CN114785649B - Satellite communication signal identification method based on multiport neural network - Google Patents
Satellite communication signal identification method based on multiport neural network Download PDFInfo
- Publication number
- CN114785649B CN114785649B CN202210425192.5A CN202210425192A CN114785649B CN 114785649 B CN114785649 B CN 114785649B CN 202210425192 A CN202210425192 A CN 202210425192A CN 114785649 B CN114785649 B CN 114785649B
- Authority
- CN
- China
- Prior art keywords
- signal
- neural network
- network
- diagram
- multiport
- 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 41
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 27
- 238000004891 communication Methods 0.000 title claims abstract description 17
- 238000010586 diagram Methods 0.000 claims abstract description 49
- 238000004458 analytical method Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 13
- 238000005070 sampling Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 5
- 238000001228 spectrum Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000000903 blocking effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims 1
- 238000013527 convolutional neural network Methods 0.000 abstract description 9
- 238000013135 deep learning Methods 0.000 abstract description 6
- 238000005516 engineering process Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 12
- 238000010606 normalization Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004220 aggregation Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000007493 shaping process Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000002401 inhibitory effect Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000452 restraining effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
Abstract
The invention provides a satellite communication signal identification method based on a multiport neural network, which comprises the following steps: constructing a signal sample library, and converting signals into sample expression forms which are more suitable for a neural network topological structure; adopting a signal eye diagram, a signal vector diagram and a signal time-frequency analysis diagram as sample expression forms; and introducing a deep learning technology, constructing a multi-terminal convolutional neural network, extracting, optimizing and learning signal characteristics, and finally establishing a stable signal identification network model to realize efficient and accurate identification of signals under low signal-to-noise ratio. The method can be applied to the field of communication signal identification.
Description
Technical Field
The invention relates to the field of satellite communication, in particular to a satellite communication signal identification method based on a multiport neural network.
Background
The automatic modulation identification of the signal refers to the known modulation set where the signal is located, and the modulation type of the target signal is correctly identified by using a correlation technique. It is an important research direction in the fields of communication reconnaissance and signal blind processing, and there is a considerable demand in many fields. Along with the development of communication technology, various new modulation modes are continuously appeared, and modulation recognition technology is also required to be continuously developed to adapt to recognition requirements under different conditions.
In the automatic modulation identification method, the method characterized by the domain information such as the phase, frequency and amplitude of the signal is widely applied, but is greatly affected by noise, and the performance is seriously reduced under the condition of low signal-to-noise ratio. The method based on the high-order statistics, such as the high-order accumulation or the cyclic spectrum of the signal, has good anti-noise performance, but the selection of the characteristics lacks theoretical guidance, and the decision threshold is difficult to set in the process of processing complex multi-class modulation signal identification.
Compared with the algorithm of the traditional manual design characteristics, the deep learning technology has outstanding achievement in the fields of voice and image due to the self-learning capability and potential fault tolerance to samples. In recent years, researchers engaged in the field of communications have also gradually utilized deep learning techniques to solve signal processing-related problems. In the field of modulation recognition, the main idea is to establish shallow feature expression of signals and to construct a deep learning neural network to learn samples. The convolutional neural network is an advanced deep learning neural network, and a new network structure comprising a convolutional layer, a pooling layer and the like is introduced while a loss function, a back propagation algorithm, a parameter updating strategy and a super parameter which are the same as those of a common neural network are used, so that the convolutional neural network is commonly used for classifying and identifying two-dimensional or three-dimensional input signals.
