CN110311744A - A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm - Google Patents
A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm Download PDFInfo
- Publication number
- CN110311744A CN110311744A CN201910618558.9A CN201910618558A CN110311744A CN 110311744 A CN110311744 A CN 110311744A CN 201910618558 A CN201910618558 A CN 201910618558A CN 110311744 A CN110311744 A CN 110311744A
- Authority
- CN
- China
- Prior art keywords
- user
- energy
- fusion center
- catboost
- channel
- 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
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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/14—Spectrum sharing arrangements between different networks
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention discloses a kind of the channel circumstance adaptive spectrum cognitive method based on Catboost algorithm, specific steps are as follows: 1, secondary user acquires the energy value in present channel environment, and be dealt into as one of fusion center user；2, primary user sends fusion center for occupied channel resource situation by phased manner；3, the information structuring that fusion center sends primary user and secondary user is at data set, and further construction feature vector set；4, fusion center Catboost algorithm training pattern；5, secondary user continues to send fusion center for energy value, as test vector and inputs into training pattern；6, fusion center obtain after result will whether available channel resources are sent to all secondary users, secondary user makes a response according to the judgement of fusion center；For the present invention in the case where use meets false alarm rate 0.1, verification and measurement ratio promotes 10% compared to SVM, while misclassification rate, misclassification risk are also remarkably decreased.
Description
Technical field
The invention belongs to wirelessly communicate and field of artificial intelligence.More particularly to a kind of letter based on Catboost algorithm
Channel environment self-adaption frequency spectrum sensing method.
Background technique
Wireless sensor network is a kind of wireless Ad Hoc and data-centered network, be calculated by largely having and
The micro sensing node of communication capacity forms.The characteristics of due to wireless sensor network low overhead low-power consumption, in industrial agricultural
Equal fields are applied.However as the development of wireless communication technique, frequency spectrum resource is rare to become current wireless sensor network face
Face maximum challenge.Wireless sensor network frequency range primarily now is 2.4GHz, in industrial circle such as wireless HART WIA-PA and
ISA100.11 is the IEEE 802.15.4 based on physical layer.Widely used short-range communication technique as ZigBee, bluetooth and
Wifi is work in this frequency range.Therefore a large amount of uses of this frequency range cause channel congestion and inevitably interference.Cause
This, cognition wireless network technology, which is suggested, solves the problems, such as that frequency spectrum resource is rare.Cognition wireless network technology is intelligent nothing
Line communication technology can perceive its periphery electromagnetic environment and therefrom learn, and then make under changing for electromagnetic environment to itself
The adjustment of running parameter state.Frequency spectrum perception is its basic mode for detecting primary user's signal.OPAwe et al. proposes to use SVM
Algorithm in sample covariance matrix for carrying out the frequency spectrum sensing method of eigenvalue estimate, wherein perception user equipment is more days
Line equipment.Author, which also proposed, a kind of carries out frequency spectrum perception using Kalman filter channel estimation under slow fading channel
Method.Had a large amount of research and work in response to this problem, W.Zhang et al. focus on solving multipath fading, shadow fading and
Hidden terminal problem, major way include the cognitive method of different cooperation frequency spectrum sensing method and optimization.In addition,
Umebayashi et al. propose using the thresholding set-up mode of optimization go improve detection performance, but its needs before state elder generation
Test information.B.L.Mark et al. is mainly estimated the position of primary user's transmitter using different assessment algorithms and then determined primary
The utilization rate in terms of spatial relation and then raising frequency spectrum resource space between family and secondary user.C.Liu et al. proposes to turn
Change primary user's signal detecting mode and improves space utilization rate using central symmetry feature.Gradually had using engineering in recent years
It practises algorithm and solves the problems, such as the thinking of cognition wireless network frequency spectrum perception, and be made that some research achievements.Such as using based on energy
Detection method SVM algorithm is measured, has preferable performance in frequency spectrum perception classification problem compared to other machines learning algorithm,
It is due to high classification accuracy so very popular and practical.But these work all improve frequency spectrum perception to a certain extent
Verification and measurement ratio, but it is also in need continue to improve the place promoted, verification and measurement ratio first can also continue to improve, secondly misclassification rate and
Misclassification relative risk is also required to continue to improve.The present invention aiming at the problem that be exactly to carry out in this context.
