CN110365435A - A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm - Google Patents

A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm Download PDF

Info

Publication number
CN110365435A
CN110365435A CN201910618633.1A CN201910618633A CN110365435A CN 110365435 A CN110365435 A CN 110365435A CN 201910618633 A CN201910618633 A CN 201910618633A CN 110365435 A CN110365435 A CN 110365435A
Authority
CN
China
Prior art keywords
user
frequency spectrum
energy
primary user
method based
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.)
Pending
Application number
CN201910618633.1A
Other languages
Chinese (zh)
Inventor
邢焕来
王成玮
詹大为
戴朋林
罗寿西
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwest Jiaotong University
Original Assignee
Southwest Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Southwest Jiaotong University filed Critical Southwest Jiaotong University
Priority to CN201910618633.1A priority Critical patent/CN110365435A/en
Publication of CN110365435A publication Critical patent/CN110365435A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/382—Monitoring; Testing of propagation channels for resource allocation, admission control or handover

Landscapes

  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm, specifically: in cognition wireless network, secondary user detects the energy in present channel environment and result is sent to one user as fusion center, and label whether primary user is by phased manner itself using frequency spectrum resource issues fusion center；Data set, the specifically used unilateral sampling technique and exclusive feature binding technology based on gradient of Lightgbm algorithm are established using the method based on energy measuring；Then by the feature parameter vectors collection of building, it is divided into training set and test set, is trained and tests respectively；Fusion center differentiates whether channel can be used, and all secondary users are reinformed after obtaining result；For the present invention using false alarm rate 0.1 for meeting the requirement of IEEE 802.11, verification and measurement ratio has 6% to 7% compared to SVM promotion, while misclassification rate, misclassification risk are also remarkably decreased.

