CN112702132B - Broadband spectrum sensing method based on convolutional neural network classifier - Google Patents
Broadband spectrum sensing method based on convolutional neural network classifier Download PDFInfo
- Publication number
- CN112702132B CN112702132B CN202011539795.5A CN202011539795A CN112702132B CN 112702132 B CN112702132 B CN 112702132B CN 202011539795 A CN202011539795 A CN 202011539795A CN 112702132 B CN112702132 B CN 112702132B
- Authority
- CN
- China
- Prior art keywords
- neural network
- classifier
- convolutional neural
- spectrum sensing
- sub
- 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
Landscapes
- Mobile Radio Communication Systems (AREA)
- Error Detection And Correction (AREA)
Abstract
The invention relates to a broadband spectrum sensing method based on a convolutional neural network classifier, and belongs to the field of wireless communication. The method comprises the following steps: s1: acquiring a large amount of original training data under different environments; s2: carrying out normalization and data equalization operation on original training data; s3: respectively designing two neural network models according to the characteristics of input data, wherein the model 1 is used for classifying the initial positions of the sub-bands occupied by the PU, and the model 2 is used for classifying the bandwidths occupied by the PU; s4: respectively inputting training data sets subjected to preprocessing operations such as normalization and sample equalization and corresponding label sets into the two convolutional neural network models to train the sub-band occupation pattern classifier; s5: and respectively inputting the real-time test data subjected to the normalization processing into the trained convolutional neural network classifier, respectively executing a classification task and outputting a sub-band occupation mode label. The invention achieves better broadband spectrum sensing performance.
Description
Technical Field
The invention belongs to the field of wireless communication, and relates to a broadband spectrum sensing method based on a convolutional neural network classifier.
Background
With the continuous revolution and evolution of mobile communication, the requirements of user services on mobile networks are higher and higher.
In order to cope with the explosive growth of mobile data traffic, massive device connection and new service and application scenes, a fifth generation mobile communication system (5G) is produced. Compared with 4G, 5G can support more diversified scenes, integrates various wireless access modes, fully utilizes spectrum resources from low frequency to high frequency, and greatly improves spectrum efficiency, energy efficiency and cost efficiency. In order to realize key technical indexes such as ultra-high frequency spectrum utilization rate, transmission rate and resource utilization rate of a 5G mobile communication system, different service types need to be provided, transmission bandwidth needs to be increased, and communication system capacity needs to be enlarged according to characteristics of different frequency bands. However, nowadays, spectrum resources are increasingly scarce, and the most important strategy of the wireless communication technology is static spectrum access, that is, a fixed spectrum at a specific location at a specific time is authorized by the government to be dedicated to a Primary User (PU), while a Secondary User (SU) cannot use the spectrum even if the PU does not use the spectrum. Extensive research by the federal communications commission in the united states, as well as other countries, has shown that this traditional fixed spectrum allocation results in low utilization of most existing frequency bands. In order to improve spectrum efficiency and meet the huge demand of spectrum, a Cognitive Radio (CR), which is a key technology capable of realizing spectrum sharing, is proposed.
The main working principle of cognitive radio is as follows: the CR allows the SU to sense the usage status of a PU in an authorized Frequency Band (LFB), and if the SU senses that the LFB is not used, the SU can access the LFB for wireless transmission without affecting the PU. The cognitive radio mainly comprises technologies such as spectrum sensing, dynamic spectrum management and dynamic spectrum access, wherein the spectrum sensing is a prerequisite for realizing spectrum sharing. In addition, with the rapid development of mobile communication broadband services, providing broadband access services with high rate and low latency will become one of the basic requirements of future mobile communication systems. The authorized frequency band required to be sensed by Spectrum Sensing also extends from a broadcast television frequency band of hundreds of MHz to a mixed frequency band of several GHz, and these changes make a broadband Spectrum Sensing (WSS) technology a large research hotspot in the time CR field. In the existing broadband spectrum sensing scheme, the main broadband spectrum sensing technologies mainly include: broadband energy detection techniques, compressed sensing techniques, and the like.
