CN108924847A - A kind of cognitive radio frequency spectrum prediction technique and equipment based on ANN - Google Patents
A kind of cognitive radio frequency spectrum prediction technique and equipment based on ANN Download PDFInfo
- Publication number
- CN108924847A CN108924847A CN201810562684.2A CN201810562684A CN108924847A CN 108924847 A CN108924847 A CN 108924847A CN 201810562684 A CN201810562684 A CN 201810562684A CN 108924847 A CN108924847 A CN 108924847A
- Authority
- CN
- China
- Prior art keywords
- sample
- spectrum
- primary user
- signal
- frequency spectrum
- 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.)
- Granted
Links
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W24/00—Supervisory, monitoring or testing arrangements
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention discloses frequency spectrum prediction technique and equipment in a kind of cognitive radio based on ANN, belong to cognitive radio technology field in communication, and this method includes:Cyclic Spectrum characteristic parameter relevant to modulation system is extracted to from each primary user's signal that user perceives;The characteristic ginseng value of extraction is identified to the modulation system of each primary user by trained ANN classification model;According to the characteristic of different modulation system and Cyclic Spectrum, the largest loop spectrum of signal is calculated;The energy of signal and largest loop spectrum are judged that primary user's signal whether there is by ANN detection.Compared with prior art, the present invention is modulated mode to multiple primary user's signals first and identifies, identify that different modulation systems can effectively cope with the interference of noise, the characteristic detected again based on energy and Cyclic Spectrum predicts primary user's frequency spectrum, noise jamming can be successfully managed, precision of prediction is improved.
Description
Technical field
The present invention relates to a kind of spectrum prediction method in cognitive radio based on artificial neural network (ANN) and set
It is standby, belong to cognitive radio technology field.
Background technique
Cognitive radio is to allow to be used from authorization frequency spectrum resource of the user to primary user's temporary standby, is not being interfered
Under the premise of primary user communicates, the utilization rate of existing frequency spectrum resource is improved as far as possible.Currently, the resource of radio-frequency spectrum is mainly adopted
With fixed spectrum allocation strategy and way to manage, under this allocation strategy, there is the feelings that frequency spectrum resource utilization rate is relatively low
Condition:On the one hand authorization frequency spectrum occupies the most of of entire Radio Spectrum Resource, and in some cases, primary user can't be every
When it is per quarter all in occupation of authorization frequency spectrum.Therefore many authorization frequency spectrums are had and are in idle state.Cause the availability of frequency spectrum low
Under.The it is proposed of cognitive radio technology is exactly to improve the low problem of the availability of frequency spectrum.
The first step of cognitive radio technology is exactly to capture the relevant information of spectral change.Cognitive radio needs perceive
The electromagnetic signature of ambient enviroment, and intelligent decision, the transmitting and reception ginseng of its equipment of adjust automatically are carried out according to the result of perception
Number.Main four functions of cognitive radio are:Frequency spectrum perception, frequency spectrum decision, frequency spectrum share, frequency spectrum switching.This four functions
A cognition ring in cognition wireless is constituted, wherein most important function is frequency spectrum perception in cognition wireless, i.e.,:From user couple
Frequency spectrum is perceived and is detected, and the relevant information of capture frequency spectrum finds the spectrum interposition for being possible to that communication is established on frequency spectrum.One
As for, cognitive radio frequency spectrum cognition technology can be divided into:Based on transmitting machine testing, cooperative detection, based on the detection of interference
With the detection based on receiver.Wherein the detection based on transmitter can be specifically subdivided into again:Energy measuring, matched filtering inspection
Survey, Cyclic Spectrum detection etc..