Time-frequency analysis is an important class of methods for processing non-stationary signals, which represent the non-stationary signals as a two-dimensional function of time and frequency, which can be more intuitively analyzed and processed. A conventional common method of analyzing and processing stationary signals is fourier transform (Fourier Transform, FT), which establishes a one-to-one mapping of the time and frequency domains of the signal. However, FT has the disadvantage that only the frequency content of the signal can be analyzed as a whole, and frequency variations in the local part of the signal cannot be obtained. The time-frequency analysis can overcome the defect of FT, and the two-dimensional function of time and frequency is used for representing the signal, so that the relation between the time domain and the frequency domain of the signal is intuitively reflected. The time-frequency analysis method can be divided into linear time-frequency and quadratic time-frequency, and common linear time-frequency is expressed as follows: short-time fourier transforms (Short Time Fourier Transform, STFT) and wavelet transforms. The quadratic time frequency is also called time frequency distribution, which can describe the energy distribution of signals, typically Wigner-Ville distribution (WVD) and Cohen time frequency distribution, and has the biggest characteristic of effectively restraining cross terms.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a satellite communication signal identification method based on a multiport neural network, which can realize that the identification task can be efficiently and accurately completed under the condition of low signal-to-noise ratio. The invention aims to improve the noise resistance of signal identification by introducing a deep learning technical means, overcome the defects of poor robustness and the like of manually extracted high-dimensional statistical features, and realize accurate and efficient identification of signals under low signal-to-noise ratio.
The technical scheme of the invention is as follows: a satellite communication signal identification method based on a multiport neural network mainly comprises the following steps:
s1, constructing a signal sample library
The invention converts the signals into a sample expression form which is more suitable for the neural network topology structure. The signal eye diagram, the signal vector diagram and the signal time-frequency analysis diagram are adopted as sample expression forms, and the expression forms are easier to obtain compared with the traditional high-dimensional statistical characteristics, and excessive signal original information is not lost. Compared with waveforms, the expression form can highlight the modulation characteristic of signals, and the neural network is easier to classify.
S2, constructing a multiport neural network
The multi-terminal convolution neural network designed by the invention mainly divides the extraction of signal characteristics into 3 stages. In the first stage, the convolutional neural network performs 7×7 convolutional processing on the road eye pattern and the vector diagram and the time frequency respectively, and then performs batch standardization (BN, batch normalization) on the feature diagram output by the first layer network, so as to ensure that the dynamic range of the input feature diagram of each layer of neural network is uniform. The batch normalized data is subjected to a max pool operation to reduce the feature map size. And then connecting the characteristic diagrams obtained by the eye diagram and the vector diagram. In the second stage of feature extraction, in order to eliminate degradation phenomenon caused by over-deep network and make the network converge faster, the invention adopts a ResNet-v1 structure in a residual network (ResNet) structure. And connecting the feature graphs at each end through the feature extraction of the second stage, and carrying out the feature extraction of the third stage. Because the initial picture is subjected to a series of downsampling by the convolutional neural network, global maximum sampling processing is directly carried out on the feature map after batch standardization in the third stage, so that parameters required to be trained by a subsequent network are reduced. The network adopts a ReLu activation function except the output layer which adopts a Softmax activation function. In the network optimization process, an Adam algorithm is adopted to carry out optimal solution solving of network parameters.
S3, model training and target signal identification
Training the constructed network by utilizing the pre-manufactured training samples, and storing the network when the network reaches a steady state. And carrying out carrier frequency coarse estimation on the target test signal through Fourier transformation, estimating a symbol rate by utilizing an envelope spectrum line, carrying out down-conversion on the signal by utilizing the estimated carrier frequency, and carrying out matched filtering by calculating a square root raised cosine function according to the symbol rate. If the target signal has timing deviation, the sampling point value of the signal at the optimal sampling position is extracted to ensure the opening degree of an eye pattern. Obtaining a time-frequency image of the signal by using a time-frequency conversion method; and carrying out normalization and blocking processing on the processed data, and carrying out eye diagram and vector diagram presentation on the blocked data. And finally, modulating and identifying the preprocessed signals by utilizing the stored network, and finally obtaining the signal modulation type.
Drawings
FIG. 1 is a flow chart of the algorithm signal recognition of the present invention;
FIG. 2 is a flow chart of the construction of a signal sample library according to the present invention;
FIG. 3 is a flow chart of the construction of the multi-terminal convolutional neural network of the present invention.