The existing broadband spectrum sensing technology mainly has the following defects:
the broadband energy detection technology converts a broadband signal into a plurality of paths of parallel narrowband signals, and then compares the energy value of each path of signal with a preset threshold to judge whether each path of narrowband signal is occupied. The key of the method is the threshold selection of each path of signal. In addition, in a real environment, the noise power is not constant, and the threshold value is difficult to be accurately determined due to the influence of uncertainty of noise power estimation.
Under the premise that the broadband spectrum has sparsity, the compression sensing fully utilizes the sparsity or compressibility of the signal to carry out undersampling on the broadband signal, and the original signal is recovered by a reconstruction algorithm. The precondition for realizing the method is that the signal has sparsity or compressibility, and the real-time performance and accuracy of compressed sensing are poor, so that the method cannot necessarily meet the requirements of practical application.
In summary, in order to overcome the various defects of the conventional wideband spectrum sensing algorithm, a new wideband spectrum sensing method is needed to solve the above problems.
Disclosure of Invention
In view of the above, the present invention provides a wideband spectrum sensing method based on a convolutional neural network classifier.
In order to achieve the purpose, the invention provides the following technical scheme:
a broadband spectrum sensing method based on a convolutional neural network classifier comprises the following steps:
s1: acquiring a large number of broadband spectrum received signal energy observation matrixes in different environments and sub-band occupation mode labels corresponding to the matrixes;
s2: normalizing the obtained original data and carrying out sample equalization operation;
s3: designing two neural network models according to the size characteristics of the energy observation matrix;
s4: respectively training the designed neural network model by utilizing the trained data set to obtain a classifier 1 for identifying the PU occupied frequency band initial position and a classifier 2 for identifying the PU occupied bandwidth;
s5: and inputting the real-time spectrum energy observation matrixes subjected to normalization processing into the two classifiers respectively, and executing classification tasks by the two classifiers simultaneously to obtain the initial position and the occupied bandwidth of the PU occupied frequency band at the current moment.
Optionally, step S1 specifically includes:
according to the initial condition, setting the current time as t, and setting a Q multiplied by M dimensional energy observation matrix formed by energy values on all Q sub-bands received by all M SUs as XtThe frequency band occupation pattern at time t isAcquiring a large number of spectrum observation energy observation matrixes and sub-band occupation modes corresponding to the matrixes under different fading environments with different signal-to-noise ratios within a fixed length of time to form a training set theta ═ X1,X2,...,XTThe label set corresponding to the training setWherein d ═ { d ═ d(1),d(2),...,d(T)},
Optionally, step S2 specifically includes: in order to achieve a better classification effect of the classifier, the energy observation matrix in the original data set Θ obtained in S1 is normalized to obtain a new sample set Θnorm={Z1,Z2,...,ZTIn which Z istIs XtCarrying out normalization operation on the matrix, and then carrying out sample equalization processing on the normalized data set;
the sample equalization processing adopts an oversampling method to perform sample equalization, and the specific operation is as follows: counting the number of samples of each class in the L, randomly copying a part of data for the class with less number of samples, and ensuring that the data amount corresponding to each class is kept balanced; obtaining a new training sample set thetaresamp={Z1,Z2,…,ZT'Frequency band occupation mode label set corresponding to training sample setWherein T' is less than or equal to T.
Optionally, the step S3 includes: determining a proper convolution kernel size, the number of convolution kernels and the size of a sampling window by adopting a 5-layer network structure according to the characteristics of an energy observation matrix, and respectively designing two convolution neural networks with the depth of 5 layers, wherein each of the two networks consists of 3 convolution layers and 2 full-connection layers; the first two convolutional layers comprise a convolutional layer and a pooling layer, and all convolutional layers adopt ReLu as an activation function to enhance the nonlinearity of the network;
the network structure adopts a maximum pooling method to achieve the purpose of reducing the calculated amount by reducing the size of the characteristic diagram; after the output characteristic diagram of the last convolution layer is obtained, flattening operation is carried out on the characteristic diagram, and the characteristic diagram is converted into a new characteristic vector; to prevent overfitting, a dropout layer is added on top of the first fully-connected layer, and then a second fully-connected layer is added.
Optionally, in the step S4, the data { Θ is converted into dataresampD } input classifier 1, { Θresamp,NpInputting the data into a classifier 2, and training the two classifiers respectively; and updating parameters of the convolutional neural network by adopting an adaptive moment estimation method, using the classified cross entropy as a target loss function, and stopping training when the target function is minimum.
