CN115062658B - Overlapping radar signal modulation type identification method based on self-adaptive threshold network - Google Patents
Overlapping radar signal modulation type identification method based on self-adaptive threshold network Download PDFInfo
- Publication number
- CN115062658B CN115062658B CN202210667010.5A CN202210667010A CN115062658B CN 115062658 B CN115062658 B CN 115062658B CN 202210667010 A CN202210667010 A CN 202210667010A CN 115062658 B CN115062658 B CN 115062658B
- Authority
- CN
- China
- Prior art keywords
- module
- network
- frequency
- time
- radar signals
- 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 38
- 238000012549 training Methods 0.000 claims abstract description 5
- 238000010586 diagram Methods 0.000 claims description 22
- 230000004913 activation Effects 0.000 claims description 21
- 230000003044 adaptive effect Effects 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 7
- AYFVYJQAPQTCCC-GBXIJSLDSA-N L-threonine Chemical compound C[C@@H](O)[C@H](N)C(O)=O AYFVYJQAPQTCCC-GBXIJSLDSA-N 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000001125 extrusion Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 27
- 238000013528 artificial neural network Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000739 chaotic effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000001627 detrimental effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an overlapping radar signal modulation type identification method based on a self-adaptive threshold network, which comprises the following steps: s1, simulating an original radar signal and modulation parameters; s2, overlapping original radar signals; s3, extracting time-frequency domain features; s4, constructing a basic module; s5, constructing Inception modules; s6, constructing a self-adaptive threshold module; s7, constructing a probability module; s8, forming an SE-INCEPATNET network; s9, training an SE-INCEPATNET network; and S10, identifying the modulation type of the overlapped radar signals by adopting the SE-INCEPATNET network after training. According to the invention, the time-frequency analysis method is utilized to extract the characteristics of the overlapped radar signals, the depth convolution network Inception is based on the characteristics of different receptive fields, the SE module is used for reducing noise influence, the self-adaptive threshold module is constructed to solve the problem of difficult threshold setting in multi-classification tasks, and the recognition accuracy under the condition of low signal-to-noise ratio is improved while the recognition of the overlapped radar signals is realized.
Description
Technical Field
The invention belongs to the technical field of radar signal processing, and is particularly suitable for radar radiation source signal identification in electronic countermeasure, in particular relates to an overlapping radar signal modulation type identification method based on a self-adaptive threshold network.
Background
Modulation identification is widely used for threat level assessment and combat strategy formulation.
As the pulse stream density increases, the radar receiver receives multiple radar signals simultaneously, which overlap not only in the time domain but also in the frequency domain. At low signal-to-noise ratios, both overlap and noise can cause certain features of the signal to become chaotic, thereby increasing the difficulty of modulation recognition. For overlap problems, the instantaneous characteristics and statistics of radar signals are often used for modulation identification, reflecting the inherent characteristics of the signal. Document "Lu Mingquan,Xiao Xianci,and Li Lemin,Ar modeling-based features extraction of multiple signals for modulation recognition,in ICSP'98.1998Fourth International Conference on Signal Processing(Cat.No.98TH8344),1998,vol.2,pp.1385–1388 vol.2" uses the instantaneous frequency and bandwidth of overlapping radar signals to achieve modulation identification, but is not applicable to signals where the frequency domains overlap. Document "Kuang-dai Li,Li-li Guo,Rong Shi,and Dan Wu,Modulation recognition method based on high order cyclic cumulants for time-frequency overlapped two-signal in the single-channel,in 2008Congress on Image and Signal Processing,2008,vol.5,pp.474–478." uses cyclic statistics and minimum error criteria to identify overlapping signals, and the identification accuracy can only reach 90% when the signal-to-noise ratio is greater than 10 dB.