Description

A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm
Technical field
The invention belongs to wirelessly communicate and field of artificial intelligence.More particularly to a kind of conjunction based on Lightgbm algorithm Make energy measuring frequency spectrum sensing method.
Background technique
Cognition wireless network has been widely accepted now as a kind of effective measures for improving frequency spectrum resource utilization rate, simultaneously Its a part that can also become future wireless network equipment.Realize that the vital step of cognition wireless network is to successfully complete Frequency spectrum perception can make unauthorized user obtain the action message of authorized user, then access unauthorized frequency range while avoid Authorized user's transmitting terminal is interfered.Wherein, authorized user is also known as primary user, and unauthorized user is also known as time user.It crosses The frequency spectrum perception problem of 10 years cognition wireless networks is gone to cause the concern and research of a large amount of scholars, many technologies are suggested, In most notably energy measuring, recycle spectral technology and matched filtering technique.In order to ensure unauthorized user can obtain nothing The information of line signal environment variation, makes that unauthorized user equipment has study and inferential capability is extremely important.Based on this reason, Solved the problems, such as that frequency spectrum perception this thinking under environment of cognitive radio network was gradually connect using machine learning techniques in recent years It receives.For example, K.M.Thilina et al. is proposed using a variety of supervised learnings and unsupervised learning methods such as KNN, SVM, K-means It realizes frequency spectrum perception, is compared in various aspects such as training time, testing time, ROC curve, verification and measurement ratios, compared to tradition Or, and or k order method, the performance of machine learning algorithm is all preferable, O.P.Awe et al. propose Kalman filter from Adaptive channel parametric learning method carries out frequency spectrum perception, it is also proposed that the Bayesian learning of variation carries out in cognition wireless network The technology of frequency spectrum perception, C.Zhao et al. propose the progress blind perception of frequency spectrum under time-varying multipath flat fading channel.It should be noted that , frequency spectrum perception needs to find spectrum interposition on the time or Spatial Dimension.However, although numerous scholars are with regard to frequency spectrum Perception problems have made intensive studies, but need to solve there are also many problems.Firstly, existing most of work only considered list One primary user, or more general topological model is not accounted for, a wide range of network model comprising multiple authorized users can More flexible such as WRAN in practical situations can be understood, but it may be just less suitable in small-scale mobile primary user's model With, such as WPANs.
Summary of the invention
To solve the above problems, especially comprising multiple primary users (authorized user) and secondary user's (unauthorized user) and The cooperative spectrum sensing carried out on diverse geographic location.It should increase than common SVM algorithm verification and measurement ratio, reduce letter again Making an uproar influences than low situation bring.Therefore the present invention proposes a kind of simpler while versatile detection method of principle.
The present invention provides a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm, specifically includes following Step:
Step 1: in cognition wireless network, secondary user detects the energy value in present channel environment and result It is sent to one user as fusion center；
Step 2: the characteristic of data set is sampled and established using the method based on energy measuring, primary user is by phased manner Whether itself is occupied the label segment that frequency spectrum resource is sent to fusion center and constructs data, this completes data sets Building；
Step 3: in fusion center, the training dataset of building is fed for the training of Lightgbm algorithm and obtains model, The specifically used unilateral sampling technique and exclusive feature binding technology based on gradient of Lightgbm algorithm；
Step 4: secondary user energy sensing and is transferred to fusion center again, as test vector；
Step 5: fusion center differentiates whether channel can be used, and all secondary users are reinformed after obtaining result；
Step 6: periodic wake front end perceptron repeats step 1 to step 5, if the repetition period, which is less than, branches to step 4。
It all can include signal transmitter and receiver in time user and primary user device in above-mentioned steps 1, while secondary user It also include front end awareness apparatus.It is paid close attention in the present invention in the frequency spectrum perception stage, is carried out to whether primary user emits signal Judgement, i.e. model simplification are that the awareness apparatus of time user emits the transmitter of primary user the perception of signal.
Above-mentioned steps 3 specifically:
3.1: data label setting:
Assuming that primary user and secondary user's share a frequency band and there is no interference between secondary user and primary user.Cognition M time users are contained in Radio Network System, are denoted as m=1 ..., M；N number of primary user is contained, n=1 ..., N is denoted as； One of them user is as fusion center., can be there are four types of scene, i.e. (0,0) if M is 2 here, (0,1), (1,0), (1,1), two primary users are online or both offline or one of them online situation, if M is 3, scene will have 8 Kind, but 2 are enough to describe the problem.WithThe geographical location two dimension for respectively representing m-th user and n-th of primary user is sat Mark.
In environment of cognitive radio network, authorized user can be exactly online or offline there are two types of working condition.Use KnGeneration The state of the busy channel resource of n-th of primary user of table, if Kn=1 to represent primary user online, currently occupies frequency spectrum resource, this When time user it cannot be interfered using frequency spectrum resource, if Kn=0 represents that primary user is offline, and current primary user does not occupy frequency spectrum Resource, frequency spectrum resource can be used in secondary user at this time；So K=(K1,...,KN)TBe represent all primary user's working conditions to Amount, is indicated by the mode of binary hypothesis test:
