CN114826857B - OFDM signal generation method based on generation countermeasure network - Google Patents
OFDM signal generation method based on generation countermeasure network Download PDFInfo
- Publication number
- CN114826857B CN114826857B CN202210310083.9A CN202210310083A CN114826857B CN 114826857 B CN114826857 B CN 114826857B CN 202210310083 A CN202210310083 A CN 202210310083A CN 114826857 B CN114826857 B CN 114826857B
- Authority
- CN
- China
- Prior art keywords
- frequency
- time
- layer
- dimensional pattern
- dimensional
- 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 19
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 34
- 239000011159 matrix material Substances 0.000 claims abstract description 33
- 239000013598 vector Substances 0.000 claims abstract description 16
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 56
- 238000006243 chemical reaction Methods 0.000 claims description 22
- 238000009826 distribution Methods 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 12
- 238000010586 diagram Methods 0.000 claims description 12
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 125000004122 cyclic group Chemical group 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000009977 dual effect Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 2
- 230000003213 activating effect Effects 0.000 claims 6
- 238000004891 communication Methods 0.000 abstract description 9
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000000605 extraction Methods 0.000 abstract description 3
- 238000012360 testing method Methods 0.000 abstract description 2
- 239000000969 carrier Substances 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 2
- 101000595198 Homo sapiens Podocalyxin Proteins 0.000 description 1
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 102100036031 Podocalyxin Human genes 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012109 statistical procedure Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000007704 transition 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
- H04L27/26—Systems using multi-frequency codes
- H04L27/2601—Multicarrier modulation systems
- H04L27/2602—Signal structure
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention discloses an OFDM signal generation method based on a generation countermeasure network, which generates a complex protocol signal using OFDM modulation and belongs to a signal reconstruction technology in the communication countermeasure field. Aiming at the problem of difficult extraction of the time domain characteristics of the OFDM signals, firstly, preprocessing the time domain OFDM signals by using FFT, and splicing the processed frequency domain symbol vectors into a two-dimensional data matrix; the data matrix is then stored as a time-frequency two-dimensional pattern in the form of a gray image for training and testing of the GAN. In addition, the dual-discriminator GAN adopts one generator and two discriminators to play games at the same time, the generator is used for generating a time-frequency two-dimensional pattern to cheat the two discriminators, and the two discriminators respectively distinguish the generated pattern and the real pattern from two aspects of a subcarrier structure and constellation density of a modulation symbol so as to ensure that the generated signal meets the requirements of OFDM time-frequency domain characteristics.
Description
Technical Field
The invention belongs to a signal reconstruction technology in the field of communication countermeasure, and particularly relates to a complex protocol signal generation method using orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) modulation.
Background
In the face of unknown communication signals, the conventional communication countermeasure mechanism needs to perform channel estimation and signal parameter estimation, such as modulation mode, code rate, carrier frequency, frequency offset, and the like, and then perform tracking interference on the signals or directly perform blocking interference under the condition that the identification of external signals fails. However, with the development and progress of communication and network technologies, electromagnetic signals are characterized by multiple modes, multiple systems, intellectualization, high self-adaption, rapid change and the like, and due to the differentiated requirements of different communication scenes, a wireless communication system designs multiple complex protocols for various scenes, and the problems of difficult identification and difficult generation of the traditional interference method are faced with unknown complex protocol signals. In addition, when generating the interference signal, the traditional interference method is realized through hardware modulation, and different hardware systems are required to realize different complex protocol signals, so that the method lacks enough flexibility and generalization capability.
Among the modulation of various complex protocol signals, OFDM modulation techniques are currently most widely used, for example, the currently mainstream 4G and 5G communication protocols in mobile communication, digital Video Broadcasting (DVB), IEEE 802.11a standard in wireless local area network, hiperLAN standard, etc. all use OFDM modulation, but these protocols use different frame structures. Facing to complex protocol signals of enemy OFDM modulation, the traditional interference mode needs different hardware systems to reconstruct different protocol and parameter signals.
The signal generation model based on machine learning can realize the generation of signals with different parameters under a unified architecture, and has more flexibility in hardware. As an emerging generative model, the generation of a countermeasure network (GENERATIVE ADVERSARIAL Networks, GAN) enables learning of high-dimensional, complex, real data distributions, which has wide application in the field of data generation. The GAN is generally composed of a neural network of two parts of a generator and a arbiter, and by the antagonistic training of the generator and the arbiter, the distribution and the statistical characteristics of the real sample data are autonomously learned, and under appropriate conditions, nash equilibrium can be reached or approximated, and at this time, the generator of the GAN can generate data similar to the real sample. GAN finds application in data generation and signal processing in many fields, such as speech signal enhancement, music sequence generation, etc., and particularly in the field of computer vision, for example, image-to-image conversion, image super resolution, video generation, text generation, etc. For a set of pictures or data with a good spatial correlation, it is feasible to generate data resembling real samples with a GAN network.
