CN114285701B - Method, system, equipment and terminal for identifying transmitting power of main user - Google Patents
Method, system, equipment and terminal for identifying transmitting power of main user Download PDFInfo
- Publication number
- CN114285701B CN114285701B CN202111441610.1A CN202111441610A CN114285701B CN 114285701 B CN114285701 B CN 114285701B CN 202111441610 A CN202111441610 A CN 202111441610A CN 114285701 B CN114285701 B CN 114285701B
- Authority
- CN
- China
- Prior art keywords
- data
- convolutional neural
- neural network
- sampling
- network
- 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 69
- 238000005070 sampling Methods 0.000 claims abstract description 67
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 63
- 238000012549 training Methods 0.000 claims abstract description 34
- 230000001149 cognitive effect Effects 0.000 claims abstract description 24
- 238000013461 design Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000011056 performance test Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 57
- 230000006870 function Effects 0.000 claims description 42
- 230000005540 biological transmission Effects 0.000 claims description 32
- 239000011159 matrix material Substances 0.000 claims description 24
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000001228 spectrum Methods 0.000 claims description 13
- 230000008447 perception Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 5
- 238000005562 fading Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 9
- 230000008569 process Effects 0.000 abstract description 3
- 206010063385 Intellectualisation Diseases 0.000 abstract 1
- 238000004088 simulation Methods 0.000 description 9
- 230000007547 defect Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013179 statistical model Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000015556 catabolic process Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000010363 phase shift Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The invention belongs to the technical field of wireless communication, and discloses a method, a system, equipment and a terminal for identifying the transmitting power of a main user, which are used for identifying the transmitting power of the main user with single input and multiple output, wherein the number of transmitting antennas configured by the main user is 1, the number of sensing antennas configured by cognitive radio equipment is K, and K is more than 1, and the method for identifying the transmitting power of the main user comprises the following steps: sampling data; obtaining observation data; designing a convolutional neural network structure; training a network; and (5) identifying a performance test. The main user transmitting power identification method introduces a convolutional neural network to design test statistics, utilizes the strong capability of the convolutional neural network to extract high-dimensional characteristics from observed data, solves the problem that the design of the test statistics of the traditional power identification method needs to involve a large number of manual processes and relies on excessive expert knowledge to cause lower accuracy, realizes the intellectualization of power identification, and solves the problem of deteriorated identification performance of the traditional method in a non-ideal environment with low signal-to-noise ratio.
Description
Technical Field
The present invention belongs to the technical field of wireless communications, and in particular, relates to a method, a system, a device, and a terminal for identifying transmission power of a primary user.
Background
Currently, with the large-scale deployment of 5G networks, the rapid emergence of internet of things, and the rapid growth of spectrum usage demands by various emerging technologies, wireless spectrum resources will become increasingly strained. Cognitive radio has become a very promising technology, and the main idea is to solve the problem of spectrum scarcity by identifying the surrounding radio environment and dynamically multiplexing the underutilized spectrum resources. To further improve the spectrum utilization capabilities, some studies have introduced a hybrid access spectrum based strategy for secondary users, i.e. if the presence of a primary user is detected, the secondary user needs to switch to a low transmit power below the interference temperature threshold that the primary user can tolerate for signal transmission. Meanwhile, the primary user may select multiple transmit power levels to meet different quality of service requirements, as specified by many standards (e.g., IEEE802.11 series). Therefore, in order not to affect the communication of the primary user, the secondary user needs to detect not only the occupancy state of the frequency band by the primary user but also the transmit power level value when the primary user is active before accessing the spectrum. Thus, methods of spectrum sensing and power recognition of depth have attracted the interests of many scholars.
The traditional main user transmitting power identification method mainly carries out identification based on the statistic model design detection statistic. The disadvantage of these model-driven algorithms is that the recognition performance is highly dependent on the accuracy of the detection statistic model, but in practical situations the prior model is very difficult to be identical to the actual signal-to-noise model, and once the statistic model has uncertainty or is not available, the method can lead to phenomena of sharp degradation or even failure in the communication scene with low signal-to-noise ratio (SNR). In addition, conventional approaches require a significant amount of a priori knowledge, including noise power and the transmit power pattern of the primary user (i.e., a priori probability for each power level), which are unlikely to be achievable in practice. Therefore, there is an urgent need to design a better signal-to-noise ratio robust transmit power identification scheme that does not require a priori probabilities for each hypothesis.
Through the above analysis, the problems and defects existing in the prior art are as follows:
(1) The identification performance of the traditional main user transmitting power identification method is highly dependent on the accuracy of the prior statistical model, but in actual situations, the prior statistical model is difficult to be identical to an actual signal-noise model.
(2) Conventional approaches require a significant amount of a priori knowledge including noise power and the transmit power pattern of the primary user (i.e., a priori probability for each power level), which are unlikely to be achievable in practice.
The difficulty of solving the problems and the defects is as follows: due to the complex high dynamics of the radio environment, it is extremely difficult to acquire accurate and fixed a priori models in real communication environments to build test statistics for primary user transmit power identification.