Optionally, in step S5, the normalized spectral energy observation matrix Z at the current time is input into the two classifierstThe two classifiers respectively execute classification tasks to obtain the initial position of the frequency band occupied by the PUAnd occupy bandwidth
The invention has the beneficial effects that: the invention solves the problem of low environmental adaptability of the traditional broadband spectrum sensing and improves the sensing accuracy.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof.
Drawings
For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
fig. 1 is a schematic diagram of a sub-band occupation pattern on a broadband authorized frequency band;
FIG. 2 is a schematic diagram of a neural network designed according to the present invention;
FIG. 3 is a flowchart of a sub-band occupation pattern classification method based on a convolutional neural network classifier.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
Wherein the showings are for the purpose of illustrating the invention only and not for the purpose of limiting the same, and in which there is shown by way of illustration only and not in the drawings in which there is no intention to limit the invention thereto; for a better explanation of the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numerals in the drawings of the embodiments of the present invention correspond to the same or similar components; in the description of the present invention, it should be understood that if there is an orientation or positional relationship indicated by terms such as "upper", "lower", "left", "right", "front", "rear", etc., based on the orientation or positional relationship shown in the drawings, it is only for convenience of description and simplification of description, but it is not an indication or suggestion that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes, and are not to be construed as limiting the present invention, and the specific meaning of the terms may be understood by those skilled in the art according to specific situations.
As shown in fig. 3, taking the number of subbands Q being 24 and the number of SUs being M being 32 as an example, the implementation process of the wideband spectrum sensing method based on the convolutional neural network classifier specifically includes the following three initial conditions and five main steps.
Initial condition 1:
assume that there are M SUs and a single PU in a cognitive radio network CRN. And dividing the broadband authorized frequency band corresponding to the PU into Q sub-bands which are not overlapped and have w bandwidths. The starting position of the sub-band occupied by the PU at the time point i is d(i)(d(i)∈{0,1,...,I1}) PU occupies subband width ofWherein I1Q and I are not more than Q2≤Q-I1. The set of the sub-bands occupied by the PU at the moment i isThe band occupation pattern is expressed as
Initial condition 2:
at time i, the observed signal on the q subband received by the mth SU is represented as:
wherein h ism,qRepresents the channel gain of the PU between the qth subband and the mth SU; n is a radical ofq(i) Indicating that the SU receiver has corresponding additive white Gaussian noise on the qth sub-band;transmitting power for the PU; h1And H0Respectively indicating whether the PU occupies the frequency band; sq(i) Represents the corresponding transmission signal on the q-th subband at time i of the PU transmitter, under the assumption of H1Under the conditions thatTime Sq(i) 1, otherwise Sq(i)=0。
In the t-th sensing process, the energy value of the q-th sub-band received by the mth SU is:
wherein 2w tau represents the number of signal sampling points in a sensing interval, and the duration of a single sensing interval is tau. The energy observation matrix, which is composed of the energy values on all subbands received by all SUs, is therefore:
initial condition 3:
assume that enough energy observation matrices Θ are acquired in a fixed amount of time X1,X2,…,XTAnd subband occupation mode label set corresponding to each energy observation matrix
On the basis of the initial conditions, the broadband spectrum sensing method based on the convolutional neural network classifier is implemented by the following steps:
step 1: obtaining raw training data
According to the initial condition 3, a large number of energy observation matrixes Θ and corresponding subband occupation mode labels L under different SNR conditions and different fading environments are obtained within a sufficiently long time.
Step 2: data pre-processing
2.1: carrying out the following normalization operation on the energy observation matrix in the original data set theta obtained in the step 1 to obtain a new sample set thetanorm={Z1,Z2,…,ZT}:
Wherein the content of the first and second substances,is the energy observation matrix X in ΘtMiddle elementThe value after the normalization is carried out,is a matrix XtThe average value of all the elements in (A),are respectively a matrix XtThe maximum and minimum values of all elements in (c).