By utilizing the strong data generation capability of the GAN, the GAN network can be used as a unified architecture to generate OFDM signals with different parameters. However, the OFDM symbol sequence has noise-like characteristics when observed from the time domain, and has poor correlation, which is unfavorable for feature extraction and data distribution fitting. For the generated model, the distributed characteristics of the OFDM time domain signals are difficult to learn, if the OFDM time domain signals are used as training samples for training, the generated signal constellation is easy to generate mapping conditions which are not in accordance with requirements, the time-frequency domain parameters are not in accordance with the requirements of protocol frame structures, and the like, and the OFDM time domain signal generation method has the characteristics of being in accordance with the OFDM signals. It is necessary to observe the characteristics of the OFDM signal from more dimensions, such as the time-frequency domain, so that the GAN generates a complex protocol signal of the OFDM modulation that meets the requirements.
Disclosure of Invention
In order to solve the problems, the invention provides an OFDM signal generation method based on a generation countermeasure network, by using the method, two dimensional characteristics of an OFDM signal time domain and a frequency domain can be extracted, and an OFDM signal meeting a specific protocol (comprising a subcarrier structure and a modulation mode) is generated.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
an OFDM signal generation method based on generating an countermeasure network, comprising the steps of:
a. preprocessing an input OFDM time domain signal, converting the OFDM signal into a frequency domain, normalizing the obtained frequency domain signal, splicing the time sequence of each frequency domain OFDM symbol to obtain a time-frequency two-dimensional matrix, and storing the real part and the imaginary part of the matrix in a single-channel gray level diagram form respectively to obtain a time-frequency two-dimensional pattern for generating training and verification of an countermeasure network;
b. inputting the time-frequency two-dimensional pattern obtained in the step a into a double-discriminant GAN for training, and storing a generator model after training is converged;
c. Inputting specific random noise into a trained generator model to obtain a real part or an imaginary part time-frequency two-dimensional pattern of an OFDM frequency domain signal meeting a corresponding protocol;
d. And c, adding the real part and the imaginary part of the time-frequency two-dimensional pattern obtained in the step c, and performing Inverse Fast Fourier Transform (IFFT) and parallel-serial conversion to obtain the generated OFDM time domain signal.
Further, the preprocessing in the step a comprises two steps of serial-parallel conversion and cyclic prefix removal fast Fourier transform.
Further, the normalization processing manner in the step a is as follows:
rnorm＝r./2max(abs(r))+0.5+0.5j
wherein r represents a matrix obtained by performing serial-parallel conversion, cyclic prefix removal and FFT on each OFDM symbol on an OFDM time domain signal, and splicing the obtained vectors according to rows; dividing each row of the matrix by the elements of the corresponding row in the column vector, wherein max (·) represents the column vector formed by the maximum value of each row in the matrix, abs (·) represents the amplitude of each element in the matrix, and j is an imaginary unit.
Further, the dual arbiter GAN described in step b includes one generator and two arbiters.
Further, the input of the generator in the dual-arbiter GAN is random noise, the output is a time-frequency two-dimensional pattern, and the network structure of the generator comprises 9 layers, specifically as follows:
the layer 1 is an input layer, and the dimension of the input random noise is 2 multiplied by 1;
Layers 2 to 6 and 8 to 9 are two-dimensional deconvolution layers, zero padding is 0×0, convolution kernel sizes are 2×2, 5×2, 1×2 and 1×2, step sizes are 1×1,2×2, 5×2, 1×2 and 1×2, except that layer 9 is subjected to batch normalization by using BatchNorm function and LeakyReLU function as an activation function, and layer 9 is subjected to Tanh function as an activation function;
Layer 7 is a two-dimensional convolution layer, zero padding is 1×1, convolution kernel size is 3×3, step size is 1×1, batch normalization is performed by BatchNorm functions, and LeakyReLU functions are used as activation functions;
The output time-frequency two-dimensional pattern dimension is 1×100×64.