Since in most spectrum access strategies the primary and secondary users belong to a non-cooperative communication mode, this means that the secondary user cannot obtain a priori probability of the primary user's transmit power.
The meaning of solving the problems and the defects is as follows: the method breaks through the model-driven thinking of the traditional identification method, designs the observation data adaptive convolutional neural network by introducing the data-driven thought, utilizes the strong capability of the network to extract the energy correlation characteristics, builds the test statistic based on the observation data according to the requirements of the identification task, does not need to manually design the test statistic based on the statistical model, and effectively overcomes the defect of performance deterioration of the traditional identification method under the low signal-to-noise ratio environment.
Disclosure of Invention
Aiming at the problems that the traditional method needs to involve a large amount of manual processes and relies on excessive professional knowledge, so that the recognition performance is deteriorated under the condition of low signal to noise ratio, the invention provides a method, a system, equipment and a terminal for recognizing the transmission power of a main user, and particularly relates to a method, a system, equipment and a terminal for recognizing the transmission power of the main user based on a convolutional neural network.
The invention is realized in such a way that a main user transmitting power identification method is used for a single-input multi-output power identification task, namely, the number of transmitting antennas configured by a main user is 1, the number of sensing antennas configured by cognitive radio equipment is K, and K is more than 1, the main user transmitting power identification method comprises the following steps:
step one, data sampling: the cognitive radio equipment acquires N time-discrete sampling vectors through K perception antennas;
step two, obtaining observation data: preprocessing the sampling signals to obtain covariance matrixes of the sampling signals, and taking the covariance matrixes as observation data of a convolutional neural network; preprocessing the original data into data types containing rich energy related features, effectively removing redundant feature information in the original data, and providing powerful support for training and judging of a convolutional neural network.
Step three, designing a convolutional neural network structure: the method comprises the steps of designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally using the output feature vector to assist the classifier to make identification judgment; aiming at a main user transmitting power identification task, a feature extraction network adapting to observation data is constructed, high-dimensional features facing task requirements are extracted, and an identification result is output according to a judgment rule.
Step four, network training: training the network constructed in the third step by adopting the observation data set obtained in the second step, updating the model parameters by implementing a gradient descent algorithm, and obtaining the target of maximizing posterior probability by minimizing a cost function to obtain a trained network; the training mode enables the trainable parameters to train according to the gradient descent direction of the set cost function, and provides optimal network parameters for application deployment (or recognition test).
Step five, identifying performance test: and inputting the observation data to be tested into a trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result. Is beneficial to verifying the identification performance of the application and is compared with the traditional representative method under the same condition.
Further, the data sampling in the first step includes:
there are N time-discrete sample vectors in one sample window, for a single sample vector:
y(n)=[y 1 (n),…,y k (n),…,y K (n)] T ,n∈[1,N];
wherein y is k (n) represents the nth time-discrete sampled data received at the kth sense antenna.
The single discrete point-in-time sampling vector expression is:
wherein s (n) represents a non-Gaussian signal emitted from a main user terminal;is a K x 1 vector representing the rayleigh fading channel vector from the transmitting end to each of the sensing antennas; w (n) represents colored complex Gaussian noise received by the sensing antenna; p (P) i I=0, 1, …, I represents the level value of the transmission power of the primary user, and there are I transmission power level values available for the primary user, the magnitude relation being P 0 <P 1 <…<P I The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is 0 =0, meaning that the primary user is unoccupied spectrum; h i Representing a specific primary user transmit power P i Is a hypothetical case of (a).
Further, the obtaining the observation data in the second step includes:
(1) Carrying out data preprocessing on the received sampling data to obtain a covariance matrix of the sampling signal, and constructing a labeled observation data set based on the covariance matrix;
(2) The real part data and the imaginary part data of the observed data are separated into two channels as input data of the convolutional neural network.
The original sampling signal is converted into a covariance matrix of the sampling signal to be used as observation data of the convolutional neural network. During a sampling window, after the cognitive radio device receives the N time discrete sampling vectors, calculating to obtain a covariance matrix expression of the sampling signals, wherein the covariance matrix expression is as follows:
the observation data set is set to:
wherein,c representing observed data y (N) a set, T being a set of data tags T; for single sample data, +.>Represents the mth, (m=1, 2, …, M) observation in the observation set; wherein->For observing data, vector t (m) =[t 0 ,...,t i ,...,t I ],i∈[0,I]Is a tag vector corresponding to observed data, where t i With a value of 0 or 1, when t i When=1, the corresponding observation data is +.>Belonging to hypothesis H i The category in the case is expressed as:
covariance matrix C y The real part and the imaginary part of (N) are separated to form dual-channel data D in (i; j; α), α representing the channel number; wherein channel 1 is real part data D in (i;j;1)=(Real(C y (N))) i,j Channel 2 is the imaginary data D in (i;j;2)=(Imag(C y (N))) i,j (. Cndot.) i, j represents the i row, j column element, so data D is entered in Is a two-channel matrix of data size K x 2.