Detailed Description
The following detailed description of embodiments of the invention is, therefore, to be taken in conjunction with the accompanying drawings, and it is to be understood that the scope of the invention is not limited to the specific embodiments. According to an embodiment of the present invention, there is provided a satellite communication signal identifying method based on a multiport neural network, and fig. 1 is a flowchart of algorithm signal identification according to the present invention, including:
and 1, constructing a signal sample library.
Fig. 2 is a flow chart for constructing a signal sample library, in order to make samples more diversified, sampling phase offset and frequency offset are artificially introduced in the sample generation process in consideration of the influence of multiple factors existing in the actual received signals. The method specifically comprises the following steps:
step 1.1, specific modulation mode and random bit data.
The baseband waveform of the signal can be expressed as:
wherein v (t) represents additive white gaussian noise; g (t) represents an equivalent filter including shaping filtering, channel filtering and matched filtering; a, a n Representing the symbol sequence transmitted by the transmitting end. The patterns presented by the symbol sequences are different in different modulation modes. This type of sample is generated each time and the bit stream data required for that modulation type is generated for a particular modulation mode. And a root raised cosine shaping filter is adopted to form a standard sample, the overdriving multiple is 32, and the roll-off coefficient is 0.35.
Step 1.2, random frequency offset phase offset.
The timing deviation possibly existing in the actual signal adoption process is considered, so that the timing deviation is artificially introduced in the simulation signal process. The timing deviation values are randomly valued within a certain range, and the timing deviation values are consistent within the duration of a single sample. Similarly, in consideration of inaccuracy of carrier frequency estimation in the actual parameter estimation process, frequency offset phase offset is artificially introduced in the simulation signal process. The value selection is the same as the sampling phase offset process.
Step 1.3, gaussian channel.
The present invention is directed to satellite communication systems employing gaussian channels. And then, performing matched filtering by using a root raised cosine filter to obtain a target signal.
And step 1.4, generating an eye pattern, a vector diagram and a time-frequency diagram.
During communication signal processing, the eye diagram is typically used as a means to qualitatively reflect the level of inter-symbol crosstalk and noise, and the receive filter can be adjusted by the eye diagram to improve system performance. On the other hand, different modulation schemes have a significant visual difference in eye pattern due to the characteristics of the modulated signal itself. The signal vector diagram is a symbol track obtained by recombining the waveforms of the signal I path and the signal Q path according to corresponding time, and can reflect the phase information of the signal.
Considering that both the conventional eye diagram and the vector diagram are binary images, the degree of aggregation of signals at a certain position is not fully considered. The invention improves the traditional eye diagram and vector diagram, converts the signals into the eye diagram and vector diagram based on the gray level image, and reflects the aggregation degree of the signals at a certain point on the gray level value of the image. The specific generation process is as follows: and 4 symbols are selected as a group of waveforms to generate an eye pattern, 32 times of overdriving is adopted for signals, and each picture consists of 800 symbols. The pixel points of the horizontal axis of the eye pattern correspond to the waveform duration time, so that the continuity of the picture on the horizontal axis is ensured. In order to realize that the subsequent different picture feature images can be connected more conveniently, the signal amplitude is quantized within the range of [ -1.05,1.05] at intervals of 4×32, and the value of each pixel point is the number of sample points falling in the pixel region. In order to make the details in the image more prominent, the following operations are performed on the picture:
wherein im 0 Im is the original image 1 For the enhanced image, α is a scaling factor.
The invention adopts Choi-Williams distribution (Choi-Williams Distribution, CWD) on time-frequency images. The CWD has the greatest characteristic of being capable of effectively inhibiting cross terms, and the expression is as follows:
where σ is the attenuation coefficient, controls the amplitude of the crossover term and is proportional thereto.
The characteristics of the current modulation signals are reflected on different layers by the eye diagram, the vector diagram and the time-frequency diagram, and the invention selects the signal eye diagram, the vector diagram and the time-frequency diagram as signal characteristics to construct a signal sample library and serve as input data of a neural network to realize modulation classification of the signals.
And 2, constructing a multiport neural network model.