3.2: normalized energy is horizontal:
It is assumed that authorized user and unauthorized user are static non-moving states, as base station or TV station etc., while perceiving Stage shadow fading and multipath fading are also quasi-static.The front end Energy-aware equipment of each user is adopted in time τ Sample w τ complex baseband signal sample, bandwidth are expressed as w；Energy YmIt is horizontal to represent the normalized energy that time user m is received, indicates Are as follows:
Wherein η be noise power spectral density, be defined as η=E [| Nm(i)|2], Nm(i) representative time user m receives heat and makes an uproar Sound, E indicate that time user m receives the expectation of noisy samples；Xm(i) i-th of sample of signal that time user m is received, table are represented It is shown as:
In formula, H0Indicate that primary user does not have online, H1Indicate that at least one primary user is online；hm,nIt represents between PUn and SUm Channel gain；Wn(i) the transmitting signal of primary user n is represented；
3.3: construct the set of eigenvectors based on detection energy:
M time users are contained in our cognitive wireless network system, receive energy YmThe vector of structure is expressed as:
Y=(Y1,...,YM)T (4)
Since under the operating mode of primary user, each energy value YmObey non-central chi square distribution, freedom degree and non- Center Parameter is as follows:
R=2w τ (5)
In above formulaIt is the fixed transmission power of primary user n, gm,n=| hm,n|2It is power attenuation, Its calculation formula is as follows
Here | | | | represent Euclidean distance, PL (dist)=dist-θRepresent the road about distance dist and loss coefficient θ Diameter loss；Here primary user and secondary user meet 802.22 agreements, meanwhile, νm,nAnd ψm,nRespectively represent multipath fading and shade Decline, and the fading coefficients ν in detecting period sectionm,nAnd ψm,nTo be quasi-static, as 1；
When there is multiple samples, energy Distribution value Gaussian distributed；Therefore energy vectors can be distributed from multivariate Gaussian Middle extraction, mean value and variance are as follows:
μYm=r+ ζm (8)
Therefore the mean vector of energy vectors and covariance matrix are as follows:
Trained rank module and test module are contained in cooperative spectrum sensing frame, therebetween independently of one another, parallel Work.First training module classifier once must be trained after cognition wireless network start-up operation, such classifier can To carry out real-time testing in test module, system periodically can collect training vector according to channel circumstance variation again and carry out To update training module, test module needs extremely short time delay, although the training module training time is longer, does not influence to survey for training The efficient operation of die trial block, the algorithm that we use use supervised learning technology, this requires authorized user to need in some times The case where using in section, is sent to fusion center, to construct a certain number of trained labels.
About Lightgbm algorithm: original unilateral sampling technique based on gradient: in Adaboost, sample weights conduct Change an instruction of data importance, however there is no the concept of sample weights in GBDT, so used in Adaboos This method cannot directly indiscriminately imitate.But it can be used as the useful of data sampling in the gradient in GBDT between data instance Information.If the gradient of a data instance is smaller, training error will be smaller.One is exactly to lose than relatively straightforward idea The data instance of these small gradients is abandoned, but does so the accuracy rate that can change the distribution of data and influence model.
Therefore, the invention proposes the unilateral sampling technique used based on gradient and exclusive feature binding technologies Lightgbm algorithm solves the problems, such as frequency spectrum perception.Unilateral sampling technique based on gradient remains all big gradient examples, together Stochastical sampling in Shi little gradient example；In order to make up the influence to data distribution, when calculating information gain, in small ladder It spends and uses constant multiplier in data instance, this technology first classifies data instance to the absolute value of gradient, and selects wherein A × 100%, then stochastical sampling b × 100% in remaining data, to having sampled when calculating information gain Small gradient example data multiplied byIt does so and pays close attention to untrained number in the case where not changing initial data distribution Factually example.And exclusive feature binding technology can effectively reduce the feature of data: high dimensional data is usually than sparse, sparse spy Sign data space gives a possibility that reducing the method that feature hardly affects simultaneously we provide a kind of design.Dilute It dredges in feature space, is exclusive incoherent between many features, for example it nonzero value will not occur simultaneously.Therefore exclusive feature bundle Technology is tied up exclusive feature binding into an individual feature, feature histogram becomes O from O (data × features) (data × bundles), bundles < < features here, so it accelerates training speed while without influencing accuracy rate.
Compared with prior art, beneficial effects of the present invention:
Using the cooperation energy measuring frequency spectrum perception of Lightgbm algorithm, as a result better than the SVM algorithm (line having proposed at present Property kernel function).Using the present invention, using false alarm rate 0.1 for meeting the requirement of IEEE 802.11, verification and measurement ratio is compared SVM promotion has 6% to 7%, while misclassification rate, misclassification risk are also remarkably decreased, and furthermore also learns primary user by experiment Inventive can be also substantially strong compared with SVM algorithm under different state of signal-to-noise by transmission power difference i.e. time user, has stronger Practicability, preferable the performance in the lower situation of signal-to-noise ratio, robustness is stronger.Frequency spectrum perception application is come It says, it is significant.