Further, the designing the convolutional neural network structure in the third step includes:
constructing a covariance matrix-adapted convolutional neural network, enabling an input signal to pass through a feature extractor module, then pass through a full-connection layer of a classifier module to output a scoring result, and finally normalizing the result by using a Softmax function to output a posterior probability, wherein the method comprises the following steps of:
(1) Extracting high-dimensional feature information of the observed data by constructing a feature extractor of the suitability of the observed data;
(2) The output feature vector is used for assisting the classifier to make identification judgment, and finally, the Softmax function outputs posterior probability values under various hypotheses.
Further, the network training in the step four includes:
training the network constructed in the third step by adopting the training set obtained in the second step,
(1) Designing a cost function based on a likelihood function;
(2) Reverse transmission is carried out according to a chained rule by utilizing a gradient descent algorithm based on a cost function, and trainable parameters in a network are continuously updated;
(3) And obtaining a target of maximizing posterior probability by minimizing a cost function, thereby obtaining a trained network.
Wherein, the cost function is:
setting training target as minimum cost function to obtain optimal parameter theta * So that posterior probabilityMaximum.
Based on the cost function, gradually updating parameters of DNN through a back propagation algorithm, and obtaining the trained optimal network output as follows:
wherein,representing a set of posterior probabilities output on a trained optimal network, wherein +.>Represented in the optimal network parameter set θ * The observation data is->When assuming H i The case is true posterior probability.
Further, the identifying performance test in the fifth step includes:
(1) Data preprocessing: calculating covariance matrix of the sampling signal to be identified;
(2) Input data generation: the real part and the imaginary part are separated into two channels and input into a trained deep convolutional neural network;
(3) Calculating an output vector of the deep convolutional neural network by using a forward propagation algorithm, and outputting the maximum component in the vector; wherein the method comprises the steps ofEach component in the output vector of the deep convolutional neural network corresponds to a level value of the transmission power of the main user, and based on a maximum posterior probability criterion, the hypothesis corresponding to the maximum component in the output vector is identified as the transmission power level value of the main user, namely for a hypothesis testing pair,if there is->Then for the mth sample, H i The hypothetical case of (2) will be judged to be correctly classified.
Another object of the present invention is to provide a primary user transmit power identification system applying the primary user transmit power identification method, the primary user transmit power identification system comprising:
the data sampling module is used for acquiring N time-discrete sampling vectors through K perception antennas by using the cognitive radio equipment;
the observation data acquisition module is used for preprocessing the sampling signals to obtain covariance matrixes of the sampling signals and taking the covariance matrixes as observation data of the convolutional neural network;
the convolutional neural network structure design module is used for designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally outputting feature vectors to assist the classifier in making identification decisions;
the network training module is used for training the network constructed by the convolutional neural network structure design module by adopting the observation data set obtained by the observation data obtaining module, updating the model parameters by implementing a gradient descent algorithm, and obtaining the target of maximizing the posterior probability by minimizing the cost function to obtain a trained network;
the recognition performance test module is used for inputting the observation data to be tested into the trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
It is a further object of the present invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:
the method comprises the steps that a main user configures the number of transmitting antennas to be 1, the number of sensing antennas configured by cognitive radio equipment is K, and K is more than 1, and the cognitive radio equipment acquires N time-discrete sampling vectors through K sensing antennas; preprocessing the sampling signals to obtain covariance matrixes of the sampling signals, and taking the covariance matrixes as observation data of a convolutional neural network;
the method comprises the steps of designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally using the output feature vector to assist the classifier to make identification judgment;
training the constructed network by adopting an observation data set, updating model parameters by adopting a gradient descent algorithm, and obtaining a target of maximizing posterior probability by minimizing a cost function to obtain a trained network; inputting the observation data to be tested into a trained network, obtaining the output of the optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
Another object of the present invention is to provide a computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
the method comprises the steps that a main user configures the number of transmitting antennas to be 1, the number of sensing antennas configured by cognitive radio equipment is K, and K is more than 1, and the cognitive radio equipment acquires N time-discrete sampling vectors through K sensing antennas; preprocessing the sampling signals to obtain covariance matrixes of the sampling signals, and taking the covariance matrixes as observation data of a convolutional neural network;
the method comprises the steps of designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally using the output feature vector to assist the classifier to make identification judgment;
training the constructed network by adopting an observation data set, updating model parameters by adopting a gradient descent algorithm, and obtaining a target of maximizing posterior probability by minimizing a cost function to obtain a trained network; inputting the observation data to be tested into a trained network, obtaining the output of the optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
Another object of the present invention is to provide an information data processing terminal for implementing the primary user transmission power identification system.
By combining all the technical schemes, the invention has the advantages and positive effects that: the method for identifying the transmitting power of the main user is mainly used for identifying the transmitting power of the main user under the single-input multi-output scene of non-Gaussian transmission signals. The method has the advantages of accurate identification, robustness to noise level and the like, and can be used for accurately identifying the transmitting power of the main user in spectrum sensing.