Energy measuring method in frequency spectrum perception is relatively simple and flexible, and the priori for not needing primary user's signal is known
Know, and become the preference algorithm of frequency spectrum detection algorithm, energy detection algorithm is a kind of noncoherent detection, it is necessary first to detect hair
The signal of sending end will first pass through a bandpass filter and carry out smothing filtering.Then pass through energy meter to putting down smoothed out signal
The formula of calculation calculates the energy of the signal, then by decision rule, if calculated energy value is greater than the threshold value of setting,
It is considered that primary user is existing, if the energy value calculated is lower than the threshold value of setting, cognition primary user is to be not present
's.But this method be measure a certain frequency band energy be foundation, thus when signal signal-to-noise ratio it is low-down when, as a result can
Become very unreliable.Thus can occur the very high situation of error rate in some cases using energy measuring method.Cyclic Spectrum is special
Property detection there is probabilistic ability in stronger resistance noise power, and do not need the priori knowledge of primary user's signal, because
And signal cognitive ability with higher.But the method needs to carry out Fourier transformation twice, computational complexity is excessive, can lead
Cause detection time too long.
Summary of the invention
Goal of the invention:In view of the above shortcomings of the prior art, the cognition nothing based on ANN that the object of the present invention is to provide a kind of
Line electricity spectrum prediction method and apparatus, this method are primarily based on circulation spectral property, predict the modulation system of primary user, then based on biography
The characteristic of Cyclic Spectrum detection under the energy and different modulating mode of system predicts primary user's frequency spectrum, effectively increases prediction
Precision.
Technical solution:To achieve the above object, the technical solution adopted by the present invention is:
A kind of cognitive radio frequency spectrum prediction technique based on ANN, this method comprises the following steps:
(1) relevant to modulation system based on Cyclic Spectrum to being extracted from each primary user's signal that user perceives
Characteristic parameter, the characteristic parameter includes:Frequency spectrum is in the δ pulse number that f axis is presented, period spectral peak number, spectrum on α axis
Related coefficient and normalization maximum value;
(2) using the characteristic ginseng value of extraction as the input of first order neural network model, the tune of each primary user is identified
Mode processed;The first order neural network model is trained by the characteristic value that the sample signal of known modulation system extracts
It arrives;
(3) it according to the characteristic of the modulation system of the primary user identified and Cyclic Spectrum, obtains at three a certain cycle frequencies
Largest loop spectrum, and combine the characteristic value that identifies as primary user of energy of primary user's signal of perception;
(4) input of the characteristic value that step (3) obtains as nervus opticus network model, it is primary according to output prediction of result
Family whether there is;The feature that the nervus opticus network model is extracted by the sample signal of known modulation system and objective result
Value is trained to obtain.
In preferred embodiments, characteristic parameter takes spectral correlative coefficient in the step (1)
Maximum value;WhereinFor the Cyclic Spectrum of random rotation stationary signal x (t),It isIn α=0
Value.
In preferred embodiments, the normalization maximum value β in the step (1) is maximum of the Cyclic Spectrum on α axis
The ratio of value and the maximum value on f axis.
In preferred embodiments, the first order neural network model in the step (2) is BP neural network model,
Model training method includes:
(2.1) initialization model parameter and desired output matrix, every layer of modified weight learning rate, minimum are missed
Difference and maximum number of iterations;
(2.2) P sample is inputted, if current input sample is p-th, starts to be trained model;
(2.3) according to Formula Input TechnologyAnd output yi=f (Ii) successively start to calculate each of each layer
Output, and network error E is found out by desired output matrixp, wherein xjFor the input data that neuron j is received, n is defeated
Enter the total number of data, θiIt is the threshold value of neuron, wijIt is the connection weight of two neurons i and j, IiIt is i-th of input, yi
It is i-th of neuron output data,For excitation function;
(2.4) according to formula δkl (p)=(dl (p)-yl (p))yl (p)(1-yl (p)) andPoint
The anti-pass error change δ of each layer is not calculatedkl (p).WhereinFirst of output data when being p-th of sample,It is p-th of sample
This hidden layer output as a result,First of desired output data when being p-th of sample,First when being -1 sample of pth
The anti-pass error of k-th of neuron when neuron and p-th of sample.K-th of neuron and p-th when being -1 sample of pth
The anti-pass error of j-th of neuron when sample;
(2.5) sample number learnt is recorded, if currently trained sample number p < P, (2.2) is returned and continues
It is calculated, when p=P, jumps into (2.6);
(2.6) according to the correction formula of weightAndThe weight of each layer of corrective networks one by one;Wherein vklIt is pth
The hidden layer of a sample exports as a result, n0For the number of iterations, wklIt is the input layer of network to the company of hidden layer neuron
Weight is connect,It is j-th of input of p-th of sample, constant η ∈ (0,1) indicates learning rate, Δ vklIt is expressed as hidden layer
The l layers of correction amount to kth layer, Δ wklIt is l layers and is input to the implicit correction amount of kth layer;
(2.7) total error E is calculatedAIf overall error EAMeet EA< ε reaches maximum number of iterations, then training knot
Otherwise beam continues the study of a new round back to (2.2);What wherein ε was indicated is desired minimal error.