Detailed description of the invention
Fig. 1 is the frequency spectrum perception frame based on Lightgbm algorithm.
Fig. 2 is the cooperative spectrum sensing model based on geographical location.
Fig. 3 is the box traction substation of Lightgbm algorithm and the SVM algorithm verification and measurement ratio under primary user's different transmission power.
Fig. 4 is the box traction substation of Lightgbm algorithm and SVM algorithm the misclassification relative risk under primary user's different transmission power.
Fig. 5 is the box traction substation of Lightgbm algorithm and SVM algorithm the misclassification rate under primary user's different transmission power.
Fig. 6 is the ROC curve of Lightgbm algorithm and SVM algorithm when primary user's transmission power is 100mw.
Fig. 7 is the ROC curve of Lightgbm algorithm and SVM algorithm when primary user's transmission power is 200mw.
Fig. 8 is the ROC curve of Lightgbm algorithm and SVM algorithm when primary user's transmission power is 300mw.
Fig. 9 is the ROC curve of Lightgbm algorithm and SVM algorithm when primary user's transmission power is 400mw.
Specific embodiment
Implementation of the invention is further described with reference to the accompanying drawing.
The present invention utilizes the cooperation energy measuring frequency spectrum sensing method of Lightgbm algorithm, and system structural framework figure is for example attached Shown in Fig. 1.Initially set up the cooperative spectrum sensing model (as shown in Figure 2) based on geographical location.Emulation is to use python3.6.2 In 64 PC, memory RAM 16G is carried out under six core i7 (3.2GHz) environment.
The performance indicator that the present invention compares is as follows:
A) verification and measurement ratio (Detection Probability), unauthorized user successfully is detected when authorized user is online Probability, result are the average detected rate that experiment runs 200 times.
B) misclassification risk (Misclassification Risk), classifier provides authorized user when authorized user is online The probability of offline label, result are the average misclassification relative risk that experiment runs 200 times.
C) misclassification rate (Misclassification Error Rate), the probability of classifier misjudgment, i.e., awarding Power user offline is judged as online and is judged as offline probability authorized user is online, and result be that experiment is run 200 times Average misclassification rate.
D) ROC curve (Receiver operating characteristic curve), result are experiment independent operating Averaged curve after 200 times, this index embody the whole classification performance of algorithm.
Embodiment
Verify that the present invention is based on Lightgbm algorithms compared with SVM algorithm carries out performance by emulation experiment to solve to close Make the availability and feasibility of energy measuring frequency spectrum perception problem.The present invention is set in the cooperative spectrum sensing based on geographical location In model, there are 16 unauthorized users, they are evenly distributed in the grid of 4*4, furthermore also deposit in the system model of attached drawing 2 In two authorized users, two-dimentional geographical location is respectively the coordinate position of (500m, 1500m) and (- 1500m, 0m).Simulation parameter It is provided that detecting period section is 100, bandwidth 5MHz, noise power spectral density is -174dBm, each authorized user hair Penetrating end power is 200mW, and path loss coefficient is 4, and multipath fading and shadow fading coefficient are all 1, on each authorized user The probability of line is 0.5.Linear kernel function is had been proven that in the work of early period in the outstanding performance of this problem, so SVM Selection of kernel function is linear kernel function.It is 200 test vectors in training vector is 1200.Attached drawing 3,4,5 is solid in standard The present invention examines under authorized user's different transmission power compared to the SVM algorithm to behave oneself best before this when fixed false alarm rate is 0.1 The box traction substation of survey rate, misclassification risk and misclassification rate, hollow box body is SVM algorithm in figure, and black cabinet is Lightgbm Algorithm.It can be obtained from 3 box traction substation of attached drawing, be 0.1 in the fixed false alarm rate of standard, authorized user's transmission power is in 50mW, 100mW When, unauthorized receiving end signal-to-noise ratio is too low, average detected rate 50% hereinafter, it is general at least 70% or more verification and measurement ratio It is just of practical significance, the requirement of verification and measurement ratio is met when authorized user's transmission power is in 200mW, it can be seen that Lightgbm algorithm In 200mW, 300mW, 400mW, verification and measurement ratio all does very well compared with SVM, correspondingly, seeing false detection rate and erroneous detection risk in Fig. 4,5 Rate Lightgbm algorithm is also good compared with SVM algorithm.Fig. 6,7,8,9 illustrate that the present invention exists compared to the SVM algorithm to behave oneself best before this ROC curve under authorized user 100mW, 200mW, 300mW, 400mW.See from Fig. 6, as the result being previously obtained, Signal-to-noise ratio is too low when 100mW, and the performance of two algorithms cannot all reach demand.Fig. 7,8 are authorized user's transmission power 200mW/ The performance of two algorithms when 300mW, when false alarm rate is 0.1, verification and measurement ratio of the invention improves 6% to 7%, Fig. 9 and shows awarding Power user emission power is larger, i.e., when receiving end noise is relatively high, the performance of two algorithms is closer to.This all demonstrates this hair It is bright noise is relatively low meet actual requirement in the case where compared with SVM algorithm performance preferably, have preferable practicability and feasibility, The present invention can be used to solve the problems, such as the frequency spectrum perception under cognition wireless network.