The present invention aims to construct detection statistics using the feature extraction capabilities of convolutional neural networks on observed data. The covariance matrix of the sampled signals is considered as observation data, so that the accurate identification of the transmitting power level value of the main user is realized. The method enables the detection result to avoid being influenced by the non-Gaussian distribution statistical characteristic of the transmission signal of the authorized user, and has robustness under the communication environment of non-Gaussian noise and low signal-to-noise ratio.
Meanwhile, the invention has the following advantages:
(1) The main user transmitting power identification method provided by the invention introduces a convolutional neural network to design test statistics, utilizes the strong capability of the convolutional neural network to extract high-dimensional characteristics from observed data, solves the problem that the design of the test statistics of the traditional power identification method needs to involve a large amount of manual processes and relies on excessive professional knowledge to cause lower accuracy, realizes the automation of power identification, and effectively solves the problem of degradation of identification performance of the traditional method in a non-ideal environment with low signal to noise ratio.
(2) The main user transmitting power identification method provided by the invention takes the covariance matrix containing rich energy correlation characteristics as the observation data of the identification model, effectively removes redundant characteristic information of the original sampling signal, and provides support for training and verification of the convolutional neural network.
(3) The invention designs the cost function by using the likelihood equation, and achieves the aim of maximizing the posterior probability by minimizing the cost function.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a primary user transmit power identification method according to an embodiment of the present invention.
Fig. 2 is a block diagram of a primary user transmit power identification system according to an embodiment of the present invention;
in the figure: 1. a data sampling module; 2. an observation data acquisition module; 3. a convolutional neural network structure design module; 4. a network training module; 5. and identifying a performance test module.
Fig. 3 is a block diagram of a feature extractor provided in an embodiment of the present invention.
Fig. 4 is a schematic diagram of a classifier according to an embodiment of the present invention.
Fig. 5 is a graph of recognition accuracy versus SNR using the proposed method and the hybrid higher order cumulant-based recognition method, respectively, provided by an embodiment of the present invention.
Fig. 6 is a graph of recognition accuracy as a function of sample number using the proposed method and the hybrid higher order cumulative amount-based recognition method, respectively, provided by an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Aiming at the problems existing in the prior art, the invention provides a method, a system, equipment and a terminal for identifying the transmitting power of a main user, and the invention is described in detail below with reference to the accompanying drawings.
As shown in fig. 1, the method for identifying the transmission power of the primary user provided by the embodiment of the invention includes the following steps:
s101, data sampling: the cognitive radio equipment acquires N time-discrete sampling vectors through K perception antennas;
s102, obtaining observation data: preprocessing the sampling signals to obtain covariance matrixes of the sampling signals, and taking the covariance matrixes as observation data of a convolutional neural network;
s103, designing a convolutional neural network structure: the method comprises the steps of designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally using the output feature vector to assist the classifier to make identification judgment;
s104, network training: training the network constructed in the step S103 by adopting the observation data set obtained in the step S102, updating the model parameters by adopting a gradient descent algorithm, and obtaining the target of maximizing the posterior probability by minimizing the cost function to obtain a trained network;
s105, recognition performance test: and inputting the observation data to be tested into a trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
As shown in fig. 2, the primary user transmit power identification system provided in the embodiment of the present invention includes:
the data sampling module 1 is used for acquiring N time-discrete sampling vectors through K perception antennas by using the cognitive radio equipment;
the observation data acquisition module 2 is used for preprocessing the sampling signals to obtain covariance matrixes of the sampling signals and taking the covariance matrixes as observation data of the convolutional neural network;
the convolutional neural network structure design module 3 is used for designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally outputting feature vectors to assist the classifier in making identification decisions;
the network training module 4 is used for training the network constructed by the convolutional neural network structural design module by adopting the observation data set obtained by the observation data obtaining module, updating the model parameters by implementing a gradient descent algorithm, and obtaining the target of maximizing the posterior probability by minimizing the cost function to obtain a trained network;
the recognition performance test module 5 is used for inputting the observation data to be tested into the trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
The technical scheme of the invention is further described below with reference to specific embodiments.
The method for identifying the transmitting power of the main user provided by the embodiment of the invention comprises the following steps:
step 1: receiving end sampling
The invention considers a multi-antenna cognitive radio network, in the network, a primary user transmits information in the same frequency band by a plurality of transmitting powers, and a secondary user identifies the transmitting power of the primary user through the cognitive radio equipment.
The invention contemplates that the primary user is equipped with a single antenna and the cognitive radio device is equipped with K cognitive antenna antennas. During one sampling window, the cognitive radio receives a total of N time-discrete sample point vectors, with for a single sample y (N) vector:
y(n)=[y 1 (n),…,y k (n),…,y K (n)] T ,n∈[1,N] (1)
wherein y is k (n) represents the nth time-discrete sample vector received at the kth sense antenna. Thus, a single discrete-time sample vector can be described as:
where s (n) represents a non-Gaussian signal transmitted from the primary client,is a K x 1 vector representing the rayleigh fading channel vector from the transmitting end to each of the sense antennas. w (n) represents colored complex Gaussian noise received by a sensing antenna, P i I=0, 1, …, I represents the level value of the transmission power of the primary user, and there are I transmission power level values available for the primary user, the magnitude relation being P 0 <P 1 <…<P I Wherein P is 0 The expression 0 indicates that the primary user is unoccupied spectrum. In addition, H i Representing a specific primary user transmit power P i Is a hypothetical case of (a).