Fig. 3 is a flowchart of the multi-terminal convolutional neural network designed by the present invention, and the extraction of signal features is mainly divided into 3 stages. In the first stage, the convolutional neural network performs 7×7 convolutional processing on the eye diagram, the vector diagram and the time-frequency diagram, respectively. And then, carrying out batch standardization (BN, batch normalization) on the characteristic diagrams output by the first layer of network, thereby ensuring the uniformity of the dynamic range of the input characteristic diagrams of each layer of neural network. The batch normalized data is subjected to a max pool operation to reduce the feature map size. And then connecting the obtained characteristic diagrams. In the feature extraction process of the second stage, a residual network (ResNet) is adopted to enable the network to converge more quickly, and feature extraction of the third stage is carried out on feature graph connection of each end. After batch standardization in the third stage, global maximum sampling processing is directly carried out on the feature map, and except for the output layer which adopts a Softmax activation function, the rest layers of the network adopt a ReLu activation function. In the network optimization process, an Adam algorithm is adopted to carry out optimal solution solving of network parameters. The method specifically comprises the following steps:
and 2.1, constructing a convolution layer and a pooling layer.
Each convolution layer in the convolution neural network consists of a plurality of convolution units, and the parameters of each convolution unit are optimized by a back propagation algorithm. In one convolution layer of CNN, the most important parameters are the size of the convolution kernel, the step size of the convolution kernel, the number of convolution kernels, the complement mode, and the like. The convolution kernel is generally initialized in the form of a random decimal matrix, and learns to obtain reasonable weight values in the training process of the network. A direct benefit of sharing weights (convolution kernels) is to reduce the connections between layers of the network while reducing the risk of overfitting.
The formula for calculating the characteristic plane through convolution kernel is as follows:
wherein y is i,j For the corresponding output of the convolution kernel, i.e. the elements of the feature plane, f (·) is the excitation function, θ ij Is a parameter of convolution kernel, x ij Is a parameter of the convolution block, and n is a size of the convolution block.
Assume that the convolutional layer input data structure is W 1 ×H 1 ×D 1 The number of convolution kernels is K, the size of the convolution kernels is F multiplied by F, the step length of the convolution kernels is S, and the complement value is P. Then for the output data structure W 2 ×H 2 ×D 2 The method comprises the following steps:
D 2 ＝K (7)
for this layer, the number of weights per filter is F.F.D 1 The total weight number of all filters is (F.F.D) 1 ) K, the number of bias terms is K.
Typically, a feature with a large dimension is obtained after the convolution layer, and the feature is cut into several regions, and the maximum value or average value is taken to obtain a new feature with a smaller dimension. The pooling process, i.e. the downsampling process, is less computationally intensive than the convolution process, with the aim of simplifying the features rather than extracting them. The present invention simplifies features by maximizing pooling.
Step 2.2, batch standardization.
In order to ensure the uniformity of the dynamic range of the input feature map of each layer of neural network, the invention adopts batch standardization to optimize the variance size and the mean value position, so that the new distribution is more suitable for the real distribution of the data, and the nonlinear expression capability of the model is ensured. The method comprises the following specific steps:
1) The data for each training batch was averaged.
2) The variance of each training batch data was found.
3) And normalizing the training data of the batch by using the obtained mean and variance to obtain 0-1 distribution.
4) Scaling and shifting.
And 2.3, constructing a residual layer.
The invention adds a ResNet-v1 structure in a ResNet model, the structure can rewrite the optimal mapping into H (X) =F (X) +X, and the learning process is skillfully split into two processes: first, the residual function F (X) is learned, and then H (X) =f (X) +x is learned by simple mapping.
H (X) =f (X) +x functions are implemented by adding one "shortcut Connections" to the feed forward network, the shortcuts skip one or more layers in different steps to merge with the main path, the output of this structure is:
y＝F(x,{W i })+x (8)
if the input and output dimensions are different, a linear projection is added, and the calculation formula is as follows:
y＝F(x,{W i })+W s x (9)
and 3, training a network by using the sample library data.
And constructing a network model, training the network by utilizing a pre-manufactured training sample, and storing the network when the network reaches a steady state.
And 4, extracting the characteristics of the target signal.