2.2: in order to enable the classifier to achieve a better classification effect, sample equalization processing is carried out on the data set normalized in the step 2.1, common sample equalization operations include over-sampling and under-sampling, and the over-sampling method is adopted for sample equalization because important original information is easily lost due to the under-sampling. The specific operation is as follows: and counting the number of samples of each class in the L, and randomly copying a part of data for the class with less number of samples to ensure that the data amount corresponding to each class is kept balanced. Thus obtaining a new training sample set thetaresamp={Z1,Z2,…,ZT'Frequency band occupation mode label set corresponding to training sample setWherein T' is less than or equal to T.
And step 3: designing convolutional neural networks
The invention adopts a 5-layer network structure, determines proper convolutional kernel size, convolutional kernel number and sampling window size according to the characteristics of an energy observation matrix, and respectively designs two convolutional neural networks with the depth of 5 layers, wherein the two networks consist of 3 convolutional layers and 2 full-connection layers. Fig. 2 shows a specific structure of the convolutional neural network according to the present invention, and specific parameter settings are given by taking Q-24 subbands and M-32 SUs as examples.
3.1: and (3) rolling layers: the sizes of convolution kernels of the three convolution layers are all 3 multiplied by 3, the number of the convolution kernels is respectively set to be 16, 32 and 64, the step length is 1, and the filling number is 0; the sampling window size is 2 x 2 during pooling, and the step size is set to 1. The output matrices of the three convolutional layers are respectively as follows:
wherein, Wl vAndrespectively representing convolution operational characters, a weight matrix and an offset value matrix of a v-th convolution kernel in the l layer, wherein v represents the number of convolution kernels of each convolution layer; ReLu (-) is an activation function, and the network model adopts ReLu as the activation function of each convolution layer for enhancing the nonlinearity of the neural network, namely:
the pooling operation replaces a plurality of adjacent features with one feature, and the network model adopts a maximum pooling method, namely, the maximum feature is selected from the plurality of adjacent features to be used as the pooled feature. Pooling thus achieves a reduction in computational effort by reducing the feature size.
3.2: full connection layer: after the output feature map of the last convolutional layer is obtained, the feature map is flattened and converted into a new feature vector. In order to prevent overfitting, a dropout layer is added on the basis of a first full-connection layer, the dropout parameter is 0.2 in the invention, and then a second full-connection layer is added. The output vectors of the two fully-connected layers are as follows:
wherein, XFC1、XFC2、Andthe probability that the input sample is classified into the jth class is represented by the output of the jth neuron of the softmax classifier of the last full connection layer, and the classifier 1 and the classifier 2 in the invention respectively have I ═ I1、I=I2。
And 4, step 4: training convolutional neural network model
The energy observation matrix set theta obtained in the step 2 is usedresampAnd corresponding set of labels LresampThe training data and the training labels are used as the training data and the training labels of the convolutional neural network classifier to complete the model training of the sub-band occupation pattern classifier, so that the sub-band occupation pattern recognition can be carried out from the next step. Data { theta }resamp D input classifier 1, { Θresamp,NpAnd inputting the data into a classifier 2, and training the two classifiers respectively. The invention adopts an adaptive moment estimation method to train a convolutional neural network, and uses the classified cross entropy as a target loss function:
where θ is a parameter in the CNN model, yi,sE {0,1} represents whether the sample s is divided into the ith class, if so, 1 is taken, otherwise, 0 is taken; p is a radical ofi,sIs the predicted probability value output by the softmax classifier, representing the probability that the s-th sample energy observation matrix is predicted as tag i. When y isi,s1 and pi,sCross entropy is minimal at 1 and training is stopped when the objective loss function reaches a minimum.
And 5: normalized matrix Z of energy observation matrix of input current momenttRespectively executing classification tasks by using the two classifiers obtained in the step 4, and combining output results to obtain a sub-band occupation mode label corresponding to the current moment
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.