Step 2: acquiring an observation dataset
When the cognitive radio device performs discrete sampling of the received signal, the sampled signal should be preprocessed before it is input into the convolutional neural network, since it contains too much redundant information that is not related to energy.
(2.1) considering the covariance matrix of the original sampled signal to be further converted into the covariance matrix of the sampled signal as the observation data of the convolutional neural network aiming at the task requirement of the transmission power identification because the covariance matrix of the sampled signal contains the relevant information of important energy characteristics. And in the period of a sampling window, after the cognitive radio equipment receives N time discrete sampling vectors, calculating to obtain a covariance matrix expression of the sampling signals:
the observation dataset may thus be set to:
wherein the method comprises the steps ofC representing observed data y (N) set, T is the set of data tags T. In the case of a single sample of data,represents the mth (m=1, 2, …, M) observation in the observation set, wherein +.>For observing data, vector t (m) =[t 0 ,...,t i ,...,t I ],i∈[0,I]Is a tag vector corresponding to observed data, where t i With a value of 0 or 1, when t i When=1, the corresponding observation data is +.>Belonging to hypothesis H i The category in the case can be expressed specifically as:
(2.2) at the same time due to C y (N) is a complex matrix of size KxK. C is C y (N) the matrix data is split into two input channels for describing real and imaginary information, respectively. The input data of the convolutional neural network is defined as: d (D) in (i; j; alpha), alpha representing the channel number, wherein channel 1 is the real part data D in (i;j;1)=(Real(C y (N))) i,j Channel 2 is the imaginary data D in (i;j;2)=(Imag(C y (N))) i,j (. Cndot.) i, j represents the i row, j column elements. Thus, the data D is input in Is a two-channel matrix of data size K x 2.
Step 3: design convolutional neural network structure
Optionally, the preset convolutional neural network model is based on a ResNet network model and consists of a feature extractor and a classifier.
(3.1) feature extractor: illustratively, the feature extractor as shown in fig. 3 is composed of five res net blocks, where each res net Block contains four convolution layers, taking the parameter "643*3Conv BN+RELU Stride =1 pad=1" in the first convolution layer as an example, (64) represents the number of convolution kernels, (3*3) represents the size of the convolution kernels, conv represents the convolution layer, BN represents the normalization layer, reLU represents the nonlinear activation layer, stride represents the step size of the convolution window sliding in each dimension, and pad represents the number of peripheral extension dimensions of the pooling layer. After the convolutional neural network is constructed, the data can be observedTraining and verifying it.
(3.2.1) classifier: illustratively, as shown in fig. 4, after the output feature map of the last convolution layer of the feature extractor is obtained, the feature map is flattened, and is input into a fully-connected layer, where the first layer fully-connected layer contains 1024 neuron nodes; the output layer of the deep convolutional neural network is represented by a single thermal coding vector, each component of the coding vector corresponds to different transmitting power level values selected by a main user, and the output layer is provided with 5 neuron nodes and corresponds to 5 classification score values under different assumption conditions.
(3.2.2) in order that the output of the model is a probability value between 0 and 1, the result is normalized using the Softmax function. The final output of the model can be expressed as:
it is an I x 1-dimensional probability vector. Wherein P is θ (. Cndot.) represents the set of posterior probability outputs at model parameters theta,representing the recognition model according to the observation data->At H i Let us assume the posterior probability of the lower output.
Step 4: network training
(4.1) design cost function: the training aims at maximizing the posterior probability corresponding to the true classification, so that the corresponding posterior probability extraction method is set as follows:i.e. when a single sample is hypothesis H i In the case of true, t (m) (t i ) =1, extract posterior probability ++>If not the true assumption, t (m) (t i ) =0, i.e. the corresponding posterior probability value is not extracted, so that the posterior probability under the current hypothesis does not affect the cost function.
According to the posterior probability extraction method, a likelihood function is introduced to design a cost function. The likelihood function is defined as:
further simplified into a summation: the likelihood function works log while not affecting monotonicity.
The cost function is designed as: the negative numbers of the log likelihood functions are averaged and monotonically changed.
It can be seen that minimizing the cost function maximizes the posterior probability under the correct assumption, matching the training objective.
(4.2) training procedure: and optimizing a cost function of the deep convolutional neural network by adopting a small batch random gradient descent algorithm Minibatch SGD, and updating model parameters of the deep convolutional neural network layer by a back propagation algorithm. The algorithm is converged through repeated iterative training of training data, and finally a trained convolutional neural network parameter set theta is obtained * . Obtaining observation-based dataThe following optimal network outputs are defined as:
wherein the method comprises the steps ofRepresented in the optimal network parameter set θ * The observation data is->When assuming H i The case is true posterior probability.