And carrying out carrier frequency coarse estimation on the target test signal through Fourier transformation, estimating a symbol rate by utilizing an envelope spectrum line, carrying out down-conversion on the signal by utilizing the estimated carrier frequency, and carrying out matched filtering by calculating a square root raised cosine function according to the symbol rate. If the target signal has timing deviation, the sampling point value of the signal at the optimal sampling position is required to be extracted to ensure the opening degree of an eye pattern. The specific calculation formula is as follows:
wherein L is 0 For the symbol length, N is the overdriving multiple, T is the sampling period, thenIs the sequence after the timing extraction.
And carrying out sampling rate conversion on the target signal to obtain 32 times of over-sampled baseband data. And carrying out normalization and blocking processing on the processed data, and carrying out eye diagram and vector diagram presentation on the blocked data.
And 5, identifying the target signal.
And carrying out modulation recognition on the preprocessed signals by utilizing the stored network, and finally obtaining the signal modulation type.
Claims (3)
1. The satellite communication signal identification method based on the multiport neural network is characterized by comprising the following steps of:
s1, constructing a signal sample library
Converting the signal into a sample expression form suitable for a neural network topology;
s2, constructing a multiport neural network
The multi-terminal convolution neural network mainly extracts signal characteristics in 3 stages:
the first stage, the convolution neural network carries out 7×7 convolution processing on the eye pattern and the vector diagram and time respectively, and then carries out batch standardization on the characteristic diagram output by the first layer network; carrying out maximum pooling operation on the data after batch standardization; then, connecting the characteristic diagrams obtained by the eye diagram and the vector diagram;
in the second stage of feature extraction, a ResNet-v1 structure in a residual error network structure is adopted;
connecting the feature graphs of all ends through the feature extraction of the second stage, and carrying out the feature extraction of the third stage; after batch standardization in the third stage, directly carrying out global maximum sampling treatment on the feature map;
s3, model training and target signal identification
Training the constructed network by utilizing a pre-manufactured training sample, and storing the network when the network reaches a steady state; for a target test signal, carrying out carrier frequency coarse estimation through Fourier transformation, estimating a symbol rate by utilizing an envelope spectrum line, carrying out down-conversion on the signal by using the estimated carrier frequency, and carrying out matched filtering by calculating a square root raised cosine function according to the symbol rate; if the target signal has timing deviation, sampling point values of the signal at the optimal sampling position are required to be extracted so as to ensure the opening degree of an eye pattern;
obtaining a time-frequency image of the signal by using a time-frequency conversion method; normalizing and blocking the processed data, and presenting an eye diagram and a vector diagram for the blocked data;
and finally, modulating and identifying the preprocessed signals by utilizing the stored network, and finally obtaining the signal modulation type.
2. The method for identifying satellite communication signals based on the multiport neural network according to claim 1, wherein a signal eye diagram, a signal vector diagram and a signal time-frequency analysis diagram are adopted as sample expression forms in S1.