Further, the input of the first discriminator in the dual-discriminator GAN is a real time-frequency two-dimensional pattern and the time-frequency two-dimensional pattern generated by the generator, the output is a 1-dimensional real number, and the network structure of the first discriminator comprises 7 layers, specifically as follows:
The 1 st layer is an input layer, and the dimension of the input time-frequency two-dimensional pattern is 1 multiplied by 100 multiplied by 64;
The 2 nd to 7 th layers are two-dimensional convolution layers, zero padding is respectively 1×0, 1×1,1×0, 1×1 and 0×0, convolution kernel sizes are respectively 4×1, 4×4, 4×1, 4×4 and 2×2, step sizes are respectively 2×1, 5×4, 4×4, 2×1, 2×2 and 1×1, the steps except the 7 th layer are all subjected to batch normalization by using BatchNorm functions and take LeakyReLU functions as activation functions, the 7 th layer directly outputs a number 1 or 0 as a discrimination result without using the activation functions, and 1 and 0 represent true and false;
The output dimension is 1X 1.
Further, the input of the second discriminator in the dual-discriminator GAN is a constellation density map generated by the real time-frequency two-dimensional pattern through the pattern-constellation conversion network, and a constellation density map generated by the time-frequency two-dimensional pattern generated by the generator through the pattern-constellation conversion network, and the output is a 100-dimensional real number, and the network structure of the second discriminator comprises 6 layers, specifically as follows:
layer 1 is the input layer, and the dimension of the input constellation density map is 1×100×100;
Layers 2 to 6 are two-dimensional convolution layers, zero padding is respectively 1×0,1×0 and 0×0, convolution kernel sizes are respectively 4×1, 4×1 and 2×1, step sizes are respectively 2×1, 5×1,2×1 and 1×1, except layer 6, batch normalization is carried out by using BatchNorm functions and LeakyReLU functions are used as activation functions, layer 6 does not use an activation function, numbers 1 or 0 are directly output as a discrimination result, and 1 and 0 represent true and false respectively;
The output dimension is 1×1×100.
Further, the pattern-constellation conversion network converts the time-frequency two-dimensional pattern into a constellation density map by using a neural network, and the network structure comprises 9 layers, specifically as follows:
The 1 st layer is an input layer, and the dimension of the input time-frequency two-dimensional pattern is 1 multiplied by 100 multiplied by 64;
The 2 nd to 9 th layers are two-dimensional convolution layers, the output dimensions are 64×100×32, 128×100×16, 256×100×8, 512×100×2, 256×100×4, 128×100×8, 64×100×20, 1×100×100, except the 9 th layer, all the layers are subjected to batch normalization by using BatchNorm functions and used as an activation function by using a ReLU function, and the 9 th layer uses a Tanh function as an activation function;
Training of the pattern-constellation conversion network is performed offline, an optimizer uses an Adam optimizer, the learning rate is 5×10 -6, the training input is a time-frequency two-dimensional pattern, the label is a constellation density map corresponding to the time-frequency two-dimensional pattern, and the objective function during training is MSELoss; the constellation density map as the training set label is generated as follows:
equally dividing [0,1] into 100 parts, and summing the probability density of the amplitude of each row of pixel points in the time-frequency two-dimensional pattern in the 100 intervals;
Normalizing the probability density distribution obtained by statistics of each row, and respectively storing the probability density distribution into vectors with dimensions of 1 multiplied by 100;
And splicing the obtained vectors according to rows to obtain a matrix with the dimension of 100 multiplied by 100, and storing the matrix as a single-channel gray level map, namely a constellation density map serving as a training label.
Further, the training in step b uses Adam optimizer, the learning rate of the generator is 4×10 -4, the learning rate of the first discriminator is 1×10 -4, the learning rate of the second discriminator is 1×10 -4, and the training objective function is as follows:
Wherein V (D 1)、V(D2), V (G) are the loss functions of the first discriminator, the second discriminator and the generator respectively, G is the generator, D 1 is the first discriminator, D 2 is the second discriminator, C is the pattern-constellation conversion network, z is the random noise of the input generator, x is the true real-time frequency two-dimensional pattern as the training set, E represents expectations, x-p data (x) represents the probability distribution that x obeys the real data set, z-p z (z) represents the random distribution of the input noise, and lambda 1 and lambda 2 are the weights occupied by the first discriminator and the second discriminator in the generator objective function respectively.
