CN109274625B - Information modulation mode determining method and device, electronic equipment and storage medium - Google Patents
Information modulation mode determining method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN109274625B CN109274625B CN201811337509.XA CN201811337509A CN109274625B CN 109274625 B CN109274625 B CN 109274625B CN 201811337509 A CN201811337509 A CN 201811337509A CN 109274625 B CN109274625 B CN 109274625B
- Authority
- CN
- China
- Prior art keywords
- information
- gcm
- demodulated
- cfcn
- modulation mode
- 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
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
Abstract
The embodiment of the invention provides a method and a device for determining an information modulation mode, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring information to be demodulated; converting the information to be demodulated into a grid constellation matrix GCM; and inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN. The invention realizes better determination of the information modulation mode corresponding to the information to be demodulated and reduces the computational complexity.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a method and an apparatus for determining an information modulation scheme, an electronic device, and a storage medium.
Background
The determination of the information modulation mode corresponding to the information to be demodulated is an intermediate step between information detection and demodulation, and is a key technology for identifying the modulation mode of the target information damaged by noise and interference. The information modulation mode determining method is widely applied to the military and civil fields.
The information modulation mode determination method includes a likelihood-based method. In the likelihood-based method, the modulation scheme classification problem is regarded as a hypothesis testing problem, likelihood function values between received information to be demodulated and candidate modulation schemes are calculated, and the modulation scheme corresponding to the maximum value among the candidate modulation schemes is determined as the modulation scheme corresponding to the received information. The method assumes that the probability density function of the information to be received is known, and gives an optimal information modulation mode determination scheme in the Bayesian sense by minimizing the misclassification probability. However, in practical situations, a large number of unknown parameters need to be estimated, and the method often fails to reach a theoretically optimal information modulation mode corresponding to the information to be demodulated, and also suffers from considerable computational complexity.
Therefore, how to obtain a better information modulation mode corresponding to the information to be demodulated and reduce the computational complexity is still an urgent technical problem to be solved in the information modulation mode determination method.
Disclosure of Invention
The embodiment of the invention aims to provide an information modulation mode determining method, an information modulation mode determining device, electronic equipment and a storage medium, so as to achieve better determination of an information modulation mode corresponding to information to be demodulated and reduction of calculation complexity. The specific technical scheme is as follows:
in a first aspect, an embodiment of the present invention discloses a method for determining an information modulation mode, where the method includes:
acquiring information to be demodulated;
converting the information to be demodulated into a grid constellation matrix GCM;
and inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN.
Optionally, the converting the information to be demodulated into a grid constellation matrix GCM includes:
converting the information to be demodulated into a grid constellation matrix GCM through a preset formula;
wherein the preset formula is as follows:
wherein I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents parameters for adjusting the dimensionality and sparsity of the GCMCounting; k represents the number of rows of the GCM; l represents the number of columns of the GCM; each element in the GCM is a ratio value of the number of information to the total number of information at a corresponding position in the GCM grid.
Optionally, the step of determining the target-contrast full convolutional network CFCN includes:
and inputting the information of the multiple groups of calibration modulation modes into an initial contrast full convolution network CFCN for training until a contrast loss function of the initial contrast full convolution network CFCN is converged to obtain the target contrast full convolution network CFCN.
Optionally, the contrast loss function is represented as follows:
J(W)＝Js(W)+Jr(W)+Jc(W0,W1)
wherein J (W) represents the value of the contrast loss function corresponding to W, which represents all the weights of the CFCN; j. the design is a squares(W) represents cross entropy loss; j. the design is a squarer(W) represents the L2 regularization term; j. the design is a squarec(W0,W1) All weights denoted W for the CFCN0All weights with the CFCN are W1Loss of contrast.
Alternatively, the Jc(W0,W1) Is represented as follows:
wherein 1 {. denotes an index function; (.)+＝max{·,0}；XaIndicates belonging to the y-thaα represents the difference threshold of the characteristics between the adjustment classes;
representing the euclidean distance between a pair of eigenvectors in a 128-dimensional feature space.
Optionally, the inputting the GCM into a predetermined target-to-full convolutional network CFCN, and determining a modulation mode of the to-be-demodulated information corresponding to the GCM through the target CFCN includes:
inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining probability values of the information to be demodulated corresponding to the GCM corresponding to preset modulation modes;
and determining a preset modulation mode corresponding to the maximum value in the probability values as the modulation mode of the information to be demodulated corresponding to the GCM.
In a second aspect, an embodiment of the present invention discloses an information modulation method determining apparatus, where the apparatus includes:
the demodulation information acquisition module is used for acquiring demodulation information;
the to-be-demodulated information conversion module is used for converting the to-be-demodulated information into a grid constellation matrix GCM;
and the modulation mode determining module is used for inputting the GCM into a predetermined target contrast full convolution network CFCN and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN.
Optionally, the to-be-demodulated information conversion module is specifically configured to convert the to-be-demodulated information into a grid constellation matrix GCM through a preset formula;
wherein the preset formula is as follows:
wherein I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents a parameter for adjusting the dimensionality and sparsity of the GCM; k represents the number of rows of the GCM; l represents the number of columns of the GCM; each element in the GCM is a ratio value of the number of information to the total number of information at a corresponding position in the GCM grid.
Optionally, the apparatus further includes a target CFCN determining module, configured to input multiple sets of information of the calibration modulation modes into an initial contrast full convolution network CFCN for training until a contrast loss function of the initial contrast full convolution network CFCN converges, so as to obtain the target contrast full convolution network CFCN.
Optionally, the modulation scheme determining module includes:
a probability value determining submodule, configured to input the GCM into a predetermined target contrast full convolution network CFCN, and determine probability values of the to-be-demodulated information corresponding to the GCM, where the probability values correspond to preset modulation modes;
and the modulation mode determining submodule is used for determining a preset modulation mode corresponding to the maximum value in the probability values as the modulation mode of the information to be demodulated corresponding to the GCM.