For modulation recognition at low signal-to-noise ratios, neural networks can be used for modulation recognition in general, by learning robust features of the mined signal from a large amount of data. Literature "Shunjun Wei,Qizhe Qu,Hao Su,Mou Wang,Jun Shi, and Xiaojun Hao,Intra-pulse modulation radar signal recognition based on cldn network,IET Radar,Sonar&Navigation,vol.14,no.6,pp.803–810,2020." is combined with CNN, long Short-Term Memory (LSTM) and Deep Neural Network (DNN) to construct an identification network, and the accuracy is 90% at-6 dB. However, the above documents only identify a single radar signal, and there is still a lack of correlation research for the identification of overlapping radar signals at low signal-to-noise ratios. Document "Yehan Ren,Weibo Huo,Jifang Pei,Yulin Huang,and Jianyu Yang,Automatic modulation recognition for overlapping radar signals based on multi-domain se-resnext,in 2021IEEE Radar Conference(Radar Conf21),2021,pp.1–6." proposes a network for single tag classification to identify overlapping signals, but single tag classification networks have poor flexibility. Therefore, the identification of overlapping radar signals at low signal-to-noise ratios still requires further investigation.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides the overlapping radar signal modulation type identification method based on the self-adaptive threshold network, which improves the identification accuracy under the condition of low signal-to-noise ratio while realizing the identification of the overlapping radar signals.
The aim of the invention is realized by the following technical scheme: the method for identifying the modulation type of the overlapping radar signals based on the self-adaptive threshold network comprises the following steps:
S1, simulating an original radar signal and modulation parameters;
S2, overlapping the simulated original radar signals to generate overlapped radar signals;
s3, extracting time-frequency domain characteristics of the overlapped radar signals, and representing the time-frequency domain characteristics as a time-frequency diagram in a chart form, wherein the horizontal axis represents time and the vertical axis represents frequency;
S4, constructing a basic module;
s5, constructing Inception modules: the Inception module comprises a plurality of basic modules with different convolution kernel sizes and is used for extracting information with different dimensions in the time-frequency diagram and combining the information with the information;
S6, constructing an adaptive threshold module: the method comprises a full connection layer, a ReLU activation function, a full connection layer and Sigmoid activation, which are connected in sequence; taking the output of the Inception module in the step S5 as input to obtain a corresponding threshold;
S7, constructing a probability module: adopting a structure of a full connection layer, a ReLU activation function and a Sigmoid activation function which are connected in sequence, and taking the output of the Inception module in the step S5 as input to obtain posterior probability;
s8, forming the submodules constructed in the steps S5 to S7 into an SE-INCEPATNET network, wherein the method specifically comprises the following steps: after the probability module and the self-adaptive threshold module are connected in parallel, the probability module and the self-adaptive threshold module are cascaded behind Inception modules to form an SE-INCEPATNET network; inputting a time-frequency diagram processed by the S3, and outputting a classification result;
S9, training an SE-INCEPATNET network, specifically: inputting a time-frequency diagram of the overlapped radar signals into a SE-INCEPATNET network for forward propagation, and calculating a cost function value; updating parameters of the SE-INCEPATNET network by using a backward propagation algorithm based on gradient descent; iterating the backward propagation process until the cost function converges, thereby obtaining a trained SE-INCEPATNET network;
And S10, identifying the modulation type of the overlapped radar signals by adopting the SE-INCEPATNET network trained in the step S9.
Further, the specific method for extracting the time-frequency characteristic in the step S3 is as follows: extracting time-frequency characteristics by using FSST algorithm, wherein the FSST calculation formula is as follows:
Where g (0) represents the value of the sliding window function g (t) at time 0, δ () is the Dirac impulse function; v f (η, t) denotes a signal after STFT of the superimposed radar signal; ω represents FSST frequency after reassignment, η represents FSST frequency before reassignment, t represents time, Representing the entire real number domain; /(I)Is the instantaneous frequency, defined as:
Further, the base module includes a concatenated convolutional layer, a batch normalization layer BN, a ReLU activation layer, a max-pooling layer, and an extrusion excitation module SE; the SE is used for denoising and comprises a global average pooling layer GAP, a full connection layer FC, a ReLU activation function layer, a full connection layer FC and a Sigmoid activation function layer which are connected in sequence.
Further, the detailed processing procedure in the step S5 is as follows:
S5.1, inputting the time-frequency diagram into 4 different basic modules simultaneously for processing to obtain features with different dimensions, wherein the convolution kernel sizes of the four basic modules are 3×7,7×3,5×5 and 3×3 in sequence;
s5.2, respectively inputting the output of each basic module in the step S5.1 into a basic module with the convolution kernel size of 3 multiplied by 3;
S5.3, fusing the outputs of the four basic modules in the step S5.2 together;
s5.4, inputting the fused output in the step S5.3 into 4 basic modules with the convolution kernel size of 3 multiplied by 3 in sequence to obtain an output result.