Claims (5)
1. A broadband spectrum sensing method based on a convolutional neural network classifier is characterized by comprising the following steps: the method comprises the following steps:
s1: acquiring a large number of broadband spectrum received signal energy observation matrixes in different environments and sub-band occupation mode labels corresponding to the matrixes;
s2: normalizing the obtained original data and carrying out sample equalization operation;
s3: designing two neural network models according to the size characteristics of the energy observation matrix;
s4: respectively training the designed neural network model by utilizing the trained data set to obtain a classifier 1 for identifying the PU occupied frequency band initial position and a classifier 2 for identifying the PU occupied bandwidth;
s5: inputting the real-time spectrum energy observation matrixes subjected to normalization processing into two classifiers respectively, and executing classification tasks by the two classifiers simultaneously to obtain the initial position and occupied bandwidth of the PU occupied frequency band at the current moment;
the step S1 specifically includes:
according to the initial condition, setting the current time as t, and setting a Q multiplied by M dimensional energy observation matrix formed by energy values on all Q sub-bands received by all M SUs as XtThe frequency band occupation pattern at time t isAcquiring a large number of spectrum observation energy observation matrixes under different fading environments with different signal-to-noise ratios within a fixed length of time, and forming a training set theta (X) by using a sub-band occupation mode corresponding to the matrixes1,X2,...,XTThe label set corresponding to the training setWherein d ═ { d ═ d(1),d(2),...,d(T)},
d(t)Comprises the following steps: the PU occupies the initial position of the sub-band at the time t;
2. The broadband spectrum sensing method based on the convolutional neural network classifier as claimed in claim 1, wherein: the step S2 specifically includes: in order to achieve a better classification effect of the classifier, the energy observation matrix in the original data set Θ obtained in S1 is normalized to obtain a new sample set Θnorm={Z1,Z2,...,ZTIn which Z istIs XtCarrying out sample equalization processing on the normalized data set after the matrix is subjected to normalization operation;
the sample equalization processing adopts an oversampling method to perform sample equalization, and the specific operation is as follows: counting the number of samples of each class in the L, randomly copying a part of data for the class with small number of samples, and ensuring that the data amount corresponding to each class is kept balanced; obtaining a new training sample set thetaresamp={Z1,Z2,…,ZT'Frequency band occupation mode label set corresponding to training sample setWherein T' is less than or equal to T.
3. The broadband spectrum sensing method based on the convolutional neural network classifier as claimed in claim 2, wherein: the step S3 includes: determining a proper convolution kernel size, the number of convolution kernels and the size of a sampling window by adopting a 5-layer network structure according to the characteristics of an energy observation matrix, and respectively designing two convolution neural networks with the depth of 5 layers, wherein each of the two networks consists of 3 convolution layers and 2 full-connection layers; the first two convolutional layers comprise a convolutional layer and a pooling layer, and all convolutional layers adopt ReLu as an activation function to enhance the nonlinearity of the network;
the network structure adopts a maximum pooling method to achieve the purpose of reducing the calculated amount by reducing the size of the characteristic diagram; after the output characteristic diagram of the last convolution layer is obtained, flattening operation is carried out on the characteristic diagram, and the characteristic diagram is converted into a new characteristic vector; to prevent overfitting, a dropout layer is added on top of the first fully-connected layer, and then a second fully-connected layer is added.
4. The broadband spectrum sensing method based on the convolutional neural network classifier as claimed in claim 3, wherein: in said step S4, data { Θ }resampD } input classifier 1, { Θresamp,NPInputting the data into a classifier 2, and training the two classifiers respectively; and updating parameters of the convolutional neural network by adopting an adaptive moment estimation method, using the classified cross entropy as a target loss function, and stopping training when the target function is minimum.