Summary of the invention
To solve the above problems, the present invention provides a kind of channel circumstance adaptive spectrum perception based on Catboost algorithm
Method, specifically includes the following steps:
Step 1: the energy value in secondary user front end energy acquisition equipment acquisition present channel environment, and will be in the perception period
Energy value be dealt into as one of fusion center user.Energy detection method is employed herein to count channel circumstance
It calculates, principle is when primary user is online or offline, and the secondary collected energy value of user front end energy acquisition equipment is counting
Difference is had in feature, energy measuring method does not need prior information, will not need to be directed to every kind of signal as matching matrix
Special receiving device is needed, also unlike circulation spectral technology needs big calculation amount.
Step 2: primary user sends fusion center for occupied channel resource situation by phased manner.
Step 3: the information structuring that fusion center sends primary user and time user is at data set, and further construction feature
Vector set.Use this machine learning algorithm of Catboost and energy measuring method as basic perceptive mode, so naturally
Ground is expected using the collected energy value of secondary user as feature.
Step 4: fusion center Catboost algorithm training pattern.
Step 5: secondary user continues to send fusion center for energy value, as test vector and inputs into having trained mould
Type.
Step 6: fusion center obtain after result will whether available channel resources are sent to all secondary users, secondary user according to
The judgement of fusion center is made a response.
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.Present invention employs
More actual non-equilibrium sample, ratio 7:1；And it is carried out before based on machine learning algorithm positive and negative in the work of frequency spectrum perception
Sample is more balanced.
Above-mentioned steps 3 specifically:
3.1: data setting and energy normalized:
Use energy measuring method as frequency spectrum perception basic means, include in systems P primary user, be denoted as p=1,
2 ... P and Q secondary user q=1,2 ... Q, in order to consider more general model, so P is at least more than or equal to 2, Q
It is more to can choose quantity.It is assumed that primary user and time user sharing band resource will not interfere simultaneously, it is primary in systems
There are two types of the working conditions at family: online Sp=1 or offline Sp=0；Which occupies frequency spectrum resource, secondary use when primary user is online
Family cannot use；It discharges frequency spectrum resource when primary user is offline, and frequency spectrum resource can be used in secondary user at this time；As long as having in system
One primary user occupies frequency spectrum resource, then regarding as time user does not allow to reuse frequency spectrum resource；Use gpRepresent the geography of PU
Position, gqRepresent the geographical location of SU.
The energy detector of each SU samples w τ complex baseband signal sample in detecting period period tau, and bandwidth is expressed as
w；Rq(i) i-th of sample of signal that SU is received is represented, is indicated with following formula:
H herein0Representing does not have PU in channel, so the only thermal noise that SU awareness apparatus receives, uses Nq(i) it indicates；H1
Represent at least one PU it is online when the case where, Wp(i) the transmitting sample of signal of PUp, h are representedp,qIt represents between PUp and SUq
Channel gain, SpThe as working condition of PU.In addition, SU should make correct decisions in detecting period section.The present invention is base
In the energy measuring frequency spectrum sensing method of machine learning algorithm, in machine learning, extraction is characterized in the first step.Use YqIt represents
The normalized energy that SUq is received is horizontal:
Herein η be noise power spectral density be defined as η=E [| Nq(i)|2]；Therefore, energy vectors are received comprising all SU
The energy level arrived:
Y=(Y1,...,YQ)T (3)
3.2 after obtaining energy vectors, further analyze its distribution；
Because of the operating mode of PU, each energy value YqObey non-central chi square distribution, freedom degree and non-centrality parameter
It is as follows:
R=2w τ (4)
HereIt is the fixed transmission power of PUp, lp,q=| hp,q|2It is power attenuation, calculates
Formula is as follows:
lp,q=PL (| | gq-gp||).νp,qψp,q (6)
Here | | | | represent Euclidean distance, PL (dist)=dist-θRepresent the road about distance dist and loss coefficient θ
Diameter loss；νp,qAnd ψp,qRespectively represent multipath fading and shadow fading；PU and SU meets 802.22 agreements；In addition, in perception
Between fading coefficients ν in sectionp,qAnd νp,qIt is constant to be quasi-static, as 1；