Further, the step S9 specifically includes the following steps:
S9.1, forward propagation;
S9.2, calculating a cost function value by taking a binary cross entropy loss function as a cost function, wherein the calculating method comprises the following steps of:
Wherein BCE is a binary cross entropy loss function, prob ε R 1×n is the output probability, lab ε R 1×n is the true value of the overlap signal in the form of a single hot code, thre ε R 1×n is the threshold, and n is the total number of modulation types;
And S9.3, updating network parameters based on a backward propagation algorithm of gradient descent.
The beneficial effects of the invention are as follows: aiming at the problem of less overlapping radar signal identification in the current research, the invention provides an identification method suitable for complex electromagnetic environment and high-density pulse stream conditions, the time-frequency analysis method is utilized to extract the characteristics of overlapping radar signals, the depth convolution network Inception module is based on the characteristics of different receptive fields, the SE module is used for reducing noise influence, and the self-adaptive threshold module is constructed to solve the problem of difficult threshold setting in multi-classification tasks. The effective identification of various overlapped radar signals in a low signal-to-noise ratio environment is realized; the method improves the recognition accuracy under the condition of low signal-to-noise ratio while realizing the recognition of the overlapped radar signals. The method has the advantages of flexibility, accuracy, strong generalization capability, strong robustness and the like.
Drawings
FIG. 1 is a flow chart of a method for identifying the modulation type of an overlapping radar signal according to the present invention;
FIG. 2 is a block diagram of a basic module of the present invention;
FIG. 3 is a block diagram of a Inception module of the present invention;
FIG. 4 is a block diagram of an adaptive threshold module of the present invention;
Fig. 5 is a diagram of the recognition result of the modulation type of the overlapped radar signal provided in this example.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings.
As shown in fig. 1, the method for identifying the modulation type of the overlapping radar signal based on the adaptive threshold network comprises the following steps:
S1, simulating an original radar signal and modulation parameters;
Table 1 shows the original radar signal parameters of the present embodiment. The present embodiment emulates five typical radar signal modulation types, including Linear Frequency Modulation (LFM), nonlinear frequency modulation (NLFM), normal wave (CW) signals, frequency Agility (FA) signals, and binary frequency shift keying (Costas). The simulation signals are all baseband signals, the sampling frequency is 700MHz, the bandwidth range is 100-300 MHz, the pulse width range is 3-5 us, the signal-to-noise ratio variation range is-12 dB-5 dB, and the signal overlapping rate is set to be 30% -100%.
TABLE 1
| Parameters (parameters) | Range of |
| fs | 700MHz |
| B | 10～300MHz |
| Pw | 3～5us |
| SNR | -12dB～5dB |
| Od | 30％～100％ |
S2, overlapping the simulated original radar signals to generate overlapped radar signals; table 2 shows the radar signal overlapping situation, including the case of overlapping of 2 modulation types and 3 modulation types, and also including the case of overlapping of different modulation types of the same modulation type.
TABLE 2
S3, extracting time-frequency domain characteristics of the overlapped radar signals, and representing the time-frequency domain characteristics as a time-frequency diagram in a chart form, wherein the horizontal axis represents time and the vertical axis represents frequency; the specific method for extracting the time-frequency characteristics comprises the following steps: extracting time-frequency characteristics by using FSST (Fourier synchronous compression transformation) algorithm, wherein the FSST calculation formula is as follows:
Where g (0) represents the value of the sliding window function g (t) at time 0, δ () is the Dirac impulse function; v f (η, t) denotes a signal after STFT (short time fourier transform) of the superimposed radar signal; ω represents FSST the frequency after reassignment, η represents FSST the frequency before reassignment (frequency after STFT conversion), t represents time, Representing the whole real number domain, equivalent toIs the instantaneous frequency, defined as:
the different radar signals generated in step S2 are interleaved with each other, so that the radar signals overlap in the time domain and overlap in the frequency domain. Conventional fourier transforms and the like are not suitable for this case, however, the time-frequency characteristics of the signal may reflect the frequency-time-dependent relationship thereof. Since the frequency of radar signals of different modulation types varies with time, the time-frequency characteristics can be used to distinguish between overlapping signals.
