CN114189416A - Digital modulation signal identification method based on consistency regularization - Google Patents
Digital modulation signal identification method based on consistency regularization Download PDFInfo
- Publication number
- CN114189416A CN114189416A CN202111479194.4A CN202111479194A CN114189416A CN 114189416 A CN114189416 A CN 114189416A CN 202111479194 A CN202111479194 A CN 202111479194A CN 114189416 A CN114189416 A CN 114189416A
- Authority
- CN
- China
- Prior art keywords
- data
- modulation signal
- point
- digital modulation
- samples
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention belongs to the technical field of deep learning, and particularly relates to a digital modulation signal identification method based on consistency regularization. The traditional digital modulation signal identification method based on deep learning needs a large amount of marked data samples for training. In actual communication activities, only a large number of unlabeled signal samples are readily available, and obtaining the manual labels is costly and inefficient. The invention expands the data set by symmetric transformation, and enhances the samples by combining random rearrangement. By using the consistency regularization algorithm, a large number of unmarked samples are fully utilized to assist learning under a small number of marked samples, so that the identification accuracy is effectively improved, and the feasibility and the practicability of deep learning in a digital modulation signal identification task are enhanced.
Description
Technical Field
The invention belongs to the technical field of deep learning, and particularly relates to a digital modulation signal identification method based on consistency regularization.
Background
Digital signal modulation identification is an important technology in the field of wireless communication, and has wide application in the civil and military fields. The task of digital signal modulation identification is mainly to confirm the type of signal modulation under the condition of only a small amount of a priori information or no a priori information, and is an important step between signal detection and demodulation. With the rapid development of wireless communication, the modulation modes of signals become various, and the cost of modulation identification also increases.
Modulation identification methods can generally be classified into two categories: likelihood ratio based modulation identification methods and feature based modulation identification methods. The modulation recognition method based on the likelihood ratio can obtain the optimal result under the meaning of the Bayesian criterion, but needs a large amount of prior information, and the algorithm has high complexity, is very sensitive to parameter setting and has poor robustness. The modulation recognition method based on the characteristics mainly utilizes the known modulation signal samples to train a network model, automatically learns and extracts the signal characteristics, and classifies unknown signals according to the extracted characteristics. Common network models are SVM, decision tree, LSTM, CNN, etc.
The existing modulation recognition method based on the characteristics usually needs a large amount of known marked samples to train a network model, and then modulation recognition tasks are carried out after an optimal network model is obtained. In practical situations, only a large amount of unlabeled samples are often available, and obtaining labeled samples is difficult and costly. Typical feature-based modulation identification methods do not take full advantage of the large number of unlabeled samples.
The Chinese patent with the application number of CN202110232691.8 and the application publication number of CN113014524A, the application name of electronics and technology university and the invention name of 'a digital signal modulation identification method based on deep learning' discloses a digital signal modulation identification method based on deep learning. The invention obtains better identification precision by constructing the RSN-MI neural network and optimizing the network by using a large amount of data. However, the case where the amount of usable marked data is small is not considered, and the actual application scene is not matched.
Therefore, how to fully utilize a large amount of unmarked samples to realize digital signal modulation identification under a small amount of marked samples becomes an urgent problem to be solved in the field of radio modulation identification.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a digital signal modulation identification method based on consistency regularization. The core of the invention is mainly divided into four parts: firstly, processing known signal data to realize data expansion; secondly, constructing a disturbance version of the unmarked sample through data random rearrangement, and calculating consistency regularization loss; thirdly, updating the training network by utilizing the cross entropy loss of the marked samples and the consistency regularization loss inverse gradient of the unmarked samples; and fourthly, completing type confirmation of the unknown modulation signal based on the trained network.
The technical scheme of the invention is as follows:
a method for identifying digital signal modulation based on consistency regularization, as shown in fig. 1, the method comprising the steps of:
and S1, acquiring the digital modulation signal data and the label, and marking as a data set D. Preprocessing the data set D, dividing the data set, including the marked data set
And label-free data sets
With D ═ S ═ U and M < N, x is a labeled dataset, y is the label to which x corresponds, x isuThe data is unmarked, M is the number of marked data sets, and N is the number of marked data sets.
S2, performing symmetric expansion and random rearrangement on the data, where the expansion manner of symmetric transformation is shown in fig. 2, and the three expansion manners are:
wherein the content of the first and second substances,
is a point of symmetry of the data point x about the axis of the constellation diagram Re;
is a point of symmetry of the data point x with respect to the origin of the constellation diagram;
the symmetric points of the data point x about the axis of the constellation diagram Im are respectively rotated by 90 degrees, 180 degrees and 270 degrees counterclockwise corresponding to the sample point x. Comprises the following steps:
and | D*4 | · | D |. Processing the data set D by adopting random rearrangement to obtain:
there is | D' | ═ 4 · | D |. Wherein permute (·) is a random rearrangement algorithm, and x' in the above formula is data after x random rearrangement, and the steps are: randomly selecting a certain point p of a signal sample as a truncation point, truncating the signal sample by the truncation point, exchanging the positions of the front and rear sections of the signal sample, and splicing to obtain a randomly rearranged sample:
s3, constructing a recognition network model, and calculating the cross entropy loss of the marked sample S', wherein the form is as follows:
where CE (-) is a cross-entropy loss function, fθ(. cndot.) is a network model for identification.