5. The broadband spectrum sensing method based on the convolutional neural network classifier as claimed in claim 4, wherein: in step S5, the normalized spectral energy observation matrix Z at the current time is input to the two classifierstThe two classifiers respectively execute classification tasks to obtain the initial position of the frequency band occupied by the PUAnd occupy bandwidth
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011539795.5A CN112702132B (en) | 2020-12-23 | 2020-12-23 | Broadband spectrum sensing method based on convolutional neural network classifier |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011539795.5A CN112702132B (en) | 2020-12-23 | 2020-12-23 | Broadband spectrum sensing method based on convolutional neural network classifier |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112702132A CN112702132A (en) | 2021-04-23 |
CN112702132B true CN112702132B (en) | 2022-06-21 |
Family
ID=75509351
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011539795.5A Active CN112702132B (en) | 2020-12-23 | 2020-12-23 | Broadband spectrum sensing method based on convolutional neural network classifier |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112702132B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115209418A (en) * | 2022-06-13 | 2022-10-18 | 海南大学 | Intelligent broadband spectrum sensing technology based on pre-training basic model |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111148107A (en) * | 2019-12-27 | 2020-05-12 | 北京环佳通信技术有限公司 | Broadband spectrum sensing method and device based on compression filtering |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101115129B1 (en) * | 2004-10-20 | 2012-03-13 | 콸콤 인코포레이티드 | A method of mult-frequency band operation in a wireless network |
CN103532649A (en) * | 2013-10-22 | 2014-01-22 | 北京邮电大学 | Environment cognition technique and equipment applicable to aerospace information network |
KR102156790B1 (en) * | 2014-02-11 | 2020-09-16 | 한국전자통신연구원 | Spectrum sensing method and spectrum sensing device |
CN107995628A (en) * | 2017-12-18 | 2018-05-04 | 北京工业大学 | A kind of cognition wireless network multi-user Cooperation frequency spectrum sensing method of deep learning |
CN109245840B (en) * | 2018-10-15 | 2021-02-02 | 哈尔滨工业大学 | Frequency spectrum prediction method based on convolutional neural network in cognitive radio system |
CN109547133B (en) * | 2018-12-06 | 2021-04-30 | 杭州电子科技大学 | SVM high-efficiency spectrum sensing method based on Cholesky decomposition sampling covariance matrix |
CN109495198A (en) * | 2019-01-25 | 2019-03-19 | 西安电子科技大学 | Lack sampling frequency spectrum sensing method based on the detection of residual error correlation matrix |
CN111447023B (en) * | 2020-03-10 | 2021-10-26 | 重庆邮电大学 | Spectrum sensing method based on authorized user transmitter transmission mode classification |
CN111614421B (en) * | 2020-05-19 | 2022-02-11 | 重庆邮电大学 | Spectrum sensing method based on unsupervised machine learning classification algorithm |
-
2020
- 2020-12-23 CN CN202011539795.5A patent/CN112702132B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111148107A (en) * | 2019-12-27 | 2020-05-12 | 北京环佳通信技术有限公司 | Broadband spectrum sensing method and device based on compression filtering |
Also Published As
Publication number | Publication date |
---|---|
CN112702132A (en) | 2021-04-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11190944B2 (en) | Spectral sensing and allocation using deep machine learning | |
CN109379153B (en) | Spectrum sensing method | |
Alhazmi et al. | 5G signal identification using deep learning | |
CN112422213B (en) | Efficient spectrum sensing method based on support vector machine | |
Yuan et al. | DeepMorse: A deep convolutional learning method for blind morse signal detection in wideband wireless spectrum | |
CN112702132B (en) | Broadband spectrum sensing method based on convolutional neural network classifier | |
Behura et al. | WiST ID—Deep learning-based large scale wireless standard technology identification | |
CN115296759A (en) | Interference identification method based on deep learning | |
Almazrouei et al. | A deep learning approach to radio signal denoising | |
He et al. | Identification of ISM band signals using deep learning | |
Muzaffar et al. | A review of spectrum sensing in modern cognitive radio networks | |
CN116761223B (en) | Method for realizing 4G radio frequency communication by using 5G baseband chip and vehicle-mounted radio frequency system | |
CN114114363A (en) | Opportunistic signal sensing method and system based on time frequency and convolutional neural network and opportunistic signal positioning method | |
Wang et al. | Residual learning based RF signal denoising | |
CN116865884A (en) | Broadband spectrum sensing method based on online learning | |
Hiremath et al. | Blind identification of radio access techniques based on time-frequency analysis and convolutional neural network | |
Yadav et al. | Application of Machine Learning Framework for Next‐Generation Wireless Networks: Challenges and Case Studies | |
Gu et al. | Deep learning based intelligent recognition method in heterogeneous communication networks | |
Valadão et al. | Cooperative spectrum sensing system using residual convolutional neural network | |
CN117667227A (en) | Parameter adjustment method and related equipment | |
Zhang et al. | CNN based wideband spectrum occupancy status identification for cognitive radios | |
CN106792728A (en) | Close to optimal imparametrization cognitive radio cyclo-stationary cooperation frequency spectrum sensing method | |
Sun et al. | Radar target recognition based on the multi-resolution analysis theory and neural network | |
Gu et al. | Deep learning aided friendly coexistence of WiFi and LTE in unlicensed bands | |
CN112232430A (en) | Neural network model testing method and device, storage medium and electronic device |
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 |