Step 5: testing and evaluation
(5.1) data preprocessing: calculating covariance matrix of the sampling signal to be identified;
(5.2) input data generation: the real part and the imaginary part are separated into two channels and input into a trained deep convolutional neural network;
(5.3) calculating an output vector of the deep convolutional neural network using a forward propagation algorithm, and outputting the vectorIs included in the maximum component of the composition. Wherein each component in the output vector of the deep convolutional neural network corresponds to a level value of the transmission power of the primary user, and based on a maximum posterior probability criterion, the hypothesis corresponding to the maximum component in the output vector is identified as the transmission power level value of the primary user, i.e. for a hypothesis test pair,if there is->Then for the mth sample, H i The hypothetical case of (2) will be judged to be correctly classified.
The technical scheme of the invention is further described below in connection with simulation experiments.
A. Simulation conditions
The communication link of the primary user is set to use quadrature phase shift keying (QPSK, quadrature Phase Shift Keying) signals with a symbol rate of 1MHz/s. The cognitive radio device, that is, the receiving end device, is provided with k=8 sensing antennas, and the receiving sampling frequency is 8 times of the symbol transmission rate, that is, the sampling frequency is 8MHz/s. The main user communication link has 4 non-zero transmitting powers and satisfies the relation P 1 :P 2 :P 3 :P 4 =3:5:7:9. Average signal to noise ratio is defined asThe mean value of the received white Gaussian noise is zero, and the variance is 1. The observation data set contains m=220000 observation data, including 200000 training data and 20000 validation data.
B. Emulation content
Simulation 1: under the condition that the number of sampling points N is 400 and 800 respectively, the recognition accuracy of the technical scheme is simulated along with the change of the signal to noise ratio, and the simulation result is shown in fig. 5.
Simulation 2: under the condition that the average signal-to-noise ratio is-8 dB and 2dB respectively, simulation is carried out aiming at the recognition accuracy of the technical scheme along with the change of the sampling points, and the simulation result is shown in figure 6.
For simulation 1, as can be seen from fig. 5, the recognition accuracy of the proposed example recognition method increases with increasing signal-to-noise ratio, and when the number of sample points N is 800, the recognition accuracy is better than that when the number of sample points N is 400; in addition, when the number of sampling points is 800, the recognition accuracy reaches 90% when the signal to noise ratio is-8 dB, and compared with the traditional recognition method based on mixed high-level cumulant, the method also maintains good robustness in a low signal to noise ratio area.
For simulation 2, as can be seen from fig. 6, the recognition accuracy of the proposed example method increases with the increase of the sampling points; in addition, the recognition accuracy of the proposed example method is higher than that of the conventional recognition method based on the mixed advanced accumulation amount at any sampling point number.
In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When used in whole or in part, is implemented in the form of a computer program product comprising one or more computer instructions. When loaded or executed on a computer, produces a flow or function in accordance with embodiments of the present invention, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, 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 a wired (e.g., coaxial cable, fiber optic, digital Subscriber Line (DSL), or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more 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)), etc.
The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.
Claims (7)
1. The method for identifying the transmitting power of the main user is characterized by being used for a single-input multi-output power identification task, wherein the number of transmitting antennas configured by the main user is 1, the number of sensing antennas configured by cognitive radio equipment is K, and K is more than 1, and the method for identifying the transmitting power of the main user comprises the following steps:
step one, data sampling: the cognitive radio equipment acquires N time-discrete sampling vectors through K perception antennas;
step two, obtaining observation data: preprocessing the sampling signals to obtain covariance matrixes of the sampling signals, and taking the covariance matrixes as observation data of a convolutional neural network;
step three, designing a convolutional neural network structure: the method comprises the steps of designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally using the output feature vector to assist the classifier to make identification judgment;
step four, network training: training the network constructed in the third step by adopting the observation data set obtained in the second step, updating the model parameters by implementing a gradient descent algorithm, and obtaining the target of maximizing posterior probability by minimizing a cost function to obtain a trained network;
step five, identifying performance test: inputting the observation data to be tested into a trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result;
the data sampling in the first step includes:
there are N time discrete sample vectors in one sample window, for a single sample vector y (N):
y(n)=[y 1 (n),…,y k (n),…,y K (n)] T ,n∈[1,N];
wherein y is k (n) represents the nth time-discrete sampled data received at the kth sense antenna;
the discrete point-in-time sampling vector expression is:
wherein s (n) represents a non-Gaussian signal emitted from a main user terminal;is a K x 1 vector representing the rayleigh fading channel vector from the transmitting end to each of the sensing antennas; w (n) represents colored complex Gaussian noise received by the sensing antenna; p (P) i I=0, 1, …, I represents the level value of the transmission power of the primary user, and there are I transmission power level values available for the primary user, the magnitude relation being P 0 <P 1 <…<P I The method comprises the steps of carrying out a first treatment on the surface of the Wherein P is 0 =0 indicates that the primary user is unoccupied spectrum; h i Representing a specific primary user transmit power P i Is a hypothetical case of (1);
the obtaining observation data in the second step includes:
(1) Carrying out data preprocessing on the received sampling signals to obtain covariance matrixes of the sampling signals, and constructing a labeled observation data set based on the covariance matrixes;
(2) Separating real part data and imaginary part data of the observed data into two channels to be used as input data of a convolutional neural network;
the method comprises the steps of converting an original sampling signal into a covariance matrix of the sampling signal to be used as observation data of a convolutional neural network; during a sampling window, after the cognitive radio device receives the N time discrete sampling vectors, calculating to obtain a covariance matrix expression of the sampling signals, wherein the covariance matrix expression is as follows:
the observation data set is set to:
wherein Ω Cy C representing observed data y (N) a set, T being a set of data tags T; in the case of a single sample of data,represents the mth observation in the observation set, m=1, 2, …, M; wherein->For observing data, vector t (m) =[t 0 ,...,t i ,...,t I ],i∈[0,I]Is a tag vector corresponding to observed data, where t i With a value of 0 or 1, when t i When=1, the corresponding observation data is +.>Belonging to H i The category in the case is expressed as:
covariance matrix C y The real part and the imaginary part of (N) are separated to form dual-channel data D in (i; j; α), α representing the channel number; wherein channel 1 is real part data D in (i;j;1)=(Real(C y (N))) i,j Channel 2 is the imaginary data D in (i;j;2)=(Imag(C y (N))) i,j (. Cndot.) i, j represents the i row, j column element, so data D is entered in A two-channel matrix with the data size of K multiplied by 2;
the network training in the fourth step comprises: training the network constructed in the third step by adopting the observation data set obtained in the second step,
(1) Designing a cost function based on a likelihood function;
(2) Reverse transmission is carried out according to a chained rule by utilizing a gradient descent algorithm based on a cost function, and trainable parameters in a network are continuously updated;
(3) Obtaining a target of maximizing posterior probability by minimizing a cost function, and obtaining a trained network;
wherein, the cost function is:
setting training target as minimum cost function to obtain optimal parameter theta * So that posterior probabilityMaximum;
based on the cost function, gradually updating parameters of DNN through a back propagation algorithm, and obtaining the trained optimal network output as follows:
wherein,representing a set of posterior probabilities output on a trained optimal network, wherein +.>Represented in the optimal network parameter set θ * The observation data is->When assuming H i The case is true posterior probability.
2. The method for identifying transmission power of a primary user according to claim 1, wherein the designing the convolutional neural network structure in the third step comprises: constructing a covariance matrix-adapted convolutional neural network, enabling an input signal to pass through a feature extractor module, then pass through a full-connection layer of a classifier module to output a scoring result, and finally normalizing the result by using a Softmax function to output a posterior probability, wherein the method comprises the following steps of:
(1) Extracting high-dimensional feature information of the observed data by constructing a feature extractor of the suitability of the observed data;
(2) The output feature vector is used for assisting the classifier to make identification judgment, and finally, the Softmax function outputs posterior probability values under various hypotheses.
3. The primary user transmit power identification method of claim 1, wherein the identification performance test in step five comprises:
(1) Data preprocessing: calculating covariance matrix of the sampling signal to be identified;
(2) Input data generation: the real part and the imaginary part are separated into two channels and input into a trained deep convolutional neural network;
(3) Calculating an output vector of the deep convolutional neural network by using a forward propagation algorithm, and outputting the maximum component in the vector; wherein each component in the output vector of the deep convolutional neural network corresponds to a level value of the transmission power of the main user, and based on a maximum posterior probability criterion, the hypothesis corresponding to the maximum component in the output vector is identified as the transmission power level value of the main user, namely for a hypothesis test pair,if there is->Then for the mth sample, H i The hypothetical case of (2) will be judged to be correctly classified.
4. A primary user transmit power identification system for implementing the primary user transmit power identification method of any one of claims 1-3, the primary user transmit power identification system comprising:
the data sampling module is used for acquiring N time-discrete sampling vectors through K perception antennas by using the cognitive radio equipment;
the observation data acquisition module is used for preprocessing the sampling signals to obtain covariance matrixes of the sampling signals and taking the covariance matrixes as observation data of the convolutional neural network;
the convolutional neural network structure design module is used for designing a convolutional neural network structure consisting of a feature extractor and a classifier, extracting high-dimensional feature information of observed data by constructing the feature extractor structure with the suitability of the observed data, and finally outputting feature vectors to assist the classifier in making identification decisions;
the network training module is used for training the network constructed by the convolutional neural network structure design module by adopting the observation data set obtained by the observation data obtaining module, updating the model parameters by implementing a gradient descent algorithm, and obtaining the target of maximizing the posterior probability by minimizing the cost function to obtain a trained network;
the recognition performance test module is used for inputting the observation data to be tested into the trained network, obtaining the output of the final optimal posterior probability through forward transmission, and selecting the hypothesis corresponding to the maximum posterior probability to judge the power recognition result.
5. A computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of the primary user transmit power identification method of any one of claims 1 to 3.
6. A computer readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of the primary user transmit power identification method of any one of claims 1 to 3.