Computing pairs of unlabeled samples
In the form of a loss of consistency of:
wherein JS (-) is the Jensen-Shannon divergence.
Calculating a joint loss function:
where λ (t) is an increasing coefficient of the form:
λmaxis the maximum value of λ (t); exp (·) is an exponential function based on a natural constant e; t is the training time at the current time; t is the increasing maximum training moment.
And S4, acquiring the digital modulation signal, preprocessing the digital modulation signal by adopting the method of the step S2, and inputting the digital modulation signal into the trained recognition network to finish recognition of the modulation signal.
The invention has the beneficial effects that the invention provides a digital signal modulation identification method based on consistency regularization, which utilizes symmetric expansion and random rearrangement to realize the expansion and enhancement of samples; by combining with a consistency regularization algorithm, the network model can approach the performance under the training of a large number of samples under the condition of a very small number of samples, and the robustness of the network model is stronger. The method for expanding and enhancing the digital modulation signal and the consistency regularization algorithm are applied to the digital signal modulation recognition, so that the defects of the existing method are overcome, and the signal modulation recognition algorithm is more practical and reliable.
Drawings
FIG. 1 is a flow chart of a method for consistency-based digitally modulated signal identification in accordance with the present invention;
FIG. 2 is a diagram of a symmetric transformation expansion method according to the present invention.
Detailed Description
The invention has detailed description on the types of the technical scheme, and the method extracts and utilizes a large amount of information without marked data on the basis of fully utilizing a small amount of information with marked data, so that a network model can obtain excellent performance under the condition of a small amount of marked samples, and the practicability of the digital signal modulation identification method is improved.
In the digital signal modulation identification method based on consistency regularization, signal data set is expanded and enhanced through symmetric transformation expansion and random rearrangement of signals. The expansion and enhancement of the unmarked data set can be regarded as disturbance of the original sample, the consistency between the disturbed sample and the original unmarked sample is measured through the JS divergence, and the unmarked sample is fully utilized. In the training process, the optimization direction of the network model is constrained through a specific loss function, and the emphasis bias of the training is controlled by using coefficient weighting. And determining the modulation type of the unknown signal by using the trained network model.
Claims (1)
1. A digital signal modulation identification method based on consistency regularization is characterized by comprising the following steps:
s1, acquiring digital modulation signal data and labels, and recording as a data set D; preprocessing the data set D, dividing the data set, including the number of labelsData set
And label-free data sets
With D ═ S ═ U and M < N, x is a labeled dataset, y is the label to which x corresponds, x isuThe data are unmarked data, M is the number of marked data sets, and N is the number of marked data sets;
s2, carrying out symmetrical expansion and random rearrangement on the data, wherein the expansion mode of the symmetrical transformation is as follows:
wherein the content of the first and second substances,
is a point of symmetry of the data point x about the axis of the constellation diagram Re;
is a point of symmetry of the data point x with respect to the origin of the constellation diagram;
the symmetric points of the data point x about the axis of the constellation diagram Im are respectively rotated by 90 degrees, 180 degrees and 270 degrees counterclockwise corresponding to the sample point x, and the following steps are provided:
|D*4 | · | D |; processing the data set D by adopting random rearrangement to obtain:
d' | 4 · | D |; where permute (·) is a random shuffling algorithm: randomly selecting a certain point p of a signal sample as a truncation point, truncating the signal sample by the truncation point, exchanging the positions of the front and rear sections of the signal sample, and splicing to obtain a randomly rearranged sample:
s3, constructing an identification network model, and calculating a marked sample
In the form of a cross-entropy loss of:
where CE (-) is a cross-entropy loss function, fθ() is a network model for recognition;
computing pairs of unlabeled samples
In the form of a loss of consistency of:
wherein JS (-) is Jensen-Shannon divergence;
calculating a joint loss function:
where λ (t) is an increasing coefficient of the form:
λmaxis the maximum value of λ (t); exp (·) is an exponential function based on a natural constant e; t is the training time at the current time; t is an increasing maximum training time instant;