S4, constructing a basic module; as shown in fig. 2, the base module includes a sequential convolution layer, a batch normalization layer BN, a ReLU activation layer, a max-pooling layer, and an extrusion excitation module SE; the SE is used for denoising and comprises a global average pooling layer GAP, a full connection layer FC, a ReLU activation function layer, a full connection layer FC and a Sigmoid activation function layer which are connected in sequence; the SE submodule learns the weight coefficient of each channel, and then the noise threshold is obtained by multiplying the weight by GAP output; finally, denoising the output result of the maximum pooling layer by taking the output of the SE module as a threshold: the value in each characteristic channel is compared with its noise threshold, and a value less than the threshold is considered to be caused by noise, which is detrimental to modulation recognition, and is set to zero.
S5, constructing Inception modules: the Inception module comprises a plurality of basic modules with different convolution kernel sizes and is used for extracting information with different dimensions in the time-frequency diagram and combining the information with the information;
When the frequency of the radar signal changes slowly with time, a flat curve is shown on the time-frequency chart, so that a basic module with a convolution kernel size of 3×7 can be used for capturing the characteristic of the frequency change with time. Likewise, a convolution kernel size of 7 x 3 may be used to capture characteristics of rapid changes in signal frequency over time. When a plurality of signals overlap, it appears on the time-frequency diagram that a plurality of frequency components exist in the same time dimension, and thus the number of overlapping signals can be obtained using the basic block 7×3. In addition, basic blocks with convolution kernel sizes of 3×3 and 5×5 can be used to capture more detailed features. The output of the basic module is fused and then cascaded, so that the characteristics with higher dimensionality can be extracted. "base module m×n" means a base module having a convolution kernel size of m×n. As shown in fig. 3, the detailed processing procedure of this step is:
S5.1, inputting the time-frequency diagram into 4 different basic modules simultaneously for processing to obtain features with different dimensions, wherein the convolution kernel sizes of the four basic modules are 3×7,7×3,5×5 and 3×3 in sequence;
s5.2, respectively inputting the output of each basic module in the step S5.1 into a basic module with the convolution kernel size of 3 multiplied by 3;
S5.3, fusing the outputs of the four basic modules in the step S5.2 together;
s5.4, inputting the fused output in the step S5.3 into 4 basic modules with the convolution kernel size of 3 multiplied by 3 in sequence to obtain an output result.
S6, constructing an adaptive threshold module: including fully connected layer, reLU activation function, fully connected layer and Sigmoid activation in parallel, as shown in fig. 4; taking the output of the Inception module in the step S5 as input to obtain a corresponding threshold; in the multi-label classification task, the posterior probability output by the network is compared with a threshold to obtain a classification result, and different classification performances are realized by different threshold settings. To avoid the difficulty of selecting the optimal threshold, the invention uses an adaptive threshold block to obtain the adaptive threshold of each modulation type so as to improve the classification performance and generalization capability of the network.
S7, constructing a probability module: adopting a structure of a full connection layer, a ReLU activation function and a Sigmoid activation function which are connected in sequence, and taking the output of the Inception module in the step S5 as input to obtain posterior probability;
s8, forming the submodules constructed in the steps S5 to S7 into an SE-INCEPATNET network, wherein the method specifically comprises the following steps: after the probability module and the self-adaptive threshold module are connected in parallel, the probability module and the self-adaptive threshold module are cascaded behind Inception modules to form an SE-INCEPATNET network; inputting a time-frequency diagram processed by the S3, and outputting a classification result;
After the original radar signal time sequence is processed in the step S3, a two-dimensional time-frequency diagram is obtained, after the time-frequency diagram is input into an SE-INCEPATNET network, the posterior probability corresponding to each category can be obtained through a probability sub-module, the threshold can be obtained through a self-adaptive threshold sub-module, and the category corresponding to the data with the value larger than the threshold in the posterior probability is the classification result of the network structure.