In preferred embodiments, the characteristic value that primary user identifies in the step (3) include three maximums α=
2fcCirculation spectrum value at cycle frequency, wherein fcIt is carrier frequency.
In preferred embodiments, the nervus opticus network model in the step (4) is BP neural network model, mould
The iterative formula of weight is in type training:
w(n0+ 1)=w (n0)+η(n0)d(n0), wherein w (n0) indicate by n-th0When secondary iteration, each layer weight institute group of network
At matrix, n0For the number of iterations, η (n0) indicate n-th0Learning rate when secondary iteration, as overall error EAWhen reduction, study speed
Rate increases;If overall error EAWhen increase, learning rate reduces,EAFor error.
A kind of computer equipment that another aspect of the present invention provides, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, the computer program realize when being loaded on processor it is described based on
The cognitive radio frequency spectrum prediction technique of ANN.
Beneficial effect:Compared with the prior art, the advantages of the present invention are as follows:Frequency spectrum perception in previous traditional cognition wireless
Technology, which is merely able to single detection primary user's signal, whether there is, and detection method will receive the very big interference of noise, this hair
The bright identification for being modulated mode to the signal of multiple primary users first identifies that different modulation systems can be coped with effectively and makes an uproar
The interference of sound, in the process, the training basis of model are simply input sample and are trained to obtain the classifier of modulation system,
Then according to the result of this classifier output and the characteristic based on conventional recycle spectrum, extract spectrum peak at three at different frequency and
The energy value of signal as judge primary user whether there is model input sample.It, can be to primary user by trained model
Signal judged whether there is.The method can accurately judge spectrum interposition.
Detailed description of the invention
Fig. 1 is to judge the flow diagram that primary user whether there is in the embodiment of the present invention;
Fig. 2 is the flow diagram of BP neural network model training in the embodiment of the present invention;
Fig. 3 is the flow diagram of frequency spectrum prediction technique in the embodiment of the present invention.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, present invention is further described in detail.
Fig. 1 is that the present invention judges the main flow that primary user whether there is, and specific requirement is as follows:First to many primary
Family signal is modulated the identification of mode, according to the feature of Cyclic Spectrum of the different calculating signals of modulation system at a certain frequency
Value carries out the training of primary user's identification then in this, as input sample.It can be right by this model by signal to be detected
Primary user, which whether there is, to be judged.
As shown in figure 3, a kind of cognitive radio frequency spectrum prediction technique based on ANN disclosed by the embodiments of the present invention, mainly
Include the following steps:
(1) relevant to modulation system based on Cyclic Spectrum to being extracted from each primary user's signal that user perceives
Characteristic parameter.
By being pre-processed to signal in this step, so that the characteristic extraction part in Modulation Identification is from having pre-processed
Signal in extract some parameters of modulated signal for identification.Wherein the pretreatment of primary user's signal is predominantly the same as phase point
Amount, the decomposition of quadrature component and the estimation of carrier frequency, and ensure to only have individual signals to enter subsequent Modulation Identification link every time.
Extracted characteristic parameter includes pulse number, period spectral peak number, spectral correlative coefficient and the normalization maximum value that frequency spectrum is presented
Four characteristic parameters.
Cyclic Spectrum is defined as the Fourier transformation of auto-correlation function:
Wherein x (t) is a random rotation stationary signal, and α is period frequency, and T is the period distances of measurement, Rx α(τ) is the letter
Number auto-correlation function, expression formula is:
Aforementioned four characteristic parameter is specifically described as:
(a) the δ pulse number n that frequency spectrum is presented on f axis.