We describe in front for the distribution of energy level, such as w τ, energy Distribution value when there is enough samples
Basic Gaussian distributed；Therefore energy vectors can be extracted from multivariate Gaussian distribution, and mean value and variance are as follows:
μYq=r+ ζq (7)
σ2 Yq=2 (r+2 ζq) (8)
Therefore the mean vector of energy vectors and covariance matrix are as follows:
μYq=(μY1,...,μYQ)T (9)
Above-mentioned Catboost algorithm specifically: within the perception period, secondary user sends energy value in the channel perceived
To fusion center as characteristic energy vector, primary user's discontinuity sends information whether occupying frequency spectrum resource and makees to fusion center
For label, the building of training dataset is completed in this way.Catboost algorithm training pattern is used in fusion center, is presently described
Once Catboost algorithm: Catboost algorithm proposes by Yandex, the algorithm optimization processing of category feature, and be
Training stage handles rather than data preprocessing phase, and another advantage of the algorithm is that a kind of new method has been used to select to set mould
Leaf node value is calculated when type, this help reduces over-fitting.Catboost algorithm has two kinds of operation methods of CPU and GPU,
GPU method is also faster than the GPU mode of current most popular Xgboost, and cpu mode is also such.Catboost algorithm and
Gradient method for improving is the same, and the new tree of building goes the residual error of fitting "current" model.However traditional gradient method for improving all can
The influence for receiving point gradient estimation partially is easy to cause over-fitting.Gradient, such model are assessed using identical data point every time
It is established.This will lead to compared to real gradient distribution space, and the feature space distribution of gradient to be fitted can deviate, meeting in this way
Lead to over-fitting.The method that many GBDT methods (such as Xgboost) construct next tree mainly includes two steps: selection tree framework and
The value of leaf is set.In order to select optimal table structure, algorithm will do it different fractionations enumerate, construct tree, setting leaf values,
It scores and selects divisional mode.Fitting gradient can be all calculated in the two stage leaf values.Catboost is in second stage
It is identical with conventional method, but the first stage has used improved method.Use FkRepresent first k tree in building, gk(Xh,Zh) generation
The model of table training sample building k tree gradient value building at h-th.In order to make gradient gk(Xh,Zh) unbiased, it would be desirable to do not having
There is XhLower training pattern Fk, implement training process can not, consider following skill to solve this problem: for each time
Realize Xh, we train a model Mh, go to update without the mode that gradient is estimated.Use MhIn XhOn the basis of be fitted gradient, make
It is scored in this way.Catboost produces s random alignment in training set, samples to come using various arrangement
The gradient of residual error is obtained to strengthen the robustness of algorithm, carrys out training pattern using different arrangements, then using various arrangement
Avoid over-fitting.For each arrangement σ, n model M of trainingk.This means that needing to store and count again when building is newly set
Calculate O (n2) come be fitted arrangement σ, for each model Mk, need to update Mk(X1),...,Mk(Xk), so computation complexity is O
(sn2), an important skill is used during realization the time complexity of each tree building is dropped to O (sn): for
Each arrangement is not storage and renewal time complexity O (n2) value Mk(Xj), but retention value Mk'(Xj), k=
1,...,[log2(n)], j < 2k+1, M herek(Xj) it is based on first 2kThe fitting of the sample j of sample is approximate.Then, it predicts
Mk(Xj) can be lower thanIn XhOn gradient be used to select tree construction.
Compared with prior art, beneficial effects of the present invention:
Using false alarm rate 0.1 for meeting the requirement of IEEE 802.11, verification and measurement ratio is promoted the present invention compared to SVM
10%, whole classification performance is better than SVM, while misclassification rate, misclassification risk are also remarkably decreased, and is furthermore also learnt by experiment
Inventive can be also substantially strong compared with SVM algorithm under different state of signal-to-noise by primary user's transmission power difference i.e. time user, i.e., not
Machine learning model can have stronger practicability, even if in noise with retraining to adapt to present channel under cochannel environment
Performance is also preferable in the case where relatively low, and robustness is stronger.It is significant for frequency spectrum perception application.
Detailed description of the invention
Fig. 1 is the frequency spectrum sensing method block diagram the present invention is based on machine learning.