S9, training an SE-INCEPATNET network, specifically: inputting a time-frequency diagram of the overlapped radar signals into a SE-INCEPATNET network for forward propagation, and calculating a cost function value; updating parameters of the SE-INCEPATNET network by using a backward propagation algorithm based on gradient descent; iterating the backward propagation process until the cost function converges, thereby obtaining a trained SE-INCEPATNET network; the method specifically comprises the following steps:
S9.1, forward propagation;
S9.2, calculating a cost function value by taking a binary cross entropy loss function as a cost function, wherein the calculating method comprises the following steps of:
Wherein BCE is a binary cross entropy loss function, prob ε R 1×n is the output probability, lab ε R 1×n is the true value of the overlap signal in the form of a single hot code, thre ε R 1×n is the threshold, and n is the total number of modulation types;
And S9.3, updating network parameters based on a backward propagation algorithm of gradient descent.
And S10, identifying the modulation type of the overlapped radar signals by adopting the SE-INCEPATNET network trained in the step S9.
To demonstrate the identification performance of the proposed SE-INCEPATNET network, the present invention compares the accuracy and recall of SE-INCEPATNET with AlexNet, resNet and GoogleNet. Fig. 5 shows the modulation type recognition result of the embodiment of the present invention. Wherein, fig. 5 (a) is an accuracy image of the recognition result according to SNR; fig. 5 (b) is a recall image of the recognition result as a function of SNR.
As can be seen from fig. 5, the recall and accuracy of SE-INCEPATNET are superior to other networks at low signal-to-noise ratios because the SE module reduces the effects of noise. Under the high signal-to-noise ratio, the SE module has little influence on the identification, and the identification performance is improved by the self-adaptive threshold block. Furthermore, SE-INCEPATNET performs better than ResNet and AlexNet due to Inception modules. Simulation shows that the method can effectively realize modulation identification of the overlapped radar signals under the low signal-to-noise ratio. At-10 dB, the recall rate of the method is 90.6%, and the accuracy rate is 95.0%. The method is statistically well behaved in the identification of each modulation type. The accuracy of CW is above 95%, the accuracy of LFM and NLFM is above 90%, and the accuracy of Costas and FA is above 85%.
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit thereof, and such modifications and combinations remain within the scope of the present disclosure.
Claims (5)
1. The method for identifying the modulation type of the overlapping radar signals based on the self-adaptive threshold network is characterized by comprising the following steps:
S1, simulating an original radar signal and modulation parameters;
S2, overlapping the simulated original radar signals to generate overlapped radar signals;
S3, extracting time-frequency domain characteristics of the overlapped radar signals, and representing the time-frequency domain characteristics as a time-frequency diagram in a chart form, wherein the horizontal axis represents time and the vertical axis represents frequency;
S4, constructing a basic module;
S5, constructing Inception modules: the Inception module comprises a plurality of basic modules with different convolution kernel sizes and is used for extracting information with different dimensions in the time-frequency diagram and combining the information with the information;
s6, constructing an adaptive threshold module: the method comprises a full connection layer, a ReLU activation function, a full connection layer and Sigmoid activation, which are connected in sequence; taking the output of the Inception module in the step S5 as input to obtain a corresponding threshold;
S7, constructing a probability module: adopting a structure of a full connection layer, a ReLU activation function and a Sigmoid activation function which are connected in sequence, and taking the output of the Inception module in the step S5 as input to obtain posterior probability;
S8, forming the submodules constructed in the steps S5 to S7 into an SE-INCEPATNET network, wherein the method specifically comprises the following steps: after the probability module and the self-adaptive threshold module are connected in parallel, the probability module and the self-adaptive threshold module are cascaded behind the Inception module to form an SE-INCEPATNET network; inputting a time-frequency diagram processed by the S3, and outputting a classification result;
S9, training an SE-INCEPATNET network, specifically: inputting a time-frequency diagram of the overlapped radar signals into a SE-INCEPATNET network for forward propagation, and calculating a cost function value; updating parameters of the SE-INCEPATNET network by using a backward propagation algorithm based on gradient descent; iterating the backward propagation process until the cost function converges, thereby obtaining a trained SE-INCEPATNET network;
And S10, identifying the modulation type of the overlapped radar signals by adopting the SE-INCEPATNET network trained in the step S9.