(b) the period spectral peak number m on α axis.
(c) spectral correlative coefficientIt is generally taken most to reduce calculation amount
Big value.
(d) normalization maximum value β is defined asIn the ratio of maximum value and maximum value on f axis on α axis,
In
(2) using the characteristic ginseng value of extraction as the input of first order neural network model, the tune of each primary user is identified
Mode processed.
In this step, the characteristic value extracted by the sample signal of known modulation system when first order neural network model into
Row training obtains.Specific neural network model is BP neural network model, and model training method is as shown in Fig. 2, specifically include:
(2.1) initialization model parameter and desired output matrix, every layer of modified weight learning rate, minimum are missed
Difference and maximum number of iterations;
(2.2) P sample is inputted, if current input sample is p-th, starts to be trained model;
(2.3) according to Formula Input TechnologyAnd output yi=f (Ii) successively start to calculate each of each layer
Output, and network error E is found out by desired output matrixp, wherein xjFor the input data that neuron j is received, n is defeated
Enter the total number of data, θiIt is the threshold value of neuron, wijIt is the connection weight of two neurons i and j, IiIt is i-th of input, yi
It is i-th of neuron output data,For excitation function;
(2.4) according to formula δkl (p)=(dl (p)-yl (p))yl (p)(1-yl (p)) andPoint
The anti-pass error change δ of each layer is not calculatedkl (p).WhereinFirst of output data when being p-th of sample,It is p-th of sample
This hidden layer output as a result,First of desired output data when being p-th of sample,First when being -1 sample of pth
The anti-pass error of k-th of neuron when neuron and p-th of sample.K-th of neuron and p-th when being -1 sample of pth
The anti-pass error of j-th of neuron when sample;
(2.5) sample number learnt is recorded, if currently trained sample number p < P, (2.2) is returned and continues
It is calculated, when p=P, jumps into (2.6);
(2.6) according to the correction formula of weightAndThe weight of each layer of corrective networks one by one;Wherein vklIt is pth
The hidden layer of a sample exports as a result, n0For the number of iterations, wklIt is the input layer of network to the company of hidden layer neuron
Weight is connect,It is j-th of input of p-th of sample, constant η ∈ (0,1) indicates learning rate, Δ vklIt is expressed as hidden layer l
Layer arrives the correction amount of kth layer, Δ wklIt is l layers and is input to the implicit correction amount of kth layer;
(2.7) total error E is calculatedAIf overall error EAMeet EA< ε reaches maximum number of iterations, then training knot
Otherwise beam continues the study of a new round back to (2.2);What wherein ε was indicated is desired minimal error.
(3) it according to the characteristic of the modulation system of the primary user identified and Cyclic Spectrum, obtains at three a certain cycle frequencies
Largest loop spectrum, and combine the characteristic value that identifies as primary user of energy of primary user's signal of perception.
In this step, according to the characteristic of the modulation system of primary user and Cyclic Spectrum, four values are chosen altogether and are known as primary user
Another characteristic vector.Specially:
(a) energy value of primary user's signal:Wherein X (w) is primary
The Fourier transformation of family signal x (t).
(b) modulation system different according to primary user's signal calculates three maximums in α=2fcCirculation at cycle frequency
Spectrum value, wherein fcIt is carrier frequency.
(4) input of the characteristic value that step (3) obtains as nervus opticus network model, it is primary according to output prediction of result
Family whether there is.
Nervus opticus network model in this step is also to be mentioned by the sample signal of known modulation system and objective result
What the characteristic value taken was trained.Specific neural network model is also BP neural network model, training method and the first mind
Similar through network model, difference is, the model refinement of the classification used in this step for the BP algorithm of adjusting learning rate,
There is relatively high requirement for learning rate with the presence or absence of link judging primary user, therefore is adopted under trained different phase
Suitable learning rate is taken to be very important for accelerating convergence rate.Therefore learning rate is a variable, mould herein
The iterative formula of weight is in type training:
w(n0+ 1)=w (n0)+η(n0)d(n0), wherein w (n0) indicate by n-th0When secondary iteration, each layer weight institute group of network
At matrix, n0For the number of iterations, η (n0) indicate n-th0Learning rate when secondary iteration, as overall error EAWhen reduction, study speed
Rate increases;If overall error EAWhen increase, learning rate reduces.