Fig. 2 is 7*7 cooperative spectrum sensing system structure model.
Fig. 3 is the ROC curve of Catboost algorithm and SVM algorithm (linear kernel function) in 7*7 model.
Specific embodiment
Implementation of the invention is further described with reference to the accompanying drawing.
The present invention is based on the channel circumstance adaptive spectrum cognitive methods of Catboost algorithm, and system structural framework figure is such as
Shown in attached drawing 1.
2 illustrate the cooperative spectrum sensing model in the present invention based on geographical location with reference to the accompanying drawing, illustrate this in conjunction with Fig. 3
Invention compared to the SVM algorithm to behave oneself best before this under PU different transmission power ROC curve, verification and measurement ratio, misclassification risk with
And the performance of misclassification rate.Emulation is with python3.6.2 in 64 PC, memory RAM 16G, six core i7 (3.2GHz) rings
It is carried out under border.
The performance indicator that the present invention compares is as follows:
A) 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.
B) 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.
C) 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.
D) 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.
Embodiment
In order to verify the available of the frequency spectrum perception solved the problems, such as under cognition wireless network the present invention is based on Catboost algorithm
Property and feasibility, carried out emulation experiment and with SVM algorithm carry out algorithm performance compared with.When simulation parameter is provided that perception
Between section τ be 100 μ s, bandwidth 5MHz, noise power spectral density be -174dBm, each PU transmission power be 200mW, path
Loss coefficient is 4, and multipath fading and shadow fading coefficient are all 1, and the kernel function that each PU online probability is 0.5.SVM is selected
It is selected as linear kernel function, because having been proven that linear kernel function in the outstanding performance of this problem in the work of early period.Training
Vector is that 160 test vectors are 640.Positive and negative sample proportion is 7*7 cooperative spectrum sensing system structure of the 7:1. in attached drawing 2
In model, have 49 SU, be evenly distributed in the grid of 7*7, have 3 PU respectively (- 1100m, -1000m), (750m,
890m), the position (1500m, -1000m).ROC curve can see in fig. 3, mark SVM algorithm with solid line, with dotted line mark
Catboost algorithm is gone out, when false alarm rate is 0.1, Catboost ratio SVM verification and measurement ratio improves 10%, and whole classification performance is excellent
In SVM.
Compared to SVM algorithm under the different transmission power of PU, the present invention have higher verification and measurement ratio, misclassification risk and
Misclassification rate is specifically as shown in table 1:
1 Catboost algorithm of table and SVM algorithm performance table in the case that primary user's transmission power is different in 7*7 model
Existing index
It can all be promoted as signal-to-noise ratio improves two algorithm indexs, when noise is relatively high, Catboost reaches convergence.This
Feasibility and availability of the invention are all demonstrated, the frequency spectrum perception that the present invention is used to solve under cognition wireless network can be asked
Topic.
Claims (4)
1. a kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm, which is characterized in that including following step
It is rapid:
Step 1: the energy value in secondary user front end energy acquisition equipment acquisition present channel environment, and the energy in the period will be perceived
Magnitude is dealt into one user as fusion center；
Step 2: primary user sends fusion center for occupied channel resource situation by phased manner；
Step 3: the information structuring that fusion center sends primary user and time user is at data set, and further construction feature vector
Collection；
Step 4: fusion center Catboost algorithm training pattern；
Step 5: secondary user continues to send fusion center for energy value, as test vector and inputs into training pattern；
Step 6: fusion center obtain after result will whether available channel resources are sent to all secondary users, secondary user is according to fusion
The judgement at center is made a response.
2. a kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm according to claim 1,
It is characterized in that, non-equilibrium sample, ratio 7:1 is used in the step 1；And frequency spectrum is carried out based on machine learning algorithm before
Positive negative sample is balanced in the work of perception.