In a third aspect, an embodiment of the present invention discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete mutual communication through the communication bus;
the memory is used for storing a computer program;
the processor is configured to implement the method steps of any one of the above information modulation scheme determining methods when executing the program stored in the memory.
In another aspect, an embodiment of the present invention discloses a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the method steps of any one of the above information modulation mode determining methods are implemented.
In another aspect, an embodiment of the present invention discloses a computer program product containing instructions, which when run on a computer, causes the computer to perform the method steps of any one of the above-mentioned information modulation scheme determination methods.
The embodiment of the invention provides a method, a device, electronic equipment and a storage medium for determining an information modulation mode, which particularly convert acquired information to be demodulated into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, which needs to estimate a large number of unknown parameters and has large calculation complexity, the method and the device can more efficiently extract effective information and eliminate redundant information based on the grid constellation matrix GCM, and can effectively reduce the calculation complexity by automatically identifying the modulation mode of the information to be demodulated through the full convolution network CFCN. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a flowchart of a method for determining an information modulation scheme according to an embodiment of the present invention;
fig. 2(a) is a grayscale image of the information modulation method according to the embodiment of the present invention, when the signal-to-noise ratio is 5dB and the modulation method is binary phase shift keying BPSK, the information to be demodulated is converted into GCM;
fig. 2(b) is a gray image obtained by converting information to be demodulated into GCM when the modulation scheme with the signal-to-noise ratio of 10dB is QPSK according to the information modulation scheme determining method in the embodiment of the present invention;
fig. 2(c) is a gray image obtained by converting information to be demodulated into GCM when the snr is 15dB and the modulation scheme is 16QAM according to the embodiment of the present invention;
fig. 3 is a schematic structural diagram of a target CFCN in an information modulation method determining method according to an embodiment of the present invention;
fig. 4 is a comparison graph of the performance impact of the GCM grid size on the information modulation mode determination method in the information modulation mode determination method according to the embodiment of the present invention;
fig. 5 is a graph comparing the effect of symbol length on the performance of an information modulation method determination method under different signal-to-noise ratios in an information modulation method determination method according to an embodiment of the present invention;
FIG. 6 is a performance comparison graph of different information modulation mode determination methods;
FIG. 7(a) is a performance comparison graph of different information modulation mode determination methods when phase offset is an argument;
fig. 7(b) is a performance comparison diagram of different information modulation mode determination methods when the normalized frequency shift is an argument;
fig. 8 is a schematic structural diagram of an information modulation scheme determining apparatus according to an embodiment of the present invention;
fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The information modulation mode determining method also comprises a characteristic-based method, wherein the characteristic-based method is to extract a series of characteristics from the received information to be demodulated, judge the modulation type of the information to be demodulated according to the characteristics, has lower calculation complexity and strong robustness on model uncertainty, and can realize the optimal determination of the information modulation mode corresponding to the information to be demodulated in Bayesian sense. Therefore, the present invention focuses on feature-based methods and aims to improve the performance of feature-based methods.
Recent studies show that deep learning based algorithms can learn discriminant representations from complex data, and can achieve better classification performance than traditional feature-based methods. Therefore, an information modulation mode determination method based on a fully-connected neural network is proposed, which uses a plurality of high-order cumulants as input, but has a long training time. In addition, a stacked convolution automatic encoder is adopted to extract characteristics from the cumulant of signals in a multipath fading channel, and then the modulation type of the information to be demodulated is judged according to the characteristics. However, the computational complexity of calculating the accumulation and training the autoencoder is still high. In addition, the statistical features such as the accumulated amount cannot completely acquire the information in the received information to be demodulated, which may limit the representation learning capability of the deep learning algorithm, resulting in poor effect of determining the information modulation mode corresponding to the information to be demodulated.
In order to better determine the information modulation mode corresponding to the information to be demodulated, the prior art is further improved to obtain a neural network so as to extract more identifying information representation from the received information to be demodulated. For example, the modulation scheme is identified by extracting features from I/Q (In-phase/Quadrature) symbols of the received information to be demodulated. This method lacks data preprocessing and the network structure is simple. In addition, the method is sensitive to frequency offset and the like. And the automatic constellation map modulation mode classification method based on the deep belief network is also adopted, and because the network does not have convolution operation, the representation with identification capability cannot be extracted from the constellation map. In addition, the training process introduces significant computational overhead. Therefore, the existing information modulation mode determining method using the neural network still limits the effect of determining the information modulation mode corresponding to the information to be demodulated.
In the existing information modulation mode determining method based on characteristics, the characteristics of certain signals are usually calculated from received signals, no processing or some simple processing is performed on the characteristics, and the characteristics are directly input into a classifier for modulation mode classification and identification. However, in an actual working environment, especially under the conditions of low signal-to-noise ratio and few sampling points, due to the influence of noise and interference, a satisfactory determination result cannot be obtained by directly using some characteristics with a theoretically good classification effect. In addition, the existing modulation mode classification method based on the characteristics by adopting the deep learning method has the following problems: the calculation complexity is high, the representation learning capability of deep learning is limited by the limitation of statistical characteristics, and the accuracy and robustness of modulation mode classification are limited by unreasonable network structure.
