the basis of extremely limited training time and training sample number, thereby realizing reliable and efficient transmission of a communication system.
The method firstly proposes to adopt a field self-adaptive theory, utilize an obtained common constellation forming optimization model as a source domain model, utilize a small amount of high-order constellation forming training samples and source domain data, establish a reliable model to carry out data analysis on the high-order optimization target field, extract the characteristic distribution such as amplitude phase and the like in the constellation, utilize the characteristics to carry out comprehensive analysis at a receiving end, and finally accurately identify an original sending signal.
The invention relates to a constellation shaping rapid identification method based on transfer learning. The first part is a training module of a network model, and the network model is trained by applying a field adaptive method to realize the identification of constellation forming signals. Firstly, checking whether a pre-training model exists or not, if not, downloading the pre-training model, otherwise, directly reading parameters of the pre-training model, and dividing a labeled data set into a training set, a verification set and a test set which are used as the input of the pre-training model; and after loading the pre-training model, performing iterative training on the network model by adopting different data samples. The second part is the application part of the network: at a receiving end of a communication system, signals are subjected to preprocessing such as dispersion compensation and carrier phase recovery, then signal classification features of the signals output by a receiver are extracted through a trained neural network, and finally a final optical signal ultrahigh-order debugging format recognition result is output by a classifier.
1. Modulation format identification based on domain adaptive theory
The transfer learning is a new strategy of the small sample in supervised learning, wherein in the aspect of processing domain allocation problems, the domain adaptation can realize the target domain calibration by the source domain under the condition that the labeled source domain and the unlabeled target domain share the same characteristics and categories, but the characteristic distribution is different. Through model parameter sharing, the pre-trained network model in the large-scale common constellation forming signal data set is embedded into other higher-order task models to serve as feature extractors of other tasks. As shown in fig. 2, the model is trained from the source domain, the knowledge learned from the source domain and the convolutional neural network are migrated to the target domain, and then a classification layer is redesigned on the target domain and the migrated network model is reconstructed into a new convolutional neural network for training the target domain signal data. The key point of the transfer learning is that the source task field and the target task field need to have correlation, as shown in fig. 2, the target task of the invention is to identify a high-order constellation forming signal, so that a model can be pre-trained by selecting a commonly-used probability forming and geometric forming constellation as an original data domain, and finally, a better identification effect can be achieved by fine-tuning the model.
1.1 Signal feature analysis
Probability forming increases the probability of the signal with low energy by reducing the probability of the signal with high energy, reduces the average energy value of the signal, improves the gain index of the system and reduces the error rate. However, signals generated by a source are equal in probability, so that the generation of multi-probability information through probability shaping needs to add redundant bits directly or indirectly through generating a polynomial, however, most shaping causes the number of compressed signal points to be not power exponent times of 2, a higher requirement is provided for the coding of a label or an index, the addition amount of redundant information is increased, and the coding rate is reduced. In order to achieve higher transmission rates, high-order QAM modulated signals are typically used in long-distance coherent optical transmission links.
Fig. 3 (a), (b) and (c) respectively list probability-shaped 16QAM constellations, geometry-shaped constellations and distribution forms of probability-shaped 1024QAM constellations.
In the mapping process, the amplitude and the phase of the signal and the combined distribution of IQ two-path signals form the main characteristics of the signal, a convolutional neural network is adopted to model Euclidean data, the overall characteristics of the signal after mapping, the characteristic information of each signal and the correlation information between signal domains are learned, and good signal characteristics are provided for subsequent complex format analysis.
1.2 Domain adaptive migration
In learning features, mainly resolution is combined with domain invariance by jointly optimizing potential features and two discriminative classifiers that operate with the features. The two classifiers are a label classifier to predict class labels in the training process and the testing process, and a domain classifier to distinguish the source domain from the target domain during training, respectively. The three training steps of the method are respectively as follows: (i) optimizing the parameters of the two classifiers to minimize the error of the two classifiers on the source domain training set; (ii) (ii) optimizing parameters of the underlying deep feature map to minimize loss of tag classifiers (iii) optimizing parameters of the underlying deep feature map to maximize loss of domain classifiers. All three training steps described above can be embedded in a suitable deep feed-forward network (as shown in fig. 4) and can be trained using standard back-propagation algorithms based on random gradient descent or modifications thereof.