The beneficial effects of the invention are as follows:
when training the GAN network, the invention designs the double discriminants to restrict the generated pattern from the time-frequency domain structure and the symbol constellation distribution respectively, so that the generated pattern can generate OFDM signals of the subcarrier structure and the modulation mode meeting the requirements of a specific protocol.
Drawings
Fig. 1 is a schematic diagram of an OFDM signal generation method based on generation of an countermeasure network in an embodiment of the present invention.
Fig. 2 is a time-frequency two-dimensional pattern corresponding to the real part of the WIFI 802.11a signal.
Fig. 3 is a network structure diagram of the dual arbiter GAN according to an embodiment of the present invention.
Fig. 4 is a schematic diagram of a conversion flow for converting an OFDM real part time-frequency two-dimensional pattern with a subcarrier number of 64 and using QPSK modulation into a constellation density map with a resolution of 100.
Fig. 5 is a diagram illustrating an OFDM signal time-frequency two-dimensional pattern with an FFT point number of 64 according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of a generator network architecture.
Fig. 7 is a schematic diagram of the network configuration of the arbiter 1 and the arbiter 2.
Fig. 8 shows a time-frequency two-dimensional pattern (the left side is the real part, and the right side is the imaginary part) of QPSK and a signal-to-noise ratio of 20dB, respectively, as modulation schemes generated by the generator.
Fig. 9 is a real constellation density map for a generated time-frequency two-dimensional pattern and a real time-frequency two-dimensional pattern (left 1 is the real part of the real pattern, left 2 is the real part of the pattern, right 2 is the real part of the generated pattern, and right 1 is the imaginary part of the generated pattern).
Fig. 10 is a comparison graph of a data segment OFDM signal time domain waveform generated based on an OFDM signal generation model that generates an countermeasure network and a data segment OFDM real signal time domain waveform of the WIFI 802.11a standard.
Detailed Description
The present invention will be described in further detail with reference to the embodiments and the accompanying drawings, in order to make the objects, technical methods and advantages of the present invention more apparent.
An OFDM signal generation method based on generation countermeasure network, its flow is as shown in figure 1, through carrying on FFT preconditioning after removing CP to the OFDM time domain symbol which accords with a certain specific agreement, get the time-frequency two-dimensional pattern of OFDM signal, input the pattern as true sample into GAN network of the multiple discriminant to train, after generator and discriminant reach Nash equilibrium, carry on IFFT to the time-frequency two-dimensional pattern that the generator generates, can get the time domain signal which accords with the agreement too. OFDM signals with different parameters are input into the system for training, and the obtained GAN network generator can realize OFDM complex protocol signal output with different parameters.
The method comprises the following steps:
a. According to fig. 1, a received OFDM time domain signal r is first preprocessed. Firstly, converting r in series and parallel, and removing CP (cyclic prefix), then obtaining a matrix r' with dimension of N T×NFFT, wherein N T is the number of OFDM symbols, and N FFT is the number of FFT points. Each row of the matrix represents a time domain sequence of one OFDM symbol. FFT is carried out on each row in the matrix r', and the time-frequency two-dimensional matrix is obtained by transforming signals from time domain to frequency domain After the normalization processing shown in the formula (1), a two-dimensional matrix r norm is obtained.
Where,/represents the division of each row of the matrix by the elements of the corresponding row in the column vector, max (·) represents the column vector consisting of the maximum value of each row in the matrix, abs (·) represents the amplitude of each element in the matrix, j is the imaginary unit. The range of values of the real part and the imaginary part of each element of the normalized matrix r norm is 0, 1.
Extracting real part and imaginary part of each element of matrix r norm and respectively forming new time-frequency two-dimensional matrixAnd/>The matrix is stored in the form of a single-channel gray image and is used as a double-discriminant GAN training set proposed by us.
Taking an OFDM signal of a data portion in the WIFI 802.11a protocol as an example, a real part time-frequency two-dimensional pattern obtained by preprocessing the signal is shown in fig. 2. In this protocol, the FFT point number is 64, so the pattern has 64 columns, representing 64 frequency domain subcarriers, respectively, and 100 rows of the pattern represent 100 OFDM symbols, respectively. 48 data subcarriers are arranged in the 64 subcarriers, random QPSK symbols are carried on each data subcarrier, and pixel points which are staggered in black and white are arranged in the real part time-frequency two-dimensional pattern; the 4 sub-carriers carrying pilot frequencies are white; the remaining 12 columns are blank subcarriers and are gray in the pattern.