In order to solve the above problem, embodiments of the present invention provide an information modulation method determining method, an apparatus, an electronic device, and a storage medium, which implement determining an information modulation method corresponding to information to be demodulated according to characteristics of a GCM (Grid Constellation Matrix) and a CFCN (continuous functional Network, in contrast to a full volume Network). The method comprises the steps of firstly, acquiring GCM with variable size from received information to be demodulated as input of CFCN, wherein the GCM can reserve effective information in a received signal and eliminate information redundancy; then, the CFCN is used for learning high-dimensional representation of different modulation modes from the GCM so as to determine the information modulation mode corresponding to the information to be demodulated. The specific process is as follows:
in a first aspect, an embodiment of the present invention discloses a method for determining an information modulation scheme, as shown in fig. 1. Fig. 1 is a flowchart of a method for determining an information modulation scheme according to an embodiment of the present invention, where the method includes:
s101, obtaining information to be demodulated.
In this step, the receiver may acquire the information to be demodulated from one signal source by using a single antenna method.
The received information to be demodulated r (n) can be represented as:
wherein h represents a channel coefficient subject to rayleigh distribution, which is invariant in determining an information modulation scheme; n denotes the total number of symbols in the received information to be demodulated, f0Represents a frequency offset; theta0Represents a phase offset; x is the number ofs(n) represents the constellation H from the unit mean powersN generated in(complex) symbols; w (n) represents additive white Gaussian noise with a mean of zero and a variance of
Since the constellation has unity power, the received signal-to-noise ratio can be defined as
S102, the information to be demodulated is converted into a grid constellation matrix GCM.
In order to facilitate the comparison of the full convolutional network CFCN to learn an effective representation from the information to be demodulated, the received information to be demodulated is converted into the GCM, and the GCM is used as the input of the CFCN.
Optionally, the converting the information to be demodulated into the grid constellation matrix GCM in S102 includes:
converting information to be demodulated into a grid constellation matrix GCM through a preset formula;
wherein, the preset formula is as follows:
wherein, I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents parameters for adjusting the dimensionality and sparsity of the GCM; k represents the width of the GCM, L represents the length of the GCM, and each element in the GCM is a ratio value of the number of information to the total number of information at a corresponding position in a GCM grid.
Fig. 2(a) is a grayscale image obtained by converting information to be demodulated into GCM when the signal-to-noise ratio is 5dB and the modulation scheme is BPSK (Binary Phase Shift Keying); the GCM has a size of 22 × 44 elements.
Fig. 2(b) is a grayscale image of GCM converted from information to be demodulated when the signal-to-noise ratio is 10dB and the modulation scheme is QPSK (Quadrature Phase Shift Keying); wherein the GCM has a size of 32 × 33 elements.
Fig. 2(c) shows a gray image obtained by converting information to be demodulated into GCM when the signal-to-noise ratio is 15dB and the Modulation scheme is 16QAM (Quadrature Amplitude Modulation) in the information Modulation scheme determining method according to the embodiment of the present invention; the GCM has a size of 31 × 31 elements.
As can be seen from fig. 2(a), 2(b), and 2(c), the size of the GCM displayed varies with different modulation schemes under specific settings, because the size of the GCM is determined by the amplitude range of the information symbol sequence of the modulation scheme to be determined, which can avoid introducing null information and further reduce the computational complexity of the subsequent classifier training. Furthermore, the highlighted pixels in the image indicate that the received information to be demodulated is gathered in these areas. Thus, the GCM can be considered as a sample estimate of a two-dimensional probability distribution of the constellation symbols, which contains more constellation information than statistical features like cumulative quantities etc.
S103, inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN.
The target CFCN of the present embodiment aims to learn the high-dimensional representation of each modulation scheme included in the input GCM and provide strong classification performance. Its working principle is to represent learning and convolutional neural networks.
In S102, after the information to be demodulated is converted into the GCM, the GCM is input into a predetermined target CFCN, and the characteristics of the GCM are extracted through the target CFCN, thereby determining the modulation mode of the information to be demodulated corresponding to the GCM.
Optionally, the step of inputting the GCM into a predetermined target contrast full convolution network CFCN in S103, and determining a modulation mode of information to be demodulated corresponding to the GCM through the target CFCN includes:
step one, inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining probability values of information to be demodulated corresponding to the GCM corresponding to preset modulation modes.
In the embodiment of the invention, a common information modulation mode can be set in the target contrast full convolution network CFCN in advance. For example, in this embodiment, the set S of preset modulation schemes is { BPSK, QPSK, 8PSK (Phase shift keying), 16QAM, 64QAM }. The GCM is input into the target-contrast full convolution network CFCN, through which a one-dimensional vector is output. Wherein, the first data value of the one-dimensional vector is the probability that the modulation mode of the information to be demodulated is BPSK; the second data value of the one-dimensional vector is the probability that the modulation mode of the information to be demodulated is QPSK, and the third data value of the one-dimensional vector is the probability that the modulation mode of the information to be demodulated is 8 PSK; the fourth data value of the one-dimensional vector is the probability that the modulation mode of the information to be demodulated is 16 QAM; the fifth data value of the one-dimensional vector is the probability that the modulation mode of the information to be demodulated is 64 QAM.
And step two, determining a preset modulation mode corresponding to the maximum value in the probability values as a modulation mode of the information to be demodulated corresponding to the GCM.
In the method for determining the information modulation mode, the acquired information to be demodulated is converted into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, the method needs to estimate a large number of unknown parameters and has high calculation complexity, the grid constellation matrix GCM-based method can more efficiently extract effective information and eliminate redundant information, and the modulation mode of the information to be demodulated is identified by the CFCN (full convolution network) so as to effectively reduce the calculation complexity. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
Optionally, in an embodiment of the information modulation method according to the present invention, the step of determining the target-to-full convolutional network CFCN includes:
and inputting the information of the multiple groups of calibration modulation modes into the initial comparison full convolution network CFCN for training until the comparison loss function of the initial comparison full convolution network CFCN is converged to obtain the target comparison full convolution network CFCN.