Gf in fig. 4: the data is mapped to the feature space such that Gy distinguishes between the labels of the source domain data and Gd distinguishes between whether the different data is from the source domain or the target domain.
Gy: and classifying the source domain data of the feature space to separate out correct labels as much as possible.
Gd: and performing domain classification on the data in the feature space to distinguish which domain the data comes from as much as possible. Through network iterative training, the result of the Gf and Gd game is expected to be that the data of the source domain and the data of the target domain are distributed relatively uniformly on the feature space.
Wherein the data of the source domain is tagged and the data of the target domain is untagged. Gf maps the data of both the source domain and the target domain onto a feature space Z, Gy predicts the label y, Gd predicts whether the data is from the target domain or the source domain. So that flowing into Gy is tagged source data and flowing into Gd is data for the source and target domains without tags.
The method provided by the invention transfers the recognition problem of probability shaping and geometric shaping to the recognition of the ultra-high order signal modulation format by virtue of the high complexity of the probability shaping and the geometric shaping. The convolutional neural network model based on the domain self-adaptation provided by the invention analyzes the characteristic relation of data between a source domain and a target domain by using a domain self-adaptation theory, quickly constructs a receiving end data recovery model, can learn new signal vector characteristics from signal association, and has extremely high model utilization rate.
2. System model
The constellation shaping fast recognition transmission system model based on the transfer learning provided by the invention is shown in fig. 5. At the transmitting end, parameters such as constellation shaping modulation formats and transmission rates of transmission signals can be dynamically changed according to the requirements of users and the conditions of channels, system resources are reasonably configured, and efficiency maximization is achieved. When the transmitting end transmits data, information to be transmitted is loaded on an I path and a Q path of two paths of polarized light by an IQ modulator, and the multiplexed light in different modes is coupled into a few-mode optical fiber by a mode multiplexer for data transmission. At the receiving end of the signal, the mode demultiplexer will be coupled into a single mode fiber and sent to the coherent receiver for demodulation. The coherent receiver completes polarization demultiplexing, converts an analog signal into a digital signal through the digital-to-analog converter ADC, and then enters the DSP for subsequent processing of the signal.
Before the modulation format of the optical signal is identified, a series of simple processing which is independent of the modulation format, such as dispersion compensation of an optical fiber channel, clock recovery of the signal, and CMA equalization, etc., can be performed. And then entering a complex format analysis stage, firstly performing power normalization pretreatment on the signal, then inputting a pre-trained neural network, finely training a new convolutional neural network through a small number of data sets, extracting signal characteristics by using the network model, identifying a constellation forming mode, and then selecting different algorithms to perform operations such as MMA equalization, frequency offset compensation, carrier phase recovery, signal demodulation and the like on the signal. The constellation diagram obtained by simulation is shown as (a), (b) in fig. 6.
The invention provides a method for reconstructing a high-order constellation shaping recovery model by using a transfer learning method, applies field self-adaptive principle analysis, extracts the characteristic distribution of data points in a constellation, performs data recovery at a receiving end and accurately identifies effective information of signal data of the transmitting end. The method can effectively improve the error code performance of the system by using limited training samples.
The above embodiments are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modifications made on the basis of the technical scheme according to the technical idea of the present invention fall within the protection scope of the present invention.