3. a kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm according to claim 1,
It is characterized in that, the step 3 specifically:
3.1: data setting and energy normalized:
Use energy measuring method as frequency spectrum perception basic means, includes in systems P primary user, be denoted as p=1,2 ... P
With Q user q=1,2 ... Q；Primary user and time user sharing band resource will not interfere simultaneously herein, be
In system there are two types of the working conditions of primary user: online Sp=1 or offline Sp=0；Which occupies frequency spectrum moneys when primary user is online
Source, secondary user cannot use；It discharges frequency spectrum resource when primary user is offline, and frequency spectrum resource can be used in secondary user at this time；System
There is a primary user to occupy frequency spectrum resource in as long as, then regarding as time user does not allow to reuse frequency spectrum resource；Use gpIt represents
The geographical location of PU, gqRepresent the geographical location of SU；
The energy detector of each SU samples w τ complex baseband signal sample in detecting period period tau, and bandwidth is expressed as w；Rq
(i) i-th of sample of signal that SU is received is represented, is indicated with following formula:
H herein0Representing does not have PU in channel, so the only thermal noise that SU awareness apparatus receives, uses Nq(i) it indicates；H1It represents
The case where when at least one PU is online, Wp(i) the transmitting sample of signal of PUp, h are representedp,qRepresent the channel between PUp and SUq
Gain, SpThe as working condition of PU；Use YqIt is horizontal to represent the normalized energy that SUq is received:
Herein η be noise power spectral density be defined as η=E [| Nq(i)|2]；Therefore, energy vectors include what all SU were received
Energy level:
Y=(Y1,...,YQ)T (3)
3.2 after obtaining energy vectors, further analyze its distribution；
Because of the operating mode of PU, each energy value YqNon-central chi square distribution is obeyed, freedom degree and non-centrality parameter are as follows:
R=2w τ (4)
HereIt is the fixed transmission power of PUp, lp,q=| hp,q|2It is power attenuation, calculation formula
It is as follows:
lp,q=PL (| | gq-gp||).νp,qψp,q (6)
Here | | | | represent Euclidean distance, PL (dist)=dist-θIt represents and is damaged about the path of distance dist and loss coefficient θ
It loses；νp,qAnd ψp,qRespectively represent multipath fading and shadow fading；PU and SU meets 802.22 agreements；In addition, in detecting period section
Interior fading coefficients νp,qAnd νp,qIt is constant to be quasi-static, as 1；
Reach w τ, energy Distribution value Gaussian distributed in sample size；Therefore energy vectors can be from multivariate Gaussian distribution
It extracts, mean value and variance are as follows:
μYq=r+ ζq (7)
σ2 Yq=2 (r+2 ζq) (8)
Therefore the mean vector of energy vectors and covariance matrix are as follows:
4. a kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm according to claim 1,
It is characterized in that, the Catboost algorithm training pattern specifically:
The second stage of Catboost is identical with conventional method, and the first stage has used improved method: using FkIt represents the of building
One k tree, gk(Xh,Zh) represent the model constructed in h-th of training sample building k tree gradient value；In order to make gk(Xh,Zh) nothing
Partially, for realizing X each timeh, we train a model Mh, go to update without the mode that gradient is estimated；Use MhIn XhBasis
Upper fitting gradient makes to be scored in this way；Catboost produces s random alignment in training set, uses
Various arrangement sampling strengthens the robustness of algorithm to obtain the gradient of residual error, carrys out training pattern using different arrangements, then
Over-fitting is avoided using various arrangement；
For each arrangement σ, n model M of trainingk, store when constructing new tree and recalculate O (n2) come be fitted arrangement σ,
For each model Mk, need to update Mk(X1),...,Mk(Xk), so computation complexity is O (sn2), make during realization
The time complexity of each tree building is dropped to O (sn) with following method: not being storage and update for each arrangement
Time complexity O (n2) value Mk(Xj), but retention value Mk'(Xj), k=1 ..., [log2(n)], j < 2k+1, M herek(Xj)
It is based on first 2kThe fitting of the sample j of sample is approximate；Then, the M of predictionk(Xj) can be lower than
In XhOn gradient be used to select tree construction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618558.9A CN110311744A (en) | 2019-07-10 | 2019-07-10 | A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910618558.9A CN110311744A (en) | 2019-07-10 | 2019-07-10 | A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110311744A true CN110311744A (en) | 2019-10-08 |
Family
ID=68080937