In the embodiment of the invention, an initial comparison full convolution network CFCN is preset and a comparison loss function is set for the initial CFCN. And inputting the information of the calibrated modulation modes into the initial comparison full convolution network CFCN aiming at a plurality of groups of information of the calibrated modulation modes, extracting the characteristics of the information of the calibrated modulation modes through the initial comparison full convolution network, further outputting the value of a comparison loss function corresponding to the information of the calibrated modulation modes through the initial comparison full convolution network, and feeding back and adjusting the parameter values in the initial comparison full convolution network until the comparison loss function of the initial comparison full convolution network CFCN is converged to finally obtain the target comparison full convolution network CFCN of the embodiment of the invention.
Alternatively, the contrast loss function is expressed as follows:
J(W)＝Js(W)+Jr(W)+Jc(W0,W1)
wherein j (W) represents the value of the contrast loss function corresponding to W, which represents all the weights of CFCN; j. the design is a squares(W) represents cross entropy loss; j. the design is a squarer(W) represents the L2 regularization term; j. the design is a squarec(W0,W1) All weights denoted CFCN are W0All weights with CFCN are W1Loss of contrast.
Alternatively, Jc(W0,W1) Is represented as follows:
wherein 1 {. denotes an index function; (.)+＝max{·,0}；XaIndicates belonging to the y-thaα represents the difference threshold of the characteristics between the adjustment classes;
representing the euclidean distance between a pair of eigenvectors in a 128-dimensional feature space.
Wherein, the Euclidean distance
Can be expressed as follows:
wherein | · | purple sweet2Representing the euclidean norm; g () represents a function representing a layer module; vaThe modulation mode of the output of the fusion module is represented as yaThe 128-dimensional vector of (1); vbThe modulation mode of the output of the fusion module is represented as yb128-dimensional vector of (1).
For the other case, minimizing the contrast loss means enlarging the distance between the feature vectors of different modulation schemes, and the cost function causes the distance to exceed the threshold α.
The initial GCM is trained to minimize the value of the loss function j (w), and the derivative of the loss function j (w) is given by:
wherein the content of the first and second substances,
on this basis, the parameters in the initial CFCN are updated using a small batch stochastic gradient descent and back propagation algorithm until the loss function j (w) converges to a constant.
As shown in fig. 3, fig. 3 is a schematic structural diagram of a target CFCN in an information modulation method determining method according to an embodiment of the present invention. The target CFCN of the embodiment of the invention consists of three modules, namely a representation module, a fusion module and a classification module. The representation module is intended to extract the underlying reduced-dimension representation from the GCM. These representations are then transformed into fixed-dimension feature vectors by a fusion module. The modulation scheme is then determined using a classification module.
To compromise accuracy and efficiency, the representation module consists of three convolution blocks, each of which consists of four convolution layers. The first three layers adopt 3 multiplied by 3 small convolution kernels to extract the representation information. The continuous convolutional layer can improve the non-linearity of the network and limit its size, which helps to enhance representation learning ability and prevent overfitting. Where both the kernel padding and step size are set to 1 to ensure that the input GCM size is not reduced by convolution. The fourth layer of the convolution block uses a 2 x 2 convolution kernel without padding and the step size is set to 2 in order to reduce the dimensionality of the input data. It should be noted that conventional CNNs use this layer as a pooling layer, and down-sampling during pooling may confuse the information in the GCM. In addition, the number of channels of these convolution blocks is gradually increased, which facilitates extraction of information from information to be demodulated input from low-level channels and synthesis of effective information.
By convolutional layer, GCM is converted into tensor M of corresponding modulation modeiIt can be expressed as follows:
wherein, XiIs the ith input GCM, W0All weights representing CFCN; k represents the number of rows of the GCM; l represents the number of columns of GCM; r represents a real number set.
The fusion module is mainly used for uniformly inputting tensors for the subsequent classification modules. It consists of two layers, namely a convolutional layer and a fusogenic layer. In convolutional layers, the step sizes of both the kernel and the padding are set to 1, and the tensor is converted to
The fused layer is formed by reducing M'iOutputs a 128-dimensional feature vector ViAnd V isiThe c-th element in (b) mayExpressed as:
wherein, wk,lIs the weight of the fusion layer. k represents a line number index; l represents a column number index; c represents ViThe element index of (2).
The reduced dimension may lose the presentation information and so the number of channels in the convolutional layer is set to 128 to retain sufficient presentation information. By means of the representation module and the fusion module, the input GCM is converted into a feature vector in a 128-dimensional feature space, which can be represented by Vi＝f(Mi|W1)＝f(g(Xi|W0)|W1)∈R128Given therein, W1Representing the weight of the fusion module.
The last part of the classification module consists of two fully connected layers. In the first layer, the rejection rate is set to 50% to avoid overfitting and enhance robustness to noise. The last layer normalizes the output of each neuron using softmax as an activation function, indicating the probability that the information to be demodulated belongs to the corresponding modulation mode. The activation functions of the other layers employ modified linear elements to introduce non-linearity and sparsity.
Therefore, the target CFCN for determining the modulation mode of the information to be demodulated corresponding to the input GCM can be obtained through the embodiment of the invention. The target CFCN is a neural network for determining information modulation modes based on GCM, and can efficiently learn high-dimensional representations of different modulation modes from a grid constellation matrix converted from received information to be demodulated and realize the determination of the modulation modes. In addition, the contrast loss function of the embodiment of the invention enhances the difference between different modulation modes. The GCM can automatically adjust the size, effectively extracts useful information from the received information to be demodulated and eliminates information redundancy, and compared with other modulation mode classification methods based on deep learning in the near term, the method has lower computational complexity.