Claims (5)
1. A method for quickly identifying a modulation format of an ultra-high order signal based on transfer learning is characterized by comprising the following steps:
step 1, using a low-order signal with a label as a source domain and an ultra-high-order signal without the label as a target domain;
step 2, constructing a convolutional neural network model, wherein the model comprises a feature extractor, a label classifier and a domain classifier;
step 3, dividing the source domain into a training set, a verification set and a test set, and training the convolutional neural network model constructed in the step 2 by using the training set, the verification set and the test set and applying a field adaptive method to obtain a pre-trained convolutional neural network model;
step 4, transferring the knowledge learned from the source domain and the pre-trained convolutional neural network model into a target domain, and finely adjusting the parameters of the pre-trained convolutional neural network model by using the target domain to obtain the convolutional neural network model with the recognition capability;
and 5, at a signal receiving end, carrying out DSP pretreatment on the received signal and then sending the signal into a convolutional neural network model with identification capability to obtain a signal modulation format identification result.
2. The method for rapidly identifying the modulation format of the ultra-high-order signal based on the transfer learning of claim 1, wherein in the step 1, the low-order signal with the label comprises a PS-16QAM probability shaping constellation map and a GS-32QAM geometric shaping constellation map; the unlabeled, ultra-high order signal includes a probability-shaped 1024QAM constellation.
3. The method for rapidly identifying the ultra-high order signal modulation format based on the transfer learning of claim 1, wherein in the step 3, the convolutional neural network model constructed in the step 2 is trained by applying a domain adaptive method, and the method comprises the following steps: (i) optimizing parameters of the label classifier and the domain classifier to minimize errors thereof on the source domain training set; (ii) optimizing parameters of the underlying deep feature map to minimize loss of the tag classifier; (iii) parameters of the underlying deep feature map are optimized to maximize the loss of the domain classifier.
4. The method for rapidly identifying an ultra-high-order signal modulation format based on transfer learning of claim 1, wherein the signal features extracted by the feature extractor comprise: amplitude, phase of the signal, and combined distribution of the IQ two-way signal.
5. The method for rapidly identifying an ultra-high order signal modulation format based on transfer learning of claim 1, wherein the DSP preprocessing comprises: analog-to-digital conversion, dispersion compensation of the fiber channel, clock recovery of the signal, and blind equalization.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210145229.9A CN114598582A (en) | 2022-02-17 | 2022-02-17 | Ultra-high order signal modulation format rapid identification method based on transfer learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210145229.9A CN114598582A (en) | 2022-02-17 | 2022-02-17 | Ultra-high order signal modulation format rapid identification method based on transfer learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114598582A true CN114598582A (en) | 2022-06-07 |
Family
ID=81804514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210145229.9A Pending CN114598582A (en) | 2022-02-17 | 2022-02-17 | Ultra-high order signal modulation format rapid identification method based on transfer learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114598582A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116016069A (en) * | 2022-12-02 | 2023-04-25 | 中国人民解放军国防科技大学 | Modulation recognition system based on transfer learning |
| CN116776230A (en) * | 2023-08-22 | 2023-09-19 | 北京海格神舟通信科技有限公司 | Method and system for identifying signal based on feature imprinting and feature migration |
| CN117932413A (en) * | 2024-03-25 | 2024-04-26 | 南京信息工程大学 | High generalization mechanical intelligent fault diagnosis method based on style migration under variable working conditions |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111092836A (en) * | 2019-12-13 | 2020-05-01 | 中国人民解放军空军工程大学 | Signal modulation mode identification method and device |
| CN113626960A (en) * | 2021-07-29 | 2021-11-09 | 南京信息工程大学 | Anti-mode coupling few-mode signal complex format analysis method and device |
| CN113723556A (en) * | 2021-09-08 | 2021-11-30 | 中国人民解放军国防科技大学 | Modulation mode identification method based on entropy weighting-multi-mode domain antagonistic neural network |