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910618558.9A Pending CN110311744A (en) | 2019-07-10 | 2019-07-10 | A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110311744A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111313971A (en) * | 2020-02-28 | 2020-06-19 | 杭州电子科技大学 | Lightgbm equalization system and method for improving IMDD short-distance optical communication system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9730258B2 (en) * | 2014-11-18 | 2017-08-08 | Intel Corporation | Apparatus configured to report aperiodic channel state information for dual connectivity |
| CN107517089A (en) * | 2017-09-06 | 2017-12-26 | 南京邮电大学 | A kind of broader frequency spectrum cognitive method based on spectrum signature |
-
2019
- 2019-07-10 CN CN201910618558.9A patent/CN110311744A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9730258B2 (en) * | 2014-11-18 | 2017-08-08 | Intel Corporation | Apparatus configured to report aperiodic channel state information for dual connectivity |
| CN107517089A (en) * | 2017-09-06 | 2017-12-26 | 南京邮电大学 | A kind of broader frequency spectrum cognitive method based on spectrum signature |
Non-Patent Citations (2)
| Title |
|---|
| DOROGUSH 等: "CatBoost : gradient boosting with categorical features support", 《LEARNINGSYS.ORG》 * |
| KARAPUTUGALA MADUSHAN THILINA 等: "Machine Learning Techniques for Cooperative Spectrum Sensing in Cognitive Radio Networks", 《IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111313971A (en) * | 2020-02-28 | 2020-06-19 | 杭州电子科技大学 | Lightgbm equalization system and method for improving IMDD short-distance optical communication system |
| CN111313971B (en) * | 2020-02-28 | 2021-06-22 | 杭州电子科技大学 | Lightgbm equalization system and method for improving IMDD short-distance optical communication system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102546059B (en) | Non-supervision clustering-based distributed cooperative spectrum sensing method for cognitive self-organizing network | |
| EP4136910A1 (en) | Architecture for machine learning (ml) assisted communications networks | |
| CN105959987A (en) | Data fusion algorithm for improving energy utilization rate and service performance of wireless sensor network | |
| CN110166154A (en) | A kind of software radio spectrum monitoring knowledge method for distinguishing neural network based | |
| US20230209419A1 (en) | Machine learning handover prediction based on sensor data from wireless device | |
| WO2022047410A2 (en) | Unsupervised learning for simultaneous localization and mapping in deep neural networks using channel state information | |
| Chen et al. | Smartphone-based indoor fingerprinting localization using channel state information | |
| CN107333294A (en) | A kind of combination AdaBoost and SVMs link quality prediction method | |
| Zhang et al. | Machine learning techniques for spectrum sensing when primary user has multiple transmit powers | |
| CN110460401A (en) | A kind of cooperative frequency spectrum sensing method clustered based on matrix decomposition and particle group optimizing | |
| CN108242961A (en) | Full duplex cognitive radio networks cooperation frequency spectrum sensing method based on integrated study | |
| CN106992822A (en) | A kind of localization method of the blind node of wireless sensor network | |
| CN113365323A (en) | Low-energy-consumption multipath transmission method for wireless sensor network | |
| WO2021226109A1 (en) | Estimating features of a radio frequency band based on an inter-band reference signal | |
| Ahmed et al. | Deep residual learning-based cognitive model for detection and classification of transmitted signal patterns in 5G smart city networks | |
| Chagas et al. | Genetic algorithms and simulated annealing optimization methods in wireless sensor networks localization using artificial neural networks | |
| CN106257849B (en) | Frequency spectrum sensing method based on multi-target quantum firefly search mechanisms | |
| CN108924847A (en) | A kind of cognitive radio frequency spectrum prediction technique and equipment based on ANN | |
| CN109541537B (en) | Universal indoor positioning method based on ranging | |
| CN111417124A (en) | Method for sensing frequency spectrum in cognitive wireless network environment | |
| CN110311744A (en) | A kind of channel circumstance adaptive spectrum cognitive method based on Catboost algorithm | |
| CN111934797B (en) | Collaborative spectrum sensing method based on covariance eigenvalue and mean shift clustering | |
| CN116578932A (en) | Random forest based method for evaluating link quality of unmanned cooperative network | |
| CN116437365A (en) | Wireless sensor network link quality evaluation method and system | |
| XU et al. | Multi-task learning with dynamic splitting for open-set wireless signal recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191008 |
|
| RJ01 | Rejection of invention patent application after publication |