and S4, acquiring the digital modulation signal, preprocessing the digital modulation signal by adopting the method of the step S2, and inputting the digital modulation signal into the trained recognition network to finish recognition of the modulation signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111479194.4A CN114189416B (en) | 2021-12-02 | 2021-12-02 | Digital modulation signal identification method based on consistency regularization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111479194.4A CN114189416B (en) | 2021-12-02 | 2021-12-02 | Digital modulation signal identification method based on consistency regularization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114189416A true CN114189416A (en) | 2022-03-15 |
| CN114189416B CN114189416B (en) | 2023-01-10 |
Family
ID=80603473
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111479194.4A Active CN114189416B (en) | 2021-12-02 | 2021-12-02 | Digital modulation signal identification method based on consistency regularization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114189416B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018176889A1 (en) * | 2017-03-27 | 2018-10-04 | 华南理工大学 | Method for automatically identifying modulation mode for digital communication signal |
| CN111695417A (en) * | 2020-04-30 | 2020-09-22 | 中国人民解放军空军工程大学 | Signal modulation pattern recognition method |
| CN112004157A (en) * | 2020-08-11 | 2020-11-27 | 海信电子科技(武汉)有限公司 | Multi-round voice interaction method and display equipment |
| CN113378736A (en) * | 2021-06-18 | 2021-09-10 | 武汉大学 | Remote sensing image depth network semi-supervised semantic segmentation method based on transformation consistency regularization |
| CN113378673A (en) * | 2021-05-31 | 2021-09-10 | 中国科学技术大学 | Semi-supervised electroencephalogram signal classification method based on consistency regularization |
-
2021
- 2021-12-02 CN CN202111479194.4A patent/CN114189416B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018176889A1 (en) * | 2017-03-27 | 2018-10-04 | 华南理工大学 | Method for automatically identifying modulation mode for digital communication signal |
| CN111695417A (en) * | 2020-04-30 | 2020-09-22 | 中国人民解放军空军工程大学 | Signal modulation pattern recognition method |
| CN112004157A (en) * | 2020-08-11 | 2020-11-27 | 海信电子科技(武汉)有限公司 | Multi-round voice interaction method and display equipment |
| CN113378673A (en) * | 2021-05-31 | 2021-09-10 | 中国科学技术大学 | Semi-supervised electroencephalogram signal classification method based on consistency regularization |
| CN113378736A (en) * | 2021-06-18 | 2021-09-10 | 武汉大学 | Remote sensing image depth network semi-supervised semantic segmentation method based on transformation consistency regularization |
Non-Patent Citations (1)
| Title |
|---|
| YU, K., MA, H., LIN, T. R., & LI, X: "A consistency regularization based semi-supervised learning approach for intelligent fault diagnosis of rolling bearing", 《MEASUREMENT》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114189416B (en) | 2023-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109949317B (en) | Semi-supervised image example segmentation method based on gradual confrontation learning | |
| CN109492099B (en) | Cross-domain text emotion classification method based on domain impedance self-adaption | |
| CN109685135B (en) | Few-sample image classification method based on improved metric learning | |
| CN111079847B (en) | Remote sensing image automatic labeling method based on deep learning | |
| CN112613556B (en) | Low-sample image emotion classification method based on meta-learning | |
| CN110598530A (en) | Small sample radio signal enhanced identification method based on ACGAN | |
| CN111126563B (en) | Target identification method and system based on space-time data of twin network | |
| CN116403058B (en) | Remote sensing cross-scene multispectral laser radar point cloud classification method | |
| CN113705787B (en) | Digital modulation signal identification method based on deep collaborative training | |
| CN108805102A (en) | A kind of video caption detection and recognition methods and system based on deep learning | |
| CN114726692B (en) | SERESESESENet-LSTM-based radiation source modulation mode identification method | |
| CN113064995A (en) | Text multi-label classification method and system based on deep learning of images | |
| CN113920472A (en) | Unsupervised target re-identification method and system based on attention mechanism | |
| CN114360038A (en) | Weak supervision RPA element identification method and system based on deep learning | |
| CN111291705A (en) | Cross-multi-target-domain pedestrian re-identification method | |
| CN110705384B (en) | Vehicle re-identification method based on cross-domain migration enhanced representation | |
| CN110555125A (en) | Vehicle retrieval method based on local features | |
| Zheng et al. | Learning from the web: Webly supervised meta-learning for masked face recognition | |
| CN114189416B (en) | Digital modulation signal identification method based on consistency regularization | |
| Li et al. | Remote Sensing Image Classification with Few Labeled Data Using Semisupervised Learning | |
| CN115661539A (en) | Less-sample image identification method embedded with uncertainty information | |
| CN115329821A (en) | Ship noise identification method based on pairing coding network and comparison learning | |
| CN114842301A (en) | Semi-supervised training method of image annotation model | |
| Zhang et al. | Vehicle verification based on deep siamese network with similarity metric | |
| Lin et al. | Features fusion based automatic modulation classification using convolutional neural network |
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 |