After the completion of neural network model training, four characteristic values in above-mentioned steps (3) are carried out to the signal that needs detect
Extraction, by the detection of neural network, if last output result with objective result be it is identical, indicate primary user
It is existing, if result is not identical, indicates that primary user's signal is not present.
Based on the same technical idea, the embodiment of the invention also provides a kind of computer equipments, which can
With include memory, processor and storage on a memory and the computer program that can run on a processor, the computer
Program realizes the above-mentioned cognitive radio frequency spectrum prediction technique based on ANN when being loaded on processor.
The foregoing is merely better embodiment of the invention, protection scope of the present invention is not with above embodiment
Limit, as long as those of ordinary skill in the art's equivalent modification or variation made by disclosure according to the present invention, should all be included in power
In the protection scope recorded in sharp claim.
Claims (7)
1. a kind of cognitive radio frequency spectrum prediction technique based on ANN, it is characterised in that:This method comprises the following steps:
(1) spy based on Cyclic Spectrum relevant to modulation system is extracted to from each primary user's signal that user perceives
Parameter is levied, the characteristic parameter includes:Frequency spectrum is in the δ pulse number that f axis is presented, period spectral peak number, spectrum correlation on α axis
Coefficient and normalization maximum value;
(2) using the characteristic ginseng value of extraction as the input of first order neural network model, the modulation methods of each primary user are identified
Formula;The first order neural network model is trained to obtain by the characteristic value that the sample signal of known modulation system extracts;
(3) it according to the characteristic of the modulation system of the primary user identified and Cyclic Spectrum, obtains at three a certain cycle frequencies most
Systemic circulation spectrum, and the characteristic value for combining the energy of primary user's signal of perception to identify as primary user;
(4) input of the characteristic value that step (3) obtains as nervus opticus network model, according to output, prediction of result primary user is
No presence;The characteristic value that the nervus opticus network model is extracted by the sample signal of known modulation system and objective result into
Row training obtains.
2. a kind of cognitive radio frequency spectrum prediction technique based on ANN according to claim 1, it is characterised in that:The step
(1) characteristic parameter takes the spectral correlative coefficient of the circulation spectrum signature based on primary user's signal in
Maximum value;WhereinFor the Cyclic Spectrum of random rotation stationary signal x (t),It isIn α=0
Value.
3. a kind of cognitive radio frequency spectrum prediction technique based on ANN according to claim 1, it is characterised in that:It is described
Normalization maximum value β in step (1) is ratio of the Cyclic Spectrum in maximum value and the maximum value on f axis on α axis.
4. a kind of cognitive radio frequency spectrum prediction technique based on ANN according to claim 1, it is characterised in that:It is described
First order neural network model in step (2) is BP neural network model, and model training method includes:
(2.1) initialization model parameter and desired output matrix, every layer of modified weight learning rate, minimal error with
And maximum number of iterations;
(2.2) P sample is inputted, if current input sample is p-th, starts to be trained model;
(2.3) according to Formula Input TechnologyAnd output yi=f (Ii) successively start to calculate each output of each layer,
And network error E is found out by desired output matrixp, wherein xjFor the input data that neuron j is received, n is input data
Total number, θiIt is the threshold value of neuron, wijIt is the connection weight of two neurons i and j, IiIt is i-th of input, yiIt is i-th
Neuron output data,For excitation function;
(2.4) according to formulaAndIt calculates separately each
The anti-pass error change of layerWhereinFirst of output data when being p-th of sample,It is the implicit of p-th of sample
Layer output as a result,First of desired output data when being p-th of sample,When being -1 sample of pth first neuron with
The anti-pass error of k-th of neuron when p-th of sample,K-th of neuron and when p-th of sample when being -1 sample of pth
The anti-pass error of j-th of neuron;
(2.5) sample number learnt is recorded, if currently trained sample number p < P, (2.2) is returned and continues
It calculates, when p=P, jumps into (2.6);
(2.6) according to the correction formula of weightAnd
The weight of each layer of corrective networks one by one;
Wherein vklIt is the hidden layer output of p-th of sample as a result, n0For the number of iterations, wklIt is the input layer of network to hidden
The connection weight of the neuron containing layer,It is j-th of input of p-th of sample, constant η ∈ (0,1) indicates learning rate;
(2.7) total error E is calculatedAIf overall error EAMeet EA< ε reaches maximum number of iterations, then training terminates, no
Then, continue the study of a new round back to (2.2);What wherein ε was indicated is desired minimal error.