Claims (5)

1. a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm, which comprises the following steps:
Step 1: in cognition wireless network, secondary user detects the energy value in present channel environment and result is sent To one user as fusion center；
Step 2: the characteristic of data set is sampled and established using the method based on energy measuring, primary user is by phased manner certainly Whether body, which occupies frequency spectrum resource, is sent to fusion center and constructs the label segment of data, and this completes the structures of data set It builds；
Step 3: in fusion center, the training dataset of building is fed for the training of Lightgbm algorithm and obtains model, The specifically used unilateral sampling technique and exclusive feature binding technology based on gradient of Lightgbm algorithm；
Step 4: secondary user energy sensing and is transferred to fusion center again, as test vector；
Step 5: fusion center differentiates whether channel can be used, and all secondary users are reinformed after obtaining result；
Step 6: periodic wake front end perceptron repeats step 1 to step 5, if the repetition period, which is less than, branches to step 4.
2. a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm according to claim 1, special Sign is, in time user and primary user device all can include signal transmitter and receiver in the step 1, while secondary user is also Include front end awareness apparatus.
3. a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm according to claim 1, special Sign is, the step 3 specifically:
3.1: data label setting:
M time users are contained in cognitive wireless network system, are denoted as m=1 ..., M；N number of primary user is contained, n=is denoted as 1,...,N；One of them user is as fusion center；WithRespectively represent m-th user and n-th primary user Geographical location two-dimensional coordinate；
Use KnThe state for representing the busy channel resource of n-th of primary user, if Kn=1 to represent primary user online, current to occupy frequency Spectrum resource, secondary user cannot interfere it using frequency spectrum resource at this time, if Kn=0 represents that primary user is offline, and current primary user does not have There is occupancy frequency spectrum resource, frequency spectrum resource can be used in secondary user at this time；So K=(K1,...,KN)TIt is to represent all primary user's works The vector for making state is indicated by the mode of binary hypothesis test:
3.2: normalized energy is horizontal:
The front end Energy-aware equipment of each user samples w τ complex baseband signal sample in time τ, and bandwidth is expressed as w；Energy YmIt is horizontal to represent the normalized energy that time user m is received, indicates are as follows:
Wherein η be noise power spectral density, be defined as η=E [| Nm(i)|2], Nm(i) it represents time user m and receives thermal noise, E Indicate that time user m receives the expectation of noisy samples；Xm(i) i-th of sample of signal that time user m is received is represented, binary is false If inspection is expressed as:
In formula, H0Indicate that primary user does not have online, H1Indicate that at least one primary user is online；hm,nRepresent the letter between PUn and SUm Road gain；Wn(i) the transmitting signal of primary user n is represented；
3.3: construct the set of eigenvectors of energy:
M time users are contained in our cognitive wireless network system, receive energy YmThe vector of composition is expressed as:
Y=(Y1,...,YM)T (4)
Since under the operating mode of primary user, each energy value YmObey non-central chi square distribution, freedom degree and non-central ginseng Number is as follows:
R=2w τ (5)
In above formulaIt is the fixed transmission power of primary user n, gm,n=| hm,n|2It is power attenuation, meter It is as follows to calculate formula
Here | | | | represent Euclidean distance, PL (dist)=dist-θIt represents and is damaged about the path of distance dist and loss coefficient θ It loses；Here primary user and secondary user meet 802.22 agreements, meanwhile, νm,nAnd ψm,nIt respectively represents multipath fading and shade declines It falls, and the fading coefficients ν in detecting period sectionm,nAnd ψm,nTo be quasi-static, as 1；
When there is multiple samples, energy Distribution value Gaussian distributed；Therefore energy vectors can be mentioned from multivariate Gaussian distribution It takes, mean value and variance are as follows:
μYm=r+ ζm (8)
Therefore the mean vector of energy vectors and covariance matrix are as follows:
4. a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm according to claim 1, special Sign is, the unilateral sampling technique based on gradient specifically: first data instance is classified to the absolute value of gradient, and is selected A × 100% therein, then stochastical sampling b × 100% in remaining data, to having adopted when calculating information gain The small gradient example data of sample multiplied by
5. a kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm according to claim 1, special Sign is, the exclusive feature binding technology specifically: bundled using exclusive feature as an individual feature, thus special Sign histogram becomes exclusive feature histogram.
CN201910618633.1A 2019-07-10 2019-07-10 A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm Pending CN110365435A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910618633.1A CN110365435A (en) 2019-07-10 2019-07-10 A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910618633.1A CN110365435A (en) 2019-07-10 2019-07-10 A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm

Publications (1)

Publication Number Publication Date
CN110365435A true CN110365435A (en) 2019-10-22

Family

ID=68218455

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910618633.1A Pending CN110365435A (en) 2019-07-10 2019-07-10 A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm

Country Status (1)

Country Link
CN (1) CN110365435A (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103763043A (en) * 2013-12-06 2014-04-30 镇江坤泉电子科技有限公司 Efficient radio spectrum sensing method based on collaborative cognitive network
US9730258B2 (en) * 2014-11-18 2017-08-08 Intel Corporation Apparatus configured to report aperiodic channel state information for dual connectivity

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103763043A (en) * 2013-12-06 2014-04-30 镇江坤泉电子科技有限公司 Efficient radio spectrum sensing method based on collaborative cognitive network
US9730258B2 (en) * 2014-11-18 2017-08-08 Intel Corporation Apparatus configured to report aperiodic channel state information for dual connectivity

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
GUOLIN KE 等: "LightGBM: A Highly Efficient Gradient Boosting Decision Tree", 《31ST CONFERENCE ON NEURAL INFORMATION PROCESSING SYSTEMS》 *
KARAPUTUGALA MADUSHAN THILINA 等: "Machine Learning Techniques for Cooperative", 《IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS》 *
曾娟 等: "基于双门限的新型协作频谱感知算法", 《计算机应用》 *

Similar Documents

Publication Publication Date Title
CN110365612B (en) Deep learning beam domain channel estimation method based on approximate message transfer algorithm
Aldossari et al. Machine learning for wireless communication channel modeling: An overview
Zou et al. Freedetector: Device-free occupancy detection with commodity wifi
Sobron et al. Device-free people counting in IoT environments: New insights, results, and open challenges
CN110197127A (en) Wireless signal detection and electromagnetic interference categorizing system and method based on deep learning
CN106059972A (en) Modulation identification method under MIMO related channel based on machine learning algorithm
Zhang et al. Device-free wireless localization and activity recognition with deep learning
Tsai et al. Refined autoencoder-based CSI hidden feature extraction for indoor spot localization
US20230368065A1 (en) Integrated Sensing and Communications Empowered by Networked Hybrid Quantum-Classical Machine Learning
CN114268388A (en) Channel estimation method based on improved GAN network in large-scale MIMO
CN110460359A (en) A kind of mimo system signal acceptance method neural network based
WO2023219177A1 (en) Integrated sensing and communications empowered by networked hybrid quantum-classical machine learning
CN111817803A (en) Frequency spectrum sensing method and system based on correlation coefficient and K-means clustering algorithm and computer readable storage medium
Bouchenak et al. A semi-supervised modulation identification in MIMO systems: A deep learning strategy
CN113037668B (en) Millimeter wave point-to-point communication channel equalization method
Zhuang et al. Centralized spectrum sensing based on covariance matrix decomposition and particle swarm clustering
CN113242066B (en) Multi-cell large-scale MIMO communication intelligent power distribution method
Du et al. SVM-assisted adaptive kernel power density clustering algorithm for millimeter wave channels
Gu et al. Device‐Free Human Activity Recognition Based on Dual‐Channel Transformer Using WiFi Signals
Ning et al. Wireless channel scene recognition method based on an autocorrelation function and deep learning
CN110365435A (en) A kind of cooperation energy measuring frequency spectrum sensing method based on Lightgbm algorithm
CN116680608A (en) Signal modulation identification method based on complex graph convolutional neural network
Mei et al. Cooperative automatic modulation recognition in cognitive radio
Sun et al. Environment information-based channel prediction method assisted by graph neural network
Janu et al. Performance comparison of machine learning based multi-antenna cooperative spectrum sensing algorithms under multi-path fading scenario

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20191022

RJ01 Rejection of invention patent application after publication