In addition, in the embodiment of the present invention, the index P can be evaluatedccEvaluation of target comparison full volume for embodiments of the inventionThe performance of the product network CFCN. PccIs a standard function for evaluating the predictive performance of different models, the preset evaluation index PccIs represented as follows:
wherein S represents a preset modulation mode set; p (H)s) Represents a modulation scheme HsA priori probability of (a);
indicating that the modulation scheme is correctly estimated as HsThe probability of (c).
For example, the embodiment of the present invention presets a modulation scheme set as S ═ BPSK, QPSK, 8PSK }, and if there are 50000 symbol sequence samples of known modulation schemes, the symbol sequence samples include 20000 BPSK information sequence samples, 20000 QPSK information sequence samples, and 10000 8PSK information sequence samples. The 50000 information samples are input into the target CFCN model of the embodiment of the present invention, and 19500 samples with BPSK sequence samples correctly predicted as BPSK modulation scheme are obtained, 19000 samples with QPSK sequence samples correctly predicted as QPSK, and 9000 samples with 8PSK sequence samples correctly predicted as 8 PSK. Then pass through PccAnd calculating a formula to obtain the performance of the target CFCN model of the embodiment of the invention.
PccIn the range of 0-1, PccThe larger the value, the better the performance of the model.
Fig. 4 is a comparison diagram of the performance impact of the GCM grid size on the information modulation method determination method in the information modulation method determination method according to the embodiment of the present invention.
In fig. 4, the abscissa represents the signal-to-noise ratio (SNR), and the ordinate represents the probability of correct evaluation (P)cc. The symbol length of the received information to be demodulated is set to 4096, and each preset modulation mode set S is consideredAll preset modulation modes. The results show that the performance of the target CFCN fluctuates with the mesh size at different signal-to-noise ratios. The target CFCN performs best at low signal-to-noise ratios with a mesh size of 0.2, but performs the worst at high signal-to-noise ratios. This can be explained in two ways, at low signal-to-noise ratio, GCM with small grids is sensitive to noise due to too dense constellation scattering; when the signal-to-noise ratio is high, the GCM with a small grid can more accurately represent the information to be demodulated, while a large grid is sparse and provides insufficient representation information. On this basis, the embodiment of the present invention may set the grid size to 0.1.
Compared with other modulation mode classification methods based on deep learning in the near term, the modulation mode classification method based on deep learning has a better modulation mode classification recognition rate, and particularly has a higher performance gain under the condition of low signal-to-noise ratio.
Fig. 5 is a comparison diagram of performance impact of symbol lengths under different signal-to-noise ratios on an information modulation method determination method in an information modulation method determination method according to an embodiment of the present invention.
Wherein the parameter settings are the same as in fig. 4. As can be seen from the figure, the more information symbols to be demodulated are received, the better the target CFCN performance is, and it is further illustrated that the GCM with more symbols represents the modulation mode more accurately.
Fig. 6 is a performance comparison diagram of different information modulation mode determination methods, In which performances of a target CFCN and a DBN (Deep Belief Network), a SCAE (Stacked Convolutional Auto-Encoder), an IQ-CNN (In-phase orthogonal Neural Network) and an FCN (full Convolutional Network) feature-based method are compared, where the FCN uses the same Network structure as the target CFCN but does not use the proposed contrast loss. The symbol lengths are all set to 4096.
As can be seen from fig. 6, the target CFCN is superior to the other four methods. Specifically, CFCN is at 95% P compared to FCNccGains in excess of 2dB were achieved, demonstrating gain for contrast loss. Furthermore, the target CFCN is in95% P on DBN, SCAE and IQ-CNN, respectivelyccGains of 0.5dB, 1dB and 2.5dB are produced. This can be explained in two ways, first, CFCN can extract more discriminative features from GCM due to its fully convolved structure and specially designed loss function compared to DBN and IQ-CNN; second, other methods use statistical features as inputs, such as cumulants in the SCAE, which limit the representation learning ability of deep neural networks and lead to poor performance. In addition, under the condition of using the same data set (each modulation mode 50000GCM) and device (GTX1080GPU (Graphics Processing Unit)), the training time of the target CFCN, DBN and SCAE is 12164 seconds, 13978 seconds and 24418 seconds respectively, which shows that the target CFCN of the embodiment of the present invention has lower computational complexity.
Fig. 7(a) is a performance comparison diagram of different information modulation scheme determination methods when the phase offset is an argument. Fig. 7(a) and 7(b) are both target CFCN, DBN and SCAE, performance comparisons at flat fading channel, symbol length 4096, signal-to-noise ratio of 9 dB.
In fig. 7(a), the feasible region of the target CFCN (where P is observed)ccAbove 99%) is much wider than DBN and it is symmetrically centered at 0dB, fig. 7(a) also shows that the target CFCN works best when θ is 0 ° and the performance is substantially consistent when the phase offset increases. In addition, SCAE is slightly better than CFCN, because the cumulant has strong stability to constellation rotation.
Fig. 7(b) is a performance comparison diagram of different information modulation scheme determination methods when the normalized frequency shift is an argument. In fig. 7(b), the frequency offset is normalized by the sampling frequency and varies from 0 to 0.0002. It can be observed that the frequency offset seriously degrades the performance of each information modulation mode determination method, and the target CFCN performs best, which proves the robustness of the target CFCN to the frequency offset of the embodiment of the present invention.
Therefore, the information modulation mode determining method provided by the embodiment of the invention has the advantages of strong robustness and good phase offset resistance and frequency offset resistance. The information modulation mode determining method can extract the more effective and more identifying high-level representation of the modulation mode characteristics, so that the information modulation mode determining method has better tolerance to phase offset and frequency offset and better robustness.