7. An information data processing terminal, characterized in that the information data processing terminal is arranged to implement the primary user transmit power identification system as claimed in claim 4.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111441610.1A CN114285701B (en) | 2021-11-30 | 2021-11-30 | Method, system, equipment and terminal for identifying transmitting power of main user |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111441610.1A CN114285701B (en) | 2021-11-30 | 2021-11-30 | Method, system, equipment and terminal for identifying transmitting power of main user |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114285701A CN114285701A (en) | 2022-04-05 |
CN114285701B true CN114285701B (en) | 2024-03-29 |
Family
ID=80870589
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111441610.1A Active CN114285701B (en) | 2021-11-30 | 2021-11-30 | Method, system, equipment and terminal for identifying transmitting power of main user |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114285701B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108616470A (en) * | 2018-03-26 | 2018-10-02 | 天津大学 | Modulation Signals Recognition method based on convolutional neural networks |
CN109379153A (en) * | 2018-12-17 | 2019-02-22 | 电子科技大学 | A kind of frequency spectrum sensing method |
CN109450573A (en) * | 2018-12-17 | 2019-03-08 | 电子科技大学 | A kind of frequency spectrum sensing method based on deep neural network |
KR20210126912A (en) * | 2020-04-13 | 2021-10-21 | 엘아이지넥스원 주식회사 | CNN Based Spectrum Sensing Technique for Cognitive Radio Communications |
-
2021
- 2021-11-30 CN CN202111441610.1A patent/CN114285701B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108616470A (en) * | 2018-03-26 | 2018-10-02 | 天津大学 | Modulation Signals Recognition method based on convolutional neural networks |
CN109379153A (en) * | 2018-12-17 | 2019-02-22 | 电子科技大学 | A kind of frequency spectrum sensing method |
CN109450573A (en) * | 2018-12-17 | 2019-03-08 | 电子科技大学 | A kind of frequency spectrum sensing method based on deep neural network |
KR20210126912A (en) * | 2020-04-13 | 2021-10-21 | 엘아이지넥스원 주식회사 | CNN Based Spectrum Sensing Technique for Cognitive Radio Communications |
Non-Patent Citations (4)
Title |
---|
Deep cm-cnn for spectrum sensing in cognitive radio;C.Liu等;《IEEE Journal on Selected Areas in Communications》;20190808;第37卷(第10期);第2306-2321页 * |
一种基于特征值的多天线认知无线电盲感知算法;刘会衡等;《计算机应用研究》;20140827;第32卷(第01期);第191-193页 * |
认知网络中多传输功率感知与辨识技术研究;王丹洋;《中国博士学位论文全文数据库信息科技辑》;20190115(第1期);第1-151页 * |
高速铁路信道时频弥散特性测量、建模与仿真;丁建文;《中国博士学位论文全文数据库信息科技辑》;20191215;第1-177页 * |
Also Published As
Publication number | Publication date |
---|---|
CN114285701A (en) | 2022-04-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Hu et al. | Deep neural network for robust modulation classification under uncertain noise conditions | |
Bitar et al. | Wireless technology identification using deep convolutional neural networks | |
CN109657604B (en) | Satellite amplitude-phase signal identification and demodulation method and device based on cyclic neural network | |
CN110099019B (en) | LoRa modulation signal detection method based on deep learning | |
Oyedare et al. | Estimating the required training dataset size for transmitter classification using deep learning | |
CN113435247B (en) | Intelligent recognition method, system and terminal for communication interference | |
CN112910811B (en) | Blind modulation identification method and device under unknown noise level condition based on joint learning | |
CN111050315B (en) | Wireless transmitter identification method based on multi-core two-way network | |
CN114337880B (en) | Spectrum sensing method considering random arrival and departure of main user signals | |
CN113378644B (en) | Method for defending signal modulation type recognition attack based on generation type countermeasure network | |
CN116866129A (en) | Wireless communication signal detection method | |
CN115546608A (en) | Unmanned aerial vehicle data link electromagnetic interference classification and threat assessment method | |
CN114925720A (en) | Small sample modulation signal identification method based on space-time mixed feature extraction network | |
CN114615118A (en) | Modulation identification method based on multi-terminal convolution neural network | |
CN107770813B (en) | LTE uplink interference classification method based on PCA and two-dimensional skewness characteristics | |
CN112350790B (en) | Deep learning-based spectrum sensing detection method, device and equipment | |
CN114285701B (en) | Method, system, equipment and terminal for identifying transmitting power of main user | |
CN110351884B (en) | Spectrum opportunity access method based on double-layer multi-arm tiger machine statistical model | |
CN116566777B (en) | Frequency hopping signal modulation identification method based on graph convolution neural network | |
CN112836788A (en) | Low-power-consumption deep learning network method for interference type identification | |
Zhang et al. | Machine learning based protocol classification in unlicensed 5 GHz bands | |
CN113242201B (en) | Wireless signal enhanced demodulation method and system based on generation classification network | |
CN115409056A (en) | Automatic modulation identification method for large dynamic signal-to-noise ratio | |
CN113556157B (en) | Method and system for estimating number of transmitting antennas of MIMO system under non-Gaussian interference | |
Yang et al. | Conventional Neural Network‐Based Radio Frequency Fingerprint Identification Using Raw I/Q Data |
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 |