B. the time-frequency two-dimensional pattern obtained in the step a is input into a double-discriminator GAN shown in fig. 3 for training, and one discriminator 2 is added in the double-discriminator GAN except for the discriminator 1 in the conventional GAN. The purpose of the discriminator is to restrict the constellation distribution of the modulation symbols in the generated pattern and judge whether the modulation constellation distribution of the generated pattern is identical to the real pattern. If the two are the same, the judgment is true, and if the two are different, the judgment is false. This embodiment uses LSGAN as the base network. And respectively adopting 0 and 1 to represent labels of false samples and true samples, and enabling a decision boundary for judging whether the samples are true or false to be 1 for enabling the generator to generate the samples as true as possible. The objective function of the GAN of the dual arbiter is defined as follows:
Wherein, G is a generator, D 1 is a discriminator 1, D 2 is a discriminator 2, C is a pattern constellation conversion network, z is random noise of an input generator, x is a real-time frequency two-dimensional pattern as a training set, E represents expectations, x-p data (x) represents probability distribution of x obeying a real dataset, z-p z (z) represents that the input noise meets the random distribution, and lambda 1 and lambda 2 are weights occupied by the discriminator 1 and the discriminator 2 in a generator objective function respectively.
In order to determine the true or false of the distribution of the modulation constellation of the data symbols in the time-frequency two-dimensional pattern, the input of the arbiter 2 is the constellation density map of the true pattern and the generated pattern, respectively. The present embodiment proposes a pattern-constellation conversion network that is trained offline, and its function is to implement the conversion of a time-frequency two-dimensional pattern into a constellation density map, and the manner of converting the time-frequency two-dimensional pattern into the constellation density map is shown in fig. 4. Taking an OFDM signal carrying data in the WIFI 802.11a protocol as an example, the size of the input time-frequency two-dimensional pattern is 100×64, which represents a time-frequency two-dimensional real part pattern corresponding to a signal with a length of 100 OFDM symbol periods. First, 0,1 is equally divided into 100 parts, and the probability density of the amplitude of each row of pixels in the time-frequency two-dimensional pattern in these 100 intervals is counted. And normalizing the probability density distribution obtained by statistics of each row, and respectively storing the probability density distribution into vectors with dimensions of 1 multiplied by 100. And finally, splicing the obtained vectors according to rows to obtain a matrix with the dimension of 100 multiplied by 100, wherein the matrix is the constellation density map.
This embodiment uses a convolution-like self-encoder to fit the quantization statistical procedure described above. And extracting constellation distribution characteristics of the time-frequency two-dimensional pattern data points through operations such as convolution and deconvolution, and generating a constellation density map. The pattern-constellation transition network structure is shown in fig. 5.
The generator is composed of deconvolution and convolution neural network modules, and the specific structure is shown in fig. 6. Under the action of a pattern structure and a constellation data dual-discriminant, the input random noise is converted into a time-frequency two-dimensional pattern which approximates to a real sample. The discriminator is composed of a convolutional network module, and the specific structure is shown in fig. 7. The function is to distinguish the true sample of the input and the generated sample as correctly as possible.
In order to verify the effectiveness of the OFDM signal generation model based on the discriminator in the embodiment, a data segment in a WIFI 802.11a PLCP protocol data unit (PCLP Protocol Date Unit, PPDU) is selected as a real signal to perform a signal generation experiment. The data section is composed of OFDM symbols, the number of subcarriers is 64, wherein the number of subcarriers carrying the data symbols is 48, and the subcarriers are respectively positioned in the 2 nd to 7 th, 9 th to 21 th, 23 th to 27 th, 39 th to 43 th, 45 th to 57 th and 59 th to 64 th subcarriers; pilot occupies 4 sub-carriers and is distributed in 8 th, 22 th, 44 th and 58 th sub-carriers; the remaining 12 are null subcarriers. The signal bandwidth is 20MHz, each OFDM symbol duration is 4us, the CP length is 0.8us, and at a sampling rate of 20MHz, one OFDM symbol contains 80 samples, where the CP contains 16 samples. Tables 1 and 2 show simulation parameters and hyper parameters selected for model training, respectively.