2. The method for identifying the modulation type of the overlapping radar signals based on the adaptive threshold network according to claim 1, wherein the specific method for extracting the time-frequency characteristics in the step S3 is as follows: extracting time-frequency characteristics by using FSST algorithm, wherein the FSST calculation formula is as follows:
Where g (0) represents the value of the sliding window function g (t) at time 0, δ () is the Dirac impulse function; v f (η, t) denotes the signal after STFT of the overlapping radar signals; ω represents FSST frequency after reassignment, η represents FSST frequency before reassignment, t represents time, Representing the entire real number domain; /(I)Is the instantaneous frequency, defined as:
3. The adaptive threshold network-based overlapping radar signal modulation type identification method according to claim 1, wherein the basic module comprises a sequential convolution layer, a batch normalization layer BN, a ReLU activation layer, a max pooling layer and an extrusion excitation module SE; the SE is used for denoising and comprises a global average pooling layer GAP, a full connection layer FC, a ReLU activation function layer, a full connection layer FC and a Sigmoid activation function layer which are connected in sequence.
4. The method for identifying the modulation type of the overlapping radar signals based on the adaptive threshold network according to claim 1, wherein the detailed processing procedure in step S5 is as follows:
S5.1, inputting the time-frequency diagram into 4 different basic modules simultaneously for processing to obtain features with different dimensions, wherein the convolution kernel sizes of the four basic modules are 3×7,7×3,5×5 and 3×3 in sequence;
s5.2, respectively inputting the output of each basic module in the step S5.1 into a basic module with the convolution kernel size of 3 multiplied by 3;
S5.3, fusing the outputs of the four basic modules in the step S5.2 together;
S5.4, inputting the fused output in the step S5.3 into 4 basic modules with the convolution kernel size of 3 multiplied by 3 in sequence to obtain an output result.
5. The method for identifying the modulation type of the overlapping radar signals based on the adaptive threshold network according to claim 1, wherein said step S9 specifically comprises the steps of:
S9.1, forward propagation;
S9.2, calculating a cost function value by taking a binary cross entropy loss function as a cost function, wherein the calculating method comprises the following steps of:
Wherein BCE is a binary cross entropy loss function, prob ε R 1×n is the output probability, lab ε R 1×n is the true value of the overlap signal in the form of a single hot code, thre ε R 1×n is the threshold, and n is the total number of modulation types;
And S9.3, updating network parameters based on a backward propagation algorithm of gradient descent.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210667010.5A CN115062658B (en) | 2022-06-14 | 2022-06-14 | Overlapping radar signal modulation type identification method based on self-adaptive threshold network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210667010.5A CN115062658B (en) | 2022-06-14 | 2022-06-14 | Overlapping radar signal modulation type identification method based on self-adaptive threshold network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115062658A CN115062658A (en) | 2022-09-16 |
| CN115062658B true CN115062658B (en) | 2024-05-28 |
Family
ID=83200331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210667010.5A Active CN115062658B (en) | 2022-06-14 | 2022-06-14 | Overlapping radar signal modulation type identification method based on self-adaptive threshold network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115062658B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446877A (en) * | 2018-09-01 | 2019-03-08 | 哈尔滨工程大学 | A kind of radar emitter signal Modulation Identification method of joint multidimensional characteristic migration fusion |
| CN111505632A (en) * | 2020-06-08 | 2020-08-07 | 北京富奥星电子技术有限公司 | Ultra-wideband radar action attitude identification method based on power spectrum and Doppler characteristics |
| CN113033473A (en) * | 2021-04-15 | 2021-06-25 | 中国人民解放军空军航空大学 | ST2DCNN + SE-based radar overlapped signal identification method |
| CN113156391A (en) * | 2021-04-25 | 2021-07-23 | 电子科技大学 | Radar signal multi-dimensional feature intelligent sorting method |