In a second aspect, an embodiment of the present invention discloses an information modulation mode determining apparatus, as shown in fig. 8. Fig. 8 is a schematic structural diagram of an information modulation scheme determining apparatus according to an embodiment of the present invention, where the apparatus includes:
a to-be-demodulated information obtaining module 801, configured to obtain to-be-demodulated information;
a to-be-demodulated information conversion module 802, configured to convert the to-be-demodulated information into a grid constellation matrix GCM;
and a modulation mode determining module 803, configured to input the GCM into a predetermined target-to-full convolution network CFCN, and determine a modulation mode of information to be demodulated corresponding to the GCM through the target CFCN.
In the apparatus for determining an information modulation mode according to the embodiment of the present invention, acquired information to be demodulated is converted into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, the method needs to estimate a large number of unknown parameters and has high calculation complexity, the grid constellation matrix GCM-based method can more efficiently extract effective information and eliminate redundant information, and the modulation mode of the information to be demodulated is identified by the CFCN (full convolution network) so as to effectively reduce the calculation complexity. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
Optionally, in an embodiment of the apparatus for determining an information modulation scheme according to the present invention, the module 802 for converting information to be demodulated is specifically configured to,
converting information to be demodulated into a grid constellation matrix GCM through a preset formula;
wherein, the preset formula is as follows:
wherein, I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents parameters for adjusting the dimensionality and sparsity of the GCM; k represents the width of the GCM, L represents the length of the GCM, and each element in the GCM is a ratio value of the number of information to the total number of information at a corresponding position in a GCM grid.
Optionally, in an embodiment of the information modulation mode determining apparatus of the present invention, the apparatus further includes a target CFCN determining module, configured to input multiple sets of information of the calibrated modulation modes into the initial comparison full convolution network CFCN for training until a comparison loss function of the initial comparison full convolution network CFCN converges, so as to obtain the target comparison full convolution network CFCN.
Optionally, in an embodiment of the information modulation mode determining apparatus of the present invention, the modulation mode determining module 803 includes:
the probability value determining submodule is used for inputting the GCM into a predetermined target comparison full convolution network CFCN and determining each probability value of the information to be demodulated corresponding to the GCM corresponding to each preset modulation mode;
and the modulation mode determining submodule is used for determining the preset modulation mode corresponding to the maximum value in the probability values as the modulation mode of the information to be demodulated corresponding to the GCM.
Optionally, in an embodiment of the information modulation mode determining apparatus of the present invention, the method further includes:
the correct evaluation probability determining module is used for determining the correct evaluation probability of the probability value aiming at the target CFCN according to each probability value through a preset evaluation index; wherein, the preset evaluation index PccIs represented as follows:
wherein S represents a preset modulation mode set; p (H)s) Represents a modulation scheme HsA priori probability of (a);
indicating that the modulation scheme is correctly estimated as HsThe probability of (c).
In a third aspect, an embodiment of the invention discloses an electronic device, as shown in fig. 9. Fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, which includes a processor 901, a communication interface 902, a memory 903, and a communication bus 904, where the processor 901, the communication interface 902, and the memory 903 complete communication with each other through the communication bus 904;
a memory 903 for storing computer programs;
the processor 901 is configured to implement the following method steps when executing the program stored in the memory:
acquiring information to be demodulated;
converting the information to be demodulated into a grid constellation matrix GCM;
and inputting the GCM into a predetermined target contrast full convolution network CFCN, and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN.
The communication bus 904 mentioned in the above electronic device may be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The communication bus 904 may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not mean that there is only one bus or one type of bus.
The communication interface 902 is used for communication between the electronic apparatus and other apparatuses.
The Memory 903 may include a Random Access Memory (RAM) or a Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory 903 may also be at least one storage device located remotely from the processor 901.
The Processor 901 may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processing (DSP), Application Specific Integrated Circuit (ASIC), Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component.
In an electronic device provided by an embodiment of the present invention, acquired information to be demodulated is converted into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, the method needs to estimate a large number of unknown parameters and has high calculation complexity, the grid constellation matrix GCM-based method can more efficiently extract effective information and eliminate redundant information, and the modulation mode of the information to be demodulated is identified by the CFCN (full convolution network) so as to effectively reduce the calculation complexity. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
In another aspect, an embodiment of the present invention discloses a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the method steps of any one of the above information modulation mode determining methods are implemented.
In a computer-readable storage medium provided in an embodiment of the present invention, acquired information to be demodulated is converted into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, the method needs to estimate a large number of unknown parameters and has high calculation complexity, the grid constellation matrix GCM-based method can more efficiently extract effective information and eliminate redundant information, and the modulation mode of the information to be demodulated is identified by the CFCN (full convolution network) so as to effectively reduce the calculation complexity. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
In another aspect, an embodiment of the present invention discloses a computer program product containing instructions, which when run on a computer, causes the computer to perform the method steps of any one of the above-mentioned information modulation scheme determination methods.
In a computer program product including instructions provided by an embodiment of the present invention, the obtained information to be demodulated is converted into a grid constellation matrix GCM; and then inputting the GCM into a predetermined target contrast full convolution network CFCN, so that the modulation mode of the information to be demodulated corresponding to the GCM is determined through the target CFCN. Compared with the existing likelihood-based method, the method needs to estimate a large number of unknown parameters and has high calculation complexity, the grid constellation matrix GCM-based method can more efficiently extract effective information and eliminate redundant information, and the modulation mode of the information to be demodulated is identified by the CFCN (full convolution network) so as to effectively reduce the calculation complexity. In addition, compared with the existing likelihood-based method, the method obtains a better information modulation mode for determining the information to be demodulated through the grid constellation matrix GCM and the characteristics of comparing the full convolution network CFCN.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The procedures or functions according to the embodiments of the invention are brought about in whole or in part when the computer program instructions are loaded and executed on a computer. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wirelessly (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., Solid State Disk (SSD)), among others.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
All the embodiments in the present specification are described in a related manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the device, the electronic apparatus and the storage medium embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and the relevant points can be referred to the partial description of the method embodiments.