Table 1 simulation parameters
| Parameters (parameters) | Value of |
| Number of data subcarriers | 48 |
| Number of pilot subcarriers | 4 |
| N FFT: total subcarrier number/FFT point number | 64 |
| B: bandwidth of a communication device | 20MHz |
| Modulation scheme | QPSK |
| SNR: channel signal to noise ratio | 20dB |
| Preprocessing the obtained time-frequency two-dimensional graph size | 100*64 |
| Sample data set size per snr and modulation scheme | 10000 |
Table 2 double discriminator LSGAN training hyper-parameters
| Parameters (parameters) | Value of |
| Batch_size | 40 |
| Generator learning rate | 4e-4 |
| Discriminant 1 learning rate | 1e-4 |
| Discriminant 2 learning rate | 1e-4 |
| Optimizer | Adam |
| The discriminator 1 loss function weight lambda 1 | 0.1 |
| Discriminator 2 loss function weight λ2 | 0.4 |
C. And inputting random noise into a trained generator model to obtain a real part or an imaginary part time-frequency two-dimensional pattern of an OFDM frequency domain signal meeting the WIFI 802.11a PLCP protocol as shown in figure 8, wherein a constellation density diagram corresponding to the real and generated time-frequency two-dimensional pattern is shown in figure 9.
D. and c, adding the real part and the imaginary part of the time-frequency two-dimensional pattern obtained in the step c, performing IFFT and parallel-serial conversion to obtain the OFDM time domain signal which meets the WIFI 802.11a PLCP protocol as shown in figure 10.
In short, the method aims at the problem of difficult extraction of the time domain characteristics of the OFDM signals, firstly, the Fast Fourier Transform (FFT) is used for preprocessing the time domain OFDM signals, and the processed frequency domain symbol vectors are spliced into a two-dimensional data matrix; the data matrix is then stored as a time-frequency two-dimensional pattern in the form of a gray image for training and testing of the GAN. In addition, the invention designs the double-discriminant GAN, and uses one generator and two discriminants to play games at the same time, wherein the purpose of the generator is to generate a time-frequency two-dimensional pattern to cheat the two discriminants, and the two discriminants respectively distinguish the generated pattern and the real pattern from two aspects of subcarrier structures and constellation densities of modulation symbols so as to ensure that the generated signals meet the requirements of OFDM time-frequency domain characteristics. The invention can realize the generation of the complex protocol OFDM signal.
Claims (1)
1. An OFDM signal generation method based on generation of an countermeasure network, comprising the steps of:
a. Preprocessing an input OFDM time domain signal, converting the OFDM signal into a frequency domain, normalizing the obtained frequency domain signal, splicing the time sequence of each frequency domain OFDM symbol to obtain a time-frequency two-dimensional matrix, and storing the real part and the imaginary part of the matrix in a single-channel gray level diagram form respectively to obtain a time-frequency two-dimensional pattern for generating training and verification of an countermeasure network; the pretreatment comprises two steps of serial-parallel conversion and cyclic prefix removal fast Fourier transform;
the normalization processing mode is as follows:
Wherein, Representing a matrix obtained by performing serial-parallel conversion on an OFDM time domain signal, removing a cyclic prefix, performing FFT on each OFDM symbol, and splicing the obtained vectors according to rows; /(I)Dividing elements representing each row of the matrix by corresponding rows in the column vector,/>Column vector representing the maximum composition of each row in the matrix,/>Representing the magnitude of each element in the matrix,Is an imaginary unit;
b. Inputting the time-frequency two-dimensional pattern obtained in the step a into a double-discriminant GAN for training, and storing a generator model after training is converged; the dual arbiter GAN comprises a generator and two discriminators; the input of the generator in the dual-discriminator GAN is random noise, the output is a time-frequency two-dimensional pattern, and the network structure of the generator comprises 9 layers, specifically as follows:
Layer 1 is the input layer, the dimension of the input random noise is ；
The 2 th to 6 th layers and the 8 th to 9 th layers are two-dimensional deconvolution layers, and zero filling is thatConvolution kernel sizes are/>, respectively、/>、、/>、/>、/>And/>Step sizes are/>, respectively、/>、/>、/>、/>、/>And/>Batch normalization is carried out by BatchNorm functions except the layer 9, leakyReLU functions are used as activating functions, and the layer 9 uses Tanh functions as activating functions;
Layer 7 is a two-dimensional convolution layer with zero padding of Convolution kernel size is/>Step size is/>Batch normalization was performed with BatchNorm function and LeakyReLU function was used as the activation function;
The dimension of the output time-frequency two-dimensional pattern is ；