| CN114024810A (en) * | 2021-11-03 | 2022-02-08 | 南京信息工程大学 | Multi-core optical fiber channel modulation format identification method and device |
-
2022
- 2022-02-17 CN CN202210145229.9A patent/CN114598582A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111092836A (en) * | 2019-12-13 | 2020-05-01 | 中国人民解放军空军工程大学 | Signal modulation mode identification method and device |
| CN113626960A (en) * | 2021-07-29 | 2021-11-09 | 南京信息工程大学 | Anti-mode coupling few-mode signal complex format analysis method and device |
| CN113723556A (en) * | 2021-09-08 | 2021-11-30 | 中国人民解放军国防科技大学 | Modulation mode identification method based on entropy weighting-multi-mode domain antagonistic neural network |
| CN114024810A (en) * | 2021-11-03 | 2022-02-08 | 南京信息工程大学 | Multi-core optical fiber channel modulation format identification method and device |
Non-Patent Citations (2)
| Title |
|---|
| TONG LI等: "Domain Adaptation-Based Automatic Modulation Recognition" * |
| 杜攀非: "基于深度迁移学习的波形识别研究" * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116016069A (en) * | 2022-12-02 | 2023-04-25 | 中国人民解放军国防科技大学 | Modulation recognition system based on transfer learning |
| CN116776230A (en) * | 2023-08-22 | 2023-09-19 | 北京海格神舟通信科技有限公司 | Method and system for identifying signal based on feature imprinting and feature migration |
| CN116776230B (en) * | 2023-08-22 | 2023-11-14 | 北京海格神舟通信科技有限公司 | Method and system for identifying signal based on feature imprinting and feature migration |
| CN117932413A (en) * | 2024-03-25 | 2024-04-26 | 南京信息工程大学 | High generalization mechanical intelligent fault diagnosis method based on style migration under variable working conditions |
| CN117932413B (en) * | 2024-03-25 | 2024-06-11 | 南京信息工程大学 | High generalization mechanical intelligent fault diagnosis method based on style migration under variable working conditions |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114598582A (en) | Ultra-high order signal modulation format rapid identification method based on transfer learning | |
| US11509339B2 (en) | Method of constructing and training a neural network | |
| US10531415B2 (en) | Learning communication systems using channel approximation | |
| Zhang et al. | Non-data-aided k-nearest neighbors technique for optical fiber nonlinearity mitigation | |
| CN113626960B (en) | Anti-mode coupling few-mode signal complex format analysis method and device | |
| CN114881092A (en) | Signal modulation identification method based on feature fusion | |
| Zhang et al. | Joint modulation format identification and OSNR monitoring using cascaded neural network with transfer learning | |
| CN109740690A (en) | The KNN equalization algorithm based on Feature Engineering for short distance optic communication | |
| CN112598072A (en) | Equalization method of improved Volterra filter based on weight coefficient migration of SVM training | |
| Ding et al. | Data-and-knowledge dual-driven automatic modulation recognition for wireless communication networks | |
| CN114513394B (en) | Signal modulation format identification method, system and device based on attention mechanism diagram neural network and storage medium | |
| CN109818889A (en) | Equalization algorithm for the SVM classifier optimization in high-order PAM optical transmission system | |
| CN112242969A (en) | Novel single-bit OFDM receiver based on model-driven deep learning | |
| Le-Tran et al. | Deep learning-based collaborative constellation design for visible light communication | |
| Sheng et al. | A multi-task semantic communication system for natural language processing | |
| CN113114603A (en) | Information recovery method and device for MIMO-OFDM system | |
| CN114759997B (en) | MIMO system signal detection method based on data model double driving | |
| Gao et al. | Joint baud-rate and modulation format identification based on asynchronous delay-tap plots analyzer by using convolutional neural network | |
| Zhang et al. | Heterogeneous deep model fusion for automatic modulation classification | |
| CN111130697B (en) | Method for reducing complexity of communication physical layer transmission system based on automatic encoder | |
| CN113839744A (en) | Blind detection method of generalized wireless optical MIMO system based on deep learning | |
| CN115426048B (en) | Augmentation space signal detection method, receiving device and optical communication system | |
| Prakash et al. | Improved BER performance using novel ELM based demodulator for CAP-VLC system | |
| CN114553650B (en) | Multi-level neural network-based anti-mode coupling signal complex format analysis method | |
| Setzler et al. | Deep Learning for Spectral Filling in Radio Frequency Applications |
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: 20220607 |