| CN114202056A (en) * | 2021-11-11 | 2022-03-18 | 中国船舶重工集团公司第七一五研究所 | Small sample underwater target identification method based on multi-platform auditory perception feature deep transfer learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017127421A1 (en) * | 2016-01-18 | 2017-07-27 | Qoscience, Inc. | Method and apparatus for the detection of distortion or corruption of cellular communication signals |
-
2022
- 2022-06-14 CN CN202210667010.5A patent/CN115062658B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109446877A (en) * | 2018-09-01 | 2019-03-08 | 哈尔滨工程大学 | A kind of radar emitter signal Modulation Identification method of joint multidimensional characteristic migration fusion |
| CN111505632A (en) * | 2020-06-08 | 2020-08-07 | 北京富奥星电子技术有限公司 | Ultra-wideband radar action attitude identification method based on power spectrum and Doppler characteristics |
| CN113033473A (en) * | 2021-04-15 | 2021-06-25 | 中国人民解放军空军航空大学 | ST2DCNN + SE-based radar overlapped signal identification method |
| CN113156391A (en) * | 2021-04-25 | 2021-07-23 | 电子科技大学 | Radar signal multi-dimensional feature intelligent sorting method |
| CN114202056A (en) * | 2021-11-11 | 2022-03-18 | 中国船舶重工集团公司第七一五研究所 | Small sample underwater target identification method based on multi-platform auditory perception feature deep transfer learning |
Non-Patent Citations (4)
| Title |
|---|
| Modulation recognition for radar emitter signals based on convolutional neural network and fusion features;Jingpeng Gao等;《Emerging Telecommunications Technologies》;20190411;全文 * |
| Modulation Recognition of Overlapping Radar Signals Under Low SNR Based on Se-Incepatnet;Hao Wang等;《 IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium》;20220928;全文 * |
| 基于AlexNet模型的雷达信号调制类型识别;郭立民;陈鑫;陈涛;;吉林大学学报(工学版);20190508(03);全文 * |
| 基于深度迁移学习的窄带雷达群目标识别方法;梁复台;李宏权;张晨浩;;兵器装备工程学报;20190904(04);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115062658A (en) | 2022-09-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110532932B (en) | Method for identifying multi-component radar signal intra-pulse modulation mode | |
| CN107220606A (en) | The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks | |
| CN113298846B (en) | Interference intelligent detection method based on time-frequency semantic perception | |
| CN104794264B (en) | A kind of radar communication waveform design method based on sparse frequency | |
| CN112115821B (en) | Multi-signal intelligent modulation mode identification method based on wavelet approximate coefficient entropy | |
| CN116047427B (en) | Small sample radar active interference identification method | |
| CN112014801A (en) | Composite interference identification method based on SPWVD and improved AlexNet | |
| CN111985349B (en) | Classification recognition method and system for radar received signal types | |
| CN114897002A (en) | LPINet-based low interception probability radar signal intra-pulse modulation identification method | |
| Cotter | Multiple snapshot matching pursuit for direction of arrival (DOA) estimation | |
| Van Huynh et al. | Generative AI for physical layer communications: A survey | |
| CN111446998B (en) | Direction-of-arrival estimation method based on deep learning | |
| CN115982613A (en) | Signal modulation identification system and method based on improved convolutional neural network | |
| CN115062658B (en) | Overlapping radar signal modulation type identification method based on self-adaptive threshold network | |
| CN115267713B (en) | Intermittent sampling interference identification and suppression method based on semantic segmentation | |
| Guven et al. | Classifying LPI radar waveforms with time-frequency transformations using multi-stage CNN system | |
| Reddy et al. | Optimization of orthogonal adaptive waveform design in presence of compound Gaussian clutter for MIMO radar | |
| CN115932895A (en) | GNSS deception jamming intelligent detection method based on MHA-1DCNN | |
| CN115378777A (en) | Method for identifying underwater communication signal modulation mode in alpha stable distribution noise environment | |
| Zhang et al. | Characteristic function based parameter estimation for ocean ambient noise | |
| Li et al. | Uplink channel estimation for intelligent reflecting surface aided wireless communication systems with condition gan | |
| Chen et al. | EMRNet: Efficient Modulation Recognition Networks for Continuous-Wave Radar Signals | |
| CN116699531B (en) | Radar signal noise reduction method, system and storage medium based on complex network | |
| CN116804740B (en) | Inverted U-Net-based passive radar signal intelligent detection method | |
| CN117997690B (en) | Class comparison learning type underwater sound signal modulation recognition system and recognition method |
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 |