The input of the first discriminator in the dual-discriminator GAN is a real time-frequency two-dimensional pattern and the time-frequency two-dimensional pattern generated by the generator respectively, the output is a 1-dimensional real number, and the network structure of the first discriminator comprises 7 layers, specifically as follows:
layer 1 is the input layer, the input time-frequency two-dimensional pattern dimension is ；
The 2 nd to 7 th layers are two-dimensional convolution layers, and zero padding is respectively as follows、/>、/>、/>、/>And/>Convolution kernel sizes are/>, respectively、/>、/>、/>、/>And/>Step sizes are/>, respectively、/>、/>、/>、/>And/>Except the 7 th layer, carrying out batch normalization by using BatchNorm functions and using LeakyReLU functions as activating functions, directly outputting numbers 1 or 0 as a discrimination result without using the activating functions by the 7 th layer, wherein 1 and 0 respectively represent true and false;
The output dimension is ；
The input of the second discriminator in the dual-discriminator GAN is a constellation density map generated by the real time-frequency two-dimensional pattern through a pattern-constellation conversion network, and the constellation density map generated by the time-frequency two-dimensional pattern generated by the generator through the pattern-constellation conversion network is output as a 100-dimensional real number, and the network structure of the second discriminator comprises 6 layers, specifically as follows:
layer 1 is the input layer, the dimension of the input constellation density map is ；
The 2 nd layer to the 6 th layer are two-dimensional convolution layers, and zero filling is respectively as follows、/>、/>、/>And/>Convolution kernel sizes are/>, respectively、/>、/>、/>And/>Step sizes are/>, respectively、/>、/>、/>And/>Except the 6 th layer, carrying out batch normalization by using BatchNorm functions and using LeakyReLU functions as activating functions, directly outputting numbers 1 or 0 as a discrimination result without using the activating functions by the 6 th layer, wherein 1 and 0 respectively represent true and false;
The output dimension is ；
The pattern-constellation conversion network converts a time-frequency two-dimensional pattern into a constellation density map by using a neural network, and the network structure comprises 9 layers, and the method comprises the following steps of:
layer 1 is the input layer, the input time-frequency two-dimensional pattern dimension is ；
The 2 nd to 9 th layers are two-dimensional convolution layers, and the output dimensions are respectively、/>、/>、/>、、/>、/>、/>Except the layer 9, carrying out batch normalization by using BatchNorm functions and using a ReLU function as an activation function, wherein the layer 9 uses a Tanh function as the activation function;
training of the pattern-constellation conversion network is performed offline, and the optimizer uses an Adam optimizer with a learning rate of The input of training is a time-frequency two-dimensional pattern, the label is a constellation density diagram corresponding to the time-frequency two-dimensional pattern, and the objective function during training is MSELoss; the constellation density map as the training set label is generated as follows:
Will be Equally dividing the time-frequency two-dimensional pattern into 100 parts, and summing the probability density of the amplitude of each row of pixel points in the time-frequency two-dimensional pattern in the 100 intervals;
normalizing the probability density distribution obtained by counting each row, and respectively storing the probability density distribution into the dimension Is defined in the vector of (2);
the obtained vectors are spliced according to rows to obtain dimension as The matrix is stored as a single-channel gray level diagram, namely a constellation density diagram serving as a training label;
training using Adam optimizer, the learning rate of the generator is The learning rate of the first discriminator is/>The learning rate of the second discriminant is/>The objective function of training is as follows:
Wherein, 、/>、/>Is the loss function of the first arbiter, the second arbiter and the generator, respectively,/>Generator,/>For the first arbiter,/>For the second discriminator,/>For pattern-constellation conversion network,/>Random noise for input generator,/>For a true real-time frequency two-dimensional pattern as training set,/>Representing expectations,/>Representation/>Obeying the probability distribution of the real dataset,/>Representing that the noise of the input satisfies a random distribution,/>And/>The weights of the first discriminator and the second discriminator in the generator objective function are respectively;
c. Inputting specific random noise into a trained generator model to obtain a real part or an imaginary part time-frequency two-dimensional pattern of an OFDM frequency domain signal meeting a corresponding protocol;
d. and c, adding the real part and the imaginary part of the time-frequency two-dimensional pattern obtained in the step c, and performing Inverse Fast Fourier Transform (IFFT) and parallel-serial conversion to obtain the generated OFDM time domain signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310083.9A CN114826857B (en) | 2022-03-28 | 2022-03-28 | OFDM signal generation method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310083.9A CN114826857B (en) | 2022-03-28 | 2022-03-28 | OFDM signal generation method based on generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114826857A CN114826857A (en) | 2022-07-29 |