3. The satellite communication signal identification method based on the multiport neural network according to claim 1, wherein in S2, except for the output layer, a Softmax activation function is adopted by the network, and a ReLu activation function is adopted by all the other layers; in the network optimization process, an Adam algorithm is adopted to carry out optimal solution solving of network parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210425192.5A CN114785649B (en) | 2022-04-21 | 2022-04-21 | Satellite communication signal identification method based on multiport neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210425192.5A CN114785649B (en) | 2022-04-21 | 2022-04-21 | Satellite communication signal identification method based on multiport neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114785649A CN114785649A (en) | 2022-07-22 |
| CN114785649B true CN114785649B (en) | 2023-11-14 |
Family
ID=82430520
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210425192.5A Active CN114785649B (en) | 2022-04-21 | 2022-04-21 | Satellite communication signal identification method based on multiport neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114785649B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107342962A (en) * | 2017-07-03 | 2017-11-10 | 北京邮电大学 | Deep learning intelligence Analysis On Constellation Map method based on convolutional neural networks |
| WO2021013334A1 (en) * | 2019-07-22 | 2021-01-28 | Toyota Motor Europe | Depth maps prediction system and training method for such a system |
| CN112347871A (en) * | 2020-10-23 | 2021-02-09 | 中国电子科技集团公司第七研究所 | Interference signal modulation identification method for communication carrier monitoring system |
| CN113141325A (en) * | 2021-04-14 | 2021-07-20 | 北京邮电大学 | Training method, identification method and device for optical OFDM signal subcarrier modulation format identification model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11640523B2 (en) * | 2019-01-02 | 2023-05-02 | Rohde & Schwarz Gmbh & Co. Kg | Method for training a neural network, method for automatically characterizing a measurement signal, measurement apparatus and method for analyzing |
-
2022
- 2022-04-21 CN CN202210425192.5A patent/CN114785649B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107342962A (en) * | 2017-07-03 | 2017-11-10 | 北京邮电大学 | Deep learning intelligence Analysis On Constellation Map method based on convolutional neural networks |
| WO2021013334A1 (en) * | 2019-07-22 | 2021-01-28 | Toyota Motor Europe | Depth maps prediction system and training method for such a system |
| CN112347871A (en) * | 2020-10-23 | 2021-02-09 | 中国电子科技集团公司第七研究所 | Interference signal modulation identification method for communication carrier monitoring system |
| CN113141325A (en) * | 2021-04-14 | 2021-07-20 | 北京邮电大学 | Training method, identification method and device for optical OFDM signal subcarrier modulation format identification model |
Non-Patent Citations (1)
| Title |
|---|
| 基于卷积神经网络的调制样式识别研究;陈昌美;李艳斌;;信息技术(01);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114785649A (en) | 2022-07-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598677B (en) | Space-time multi-channel deep learning system for automatic modulation recognition | |
| CN110163282B (en) | Modulation mode identification method based on deep learning | |
| CN109657604B (en) | Satellite amplitude-phase signal identification and demodulation method and device based on cyclic neural network | |
| CN112347871B (en) | Interference signal modulation identification method for communication carrier monitoring system | |
| CN112418014A (en) | Modulation signal identification method based on wavelet transformation and convolution long-short term memory neural network | |
| Zhang et al. | A data preprocessing method for automatic modulation classification based on CNN | |
| CN112749633B (en) | Separate and reconstructed individual radiation source identification method | |
| Jeong et al. | Spectrogram-based automatic modulation recognition using convolutional neural network | |
| CN114881092A (en) | Signal modulation identification method based on feature fusion | |
| CN112910812B (en) | Modulation mode identification method for deep learning based on space-time feature extraction | |
| CN108052956A (en) | Wireless light communication subcarrier modulation constellation recognition methods under a kind of atmospheric turbulance | |
| CN114422311B (en) | Signal modulation recognition method and system combining deep neural network and expert priori features | |
| CN112737992B (en) | Underwater sound signal modulation mode self-adaptive in-class identification method | |
| CN111612130B (en) | Frequency shift keying communication signal modulation mode identification method | |
| CN112242969A (en) | Novel single-bit OFDM receiver based on model-driven deep learning | |
| Ma et al. | CNN-based automatic modulation recognition of wireless signal | |
| CN111507299A (en) | Method for identifying STBC (space time Block coding) signal on frequency domain by using convolutional neural network | |
| Kong et al. | A transformer-based CTDNN structure for automatic modulation recognition | |
| Ali et al. | Modulation format identification using supervised learning and high-dimensional features | |
| CN114785649B (en) | Satellite communication signal identification method based on multiport neural network | |
| He et al. | Deep learning-based automatic modulation recognition algorithm in non-cooperative communication systems | |
| CN116760491A (en) | Signal-to-noise ratio estimation method based on deep learning | |
| CN115913849A (en) | Electromagnetic signal identification method based on one-dimensional complex value residual error network | |
| CN116628566A (en) | Communication signal modulation classification method based on aggregated residual transformation network | |
| Essai et al. | Automatic Modulation Classification: Convolutional Deep Learning Neural Networks Approaches |
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 |