The above description is only for the preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention shall fall within the protection scope of the present invention.
Claims (4)
1. An information modulation mode determining method, the method comprising:
acquiring information to be demodulated;
converting the information to be demodulated into a grid constellation matrix GCM, wherein the information to be demodulated is converted into the grid constellation matrix GCM through a preset formula;
wherein the preset formula is as follows:
wherein I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents a parameter for adjusting the dimensionality and sparsity of the GCM; k represents the number of rows of the GCM; l represents the number of columns of the GCM; each element in the GCM is a proportional value of the number of information symbols and the total number of the information symbols at the corresponding position in the GCM grid;
inputting the GCM into a target contrast full convolution network CFCN determined by pre-training, and determining the modulation mode of the information to be demodulated corresponding to the GCM through the target CFCN determined by pre-training, wherein the CFCN is a full convolution neural network model based on a contrast loss function, and the probability value of the information to be demodulated corresponding to the GCM corresponding to each preset modulation mode is determined by inputting the GCM into the target contrast full convolution network CFCN determined by pre-training; determining a preset modulation mode corresponding to the maximum value in the probability values as a modulation mode of the to-be-demodulated information corresponding to the GCM; the CFCN model is trained by inputting a plurality of groups of information of calibration modulation modes into an initial contrast full convolution network CFCN in pairs, and obtaining a target contrast full convolution network CFCN determined by the pre-training through repeated iterative training until a contrast loss function of the initial contrast full convolution network CFCN is converged; the contrast loss function is expressed as follows:
J(W)＝Js(W)+Jr(W)+Jc(W0,W1)
wherein J (W) represents the value of the contrast loss function corresponding to W, which represents all the weights of the CFCN; j. the design is a squares(W) represents cross entropy loss; j. the design is a squarer(W) represents the L2 regularization term; j. the design is a squarec(W0,W1) All weights denoted W for the CFCN0All weights with the CFCN are W1Loss of contrast of (d);
said Jc(W0,W1) Is represented as follows:
wherein 1 {. denotes an index function; (.)+＝max{·,0}；XaIndicates belonging to the y-thaα represents the difference threshold of the characteristics between the adjustment classes;
representing the euclidean distance between a pair of eigenvectors in a 128-dimensional eigenspace; wherein the Euclidean distance
Can be expressed as follows:
wherein | · | purple sweet2Representing the euclidean norm; g () represents a function representing a layer module; vaThe modulation mode of the output of the fusion module is represented as yaThe 128-dimensional vector of (1); vbThe modulation mode of the output of the fusion module is represented as yb128-dimensional vector of (1).
2. An information modulation scheme determining apparatus, comprising:
the demodulation information acquisition module is used for acquiring demodulation information;
the to-be-demodulated information conversion module is used for converting the to-be-demodulated information into a grid constellation matrix GCM; the to-be-demodulated information conversion module is specifically configured to convert the to-be-demodulated information into a grid constellation matrix GCM through a preset formula;
wherein the preset formula is as follows:
wherein I represents the real part of the information to be demodulated; q represents the imaginary part of the information to be demodulated; max (-) represents the maximum element of the output vector corresponding to the information to be demodulated; min (-) represents the minimum element of the output vector corresponding to the information to be demodulated;
represents an upward rounding function; g represents a parameter for adjusting the dimensionality and sparsity of the GCM; k represents the number of rows of the GCM; l represents the number of columns of the GCM; each element in the GCM is a proportional value of the number of information symbols and the total number of the information symbols at the corresponding position in the GCM grid;
a modulation mode determining module, configured to input the GCM into a target contrast full convolution network CFCN determined by pre-training, and determine a modulation mode of the to-be-demodulated information corresponding to the GCM through the target CFCN, where the CFCN is a full convolution neural network model based on a contrast loss function, and determine probability values of the to-be-demodulated information corresponding to the GCM corresponding to each preset modulation mode by inputting the GCM into the target contrast full convolution network CFCN determined by pre-training; determining a preset modulation mode corresponding to the maximum value in the probability values as a modulation mode of the to-be-demodulated information corresponding to the GCM; the CFCN model is trained by inputting a plurality of groups of information of calibration modulation modes into an initial contrast full convolution network CFCN in pairs, and obtaining a target contrast full convolution network CFCN determined by pre-training through repeated iterative training until a contrast loss function of the initial contrast full convolution network CFCN is converged, wherein the contrast loss function is expressed as follows:
J(W)＝Js(W)+Jr(W)+Jc(W0,W1)
wherein J (W) represents the value of the contrast loss function corresponding to W, which represents all the weights of the CFCN; j. the design is a squares(W) represents cross entropy loss; j. the design is a squarer(W) represents the L2 regularization term; j. the design is a squarec(W0,W1) All weights denoted W for the CFCN0All weights with the CFCN are W1Loss of contrast of (d);
said Jc(W0,W1) Is represented as follows:
wherein 1 {. denotes an index function; (.)+＝max{·,0}；XaIndicates belonging to the y-thaα represents the difference threshold of the characteristics between the adjustment classes;
representing a Euclidean distance between a pair of feature vectors in a 128-dimensional feature space, wherein the Euclidean distance
Can be expressed as follows:
wherein | · | purple sweet2Representing the euclidean norm; g () represents a function representing a layer module; vaThe modulation mode of the output of the fusion module is represented as yaThe 128-dimensional vector of (1); vbThe modulation mode of the output of the fusion module is represented as yb128-dimensional vector of (1).
3. An electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete communication with each other through the communication bus;
the memory is used for storing a computer program;
the processor, when executing the program stored in the memory, performs the method steps of claim 1.
4. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the method steps of claim 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811337509.XA CN109274625B (en) | 2018-11-12 | 2018-11-12 | Information modulation mode determining method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811337509.XA CN109274625B (en) | 2018-11-12 | 2018-11-12 | Information modulation mode determining method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109274625A CN109274625A (en) | 2019-01-25 |
| CN109274625B true CN109274625B (en) | 2020-06-19 |
Family
ID=65192925
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811337509.XA Active CN109274625B (en) | 2018-11-12 | 2018-11-12 | Information modulation mode determining method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109274625B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059757B (en) * | 2019-04-23 | 2021-04-09 | 北京邮电大学 | Mixed signal classification method and device and electronic equipment |
| CN110309854A (en) * | 2019-05-21 | 2019-10-08 | 北京邮电大学 | A kind of signal modulation mode recognition methods and device |
| CN110798417B (en) | 2019-10-24 | 2020-07-31 | 北京邮电大学 | Signal modulation identification method and device based on cyclic residual error network |
| CN111343115B (en) * | 2020-02-19 | 2021-06-29 | 北京邮电大学 | 5G communication modulation signal identification method and system |
| CN111327559B (en) * | 2020-02-28 | 2021-01-08 | 北京邮电大学 | Encoding and decoding method and device |
| CN112364768B (en) * | 2020-11-11 | 2024-03-22 | 浙江工业大学 | Distributed optical fiber intrusion recognition method based on airspace characteristics and machine learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9729362B1 (en) * | 2013-03-20 | 2017-08-08 | Georgia Tech Research Corporation | Systems and methods for autonomous signal modulation format identification |
| CN107276938A (en) * | 2017-06-28 | 2017-10-20 | 北京邮电大学 | A kind of digital signal modulation mode recognition methods and device |
| 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 |
| CN108780519A (en) * | 2016-03-11 | 2018-11-09 | 奇跃公司 | Structure learning in convolutional neural networks |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8755469B1 (en) * | 2008-04-15 | 2014-06-17 | The United States Of America, As Represented By The Secretary Of The Army | Method of spectrum mapping and exploitation using distributed sensors |
| US8611410B2 (en) * | 2010-07-30 | 2013-12-17 | National Instruments Corporation | Variable modulus mechanism for performing equalization without a priori knowledge of modulation type or constellation order |
| CN108229509B (en) * | 2016-12-16 | 2021-02-26 | 北京市商汤科技开发有限公司 | Method and device for identifying object class and electronic equipment |
| US20180183532A1 (en) * | 2016-12-23 | 2018-06-28 | Intel Corporation | Modulation format selection for millimeter-wave operation in the presence of phase noise |
| CN106789788B (en) * | 2016-12-26 | 2019-05-10 | 北京邮电大学 | A kind of wireless digital signal Modulation Mode Recognition method and device |
| CN108229404B (en) * | 2018-01-09 | 2022-03-08 | 东南大学 | Radar echo signal target identification method based on deep learning |
-
2018
- 2018-11-12 CN CN201811337509.XA patent/CN109274625B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9729362B1 (en) * | 2013-03-20 | 2017-08-08 | Georgia Tech Research Corporation | Systems and methods for autonomous signal modulation format identification |
| CN108780519A (en) * | 2016-03-11 | 2018-11-09 | 奇跃公司 | Structure learning in convolutional neural networks |
| CN107276938A (en) * | 2017-06-28 | 2017-10-20 | 北京邮电大学 | A kind of digital signal modulation mode recognition methods and device |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度卷积神经网络的数字调制方式识别;彭超然;《计算机测量与控制》;20180825;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109274625A (en) | 2019-01-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109274625B (en) | Information modulation mode determining method and device, electronic equipment and storage medium | |
| CN110232341B (en) | Semi-supervised learning image identification method based on convolution-stacking noise reduction coding network | |
| CN113050042B (en) | Radar signal modulation type identification method based on improved UNet3+ network | |
| Gurram et al. | Support-vector-based hyperspectral anomaly detection using optimized kernel parameters | |
| Cai et al. | Signal modulation classification based on the transformer network | |
| CN110569752A (en) | convolutional neural network-based radar signal category determination method | |
| CN110738242A (en) | Bayes structure learning method and device for deep neural networks | |
| US20230185998A1 (en) | System and method for ai-assisted system design | |
| CN114567528A (en) | Communication signal modulation mode open set identification method and system based on deep learning | |
| CN112737992A (en) | Underwater sound signal modulation mode self-adaptive in-class identification method | |
| Meng et al. | An effective weighted vector median filter for impulse noise reduction based on minimizing the degree of aggregation | |
| CN116912568A (en) | Noise-containing label image recognition method based on self-adaptive class equalization | |
| CN111343115B (en) | 5G communication modulation signal identification method and system | |
| Su et al. | Deep non-cooperative spectrum sensing over Rayleigh fading channel | |
| CN112350790B (en) | Deep learning-based spectrum sensing detection method, device and equipment | |
| CN116628566A (en) | Communication signal modulation classification method based on aggregated residual transformation network | |
| CN111027609A (en) | Image data weighted classification method and system | |
| Kumar et al. | 2D-FFT based modulation classification using deep convolution neural network | |
| CN115620038A (en) | Common knowledge constrained remote sensing sample migration method | |
| Du et al. | D-GF-CNN algorithm for modulation recognition | |
| Hussein et al. | Automatic Modulation Classification: Convolutional Deep Learning Neural Networks Approaches | |
| CN113688655A (en) | Method and device for identifying interference signal, computer equipment and storage medium | |
| Li et al. | Automatic modulation classification based on the improved AlexNet | |
| Chen et al. | Optimisation for image salient object detection based on semantic‐aware clustering and CRF | |
| CN115664908B (en) | Cross-layer intelligent countermeasure method, system, medium and equipment for communication signal modulation recognition |
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 |