| CN114826857B true CN114826857B (en) | 2024-05-03 |
Family
ID=82531444
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210310083.9A Active CN114826857B (en) | 2022-03-28 | 2022-03-28 | OFDM signal generation method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114826857B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115801510B (en) * | 2022-11-10 | 2025-01-07 | 北京邮电大学 | Communication system and method |
| CN115905819B (en) * | 2023-03-09 | 2023-05-12 | 中国民用航空飞行学院 | rPPG signal generation method and device based on generation countermeasure network |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111178260A (en) * | 2019-12-30 | 2020-05-19 | 山东大学 | Modulation signal time-frequency diagram classification system based on generation countermeasure network and operation method thereof |
| CN112804048A (en) * | 2021-04-12 | 2021-05-14 | 南京信息工程大学 | Physical layer chaotic encryption optical transmission method based on generation countermeasure network |
| CN114065841A (en) * | 2021-10-26 | 2022-02-18 | 中国电子科技集团公司第五十四研究所 | Channel characteristic migration method based on star-type generation countermeasure network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11422217B2 (en) * | 2019-06-26 | 2022-08-23 | Siemens Healthcare Gmbh | Progressive generative adversarial network in medical image reconstruction |
-
2022
- 2022-03-28 CN CN202210310083.9A patent/CN114826857B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111178260A (en) * | 2019-12-30 | 2020-05-19 | 山东大学 | Modulation signal time-frequency diagram classification system based on generation countermeasure network and operation method thereof |
| CN112804048A (en) * | 2021-04-12 | 2021-05-14 | 南京信息工程大学 | Physical layer chaotic encryption optical transmission method based on generation countermeasure network |
| CN114065841A (en) * | 2021-10-26 | 2022-02-18 | 中国电子科技集团公司第五十四研究所 | Channel characteristic migration method based on star-type generation countermeasure network |
Non-Patent Citations (1)
| Title |
|---|
| 基于生成对抗网络的数字信号生成技术研究;杨鸿杰, 陈丽, 张君毅;电子测量技术;20201031;第43卷(第20期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114826857A (en) | 2022-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Dileep et al. | Dense layer dropout based CNN architecture for automatic modulation classification | |
| CN114826857B (en) | OFDM signal generation method based on generation countermeasure network | |
| Yang et al. | Deep learning aided method for automatic modulation recognition | |
| CN108540419B (en) | A deep learning-based OFDM detection method against inter-subcarrier interference | |
| Wang et al. | Signal detection in uplink time-varying OFDM systems using RNN with bidirectional LSTM | |
| Park et al. | Deep learning-based automatic modulation classification with blind OFDM parameter estimation | |
| Jeong et al. | Spectrogram-based automatic modulation recognition using convolutional neural network | |
| US11368349B1 (en) | Convolutional neural networks based computationally efficient method for equalization in FBMC-OQAM system | |
| Cheng et al. | A ResNet-DNN based channel estimation and equalization scheme in FBMC/OQAM systems | |
| Zhang et al. | Label-assisted transmission for short packet communications: A machine learning approach | |
| CN112242969A (en) | Novel single-bit OFDM receiver based on model-driven deep learning | |
| Ma et al. | CNN-based automatic modulation recognition of wireless signal | |
| Delamou et al. | Deep learning-based estimation for multitarget radar detection | |
| Ouyang et al. | Channel estimation for underwater acoustic OFDM communications: An image super-resolution approach | |
| An et al. | Blind high-order modulation recognition for beyond 5G OSTBC-OFDM systems via projected constellation vector learning network | |
| Kumar et al. | Automatic modulation classification for adaptive OFDM systems using convolutional neural networks with residual learning | |
| He et al. | Deep learning-based automatic modulation recognition algorithm in non-cooperative communication systems | |
| CN112104389A (en) | Staged BOC signal detection method based on cyclostationarity | |
| Huynh-The et al. | Efficient convolutional networks for robust automatic modulation classification in OFDM-based wireless systems | |
| Zhang et al. | An effective generative model based channel estimation method with reduced overhead | |
| Hao et al. | Frequency domain analysis and convolutional neural network based modulation signal classification method in OFDM system | |
| An et al. | Blind multicarrier waveform recognition based on spatial-temporal learning neural networks | |
| CN102025459A (en) | Nonparametric estimation ICA-based MIMO-OFDM system blind deconvolution method | |
| Yıldırım et al. | Deep receiver design for multi-carrier waveforms using cnns | |
| Zhang et al. | Deep learning‐based digital signal modulation identification under different multipath channels |
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 |