5. a kind of cognitive radio frequency spectrum prediction technique based on ANN according to claim 1, it is characterised in that:It is described
The characteristic value that primary user identifies in step (3) includes three maximums in α=2fcCirculation spectrum value at cycle frequency, wherein fcIt is
Carrier frequency.
6. a kind of cognitive radio frequency spectrum prediction technique based on ANN according to claim 1, it is characterised in that:It is described
Nervus opticus network model in step (4) is BP neural network model, and the iterative formula of weight is in model training:
w(n0+ 1)=w (n0)+η(n0)d(n0), wherein w (n0) indicate by n-th0When secondary iteration, composed by each layer weight of network
Matrix, n0For the number of iterations, η (n0) indicate n-th0Learning rate when secondary iteration, as overall error EAWhen reduction, learning rate increases
Add;If overall error EAWhen increase, learning rate reduces,EAFor error.
7. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, which is characterized in that the computer program realizes any one of -6 institute according to claim 1 when being loaded on processor
Spectrum prediction method in the cognitive radio based on ANN stated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810562684.2A CN108924847B (en) | 2018-06-04 | 2018-06-04 | Cognitive radio frequency spectrum prediction method and device based on ANN |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810562684.2A CN108924847B (en) | 2018-06-04 | 2018-06-04 | Cognitive radio frequency spectrum prediction method and device based on ANN |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108924847A true CN108924847A (en) | 2018-11-30 |
CN108924847B CN108924847B (en) | 2022-11-01 |
Family
ID=64418721
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810562684.2A Active CN108924847B (en) | 2018-06-04 | 2018-06-04 | Cognitive radio frequency spectrum prediction method and device based on ANN |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108924847B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109818892A (en) * | 2019-01-18 | 2019-05-28 | 华中科技大学 | Construct Cyclic Spectrum characteristic parameter extraction model and signal modulation mode recognition methods |
CN110289926A (en) * | 2019-07-22 | 2019-09-27 | 南通大学 | Frequency spectrum sensing method based on modulated signal Cyclic Autocorrelation Function asymmetric peak |
CN110336631A (en) * | 2019-06-04 | 2019-10-15 | 浙江大学 | A kind of signal detecting method based on deep learning |
CN110830124A (en) * | 2019-11-21 | 2020-02-21 | 长春理工大学 | Spectrum sensing method based on quantum particle swarm optimization extreme learning machine |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102130732A (en) * | 2011-04-01 | 2011-07-20 | 北京邮电大学 | Cooperative spectrum detection method for cognitive radio based on neural network |
CN102665221A (en) * | 2012-03-26 | 2012-09-12 | 南京邮电大学 | Cognitive radio frequency spectrum perception method based on compressed sensing and BP (back-propagation) neural network |
-
2018
- 2018-06-04 CN CN201810562684.2A patent/CN108924847B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102130732A (en) * | 2011-04-01 | 2011-07-20 | 北京邮电大学 | Cooperative spectrum detection method for cognitive radio based on neural network |
CN102665221A (en) * | 2012-03-26 | 2012-09-12 | 南京邮电大学 | Cognitive radio frequency spectrum perception method based on compressed sensing and BP (back-propagation) neural network |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109818892A (en) * | 2019-01-18 | 2019-05-28 | 华中科技大学 | Construct Cyclic Spectrum characteristic parameter extraction model and signal modulation mode recognition methods |
CN110336631A (en) * | 2019-06-04 | 2019-10-15 | 浙江大学 | A kind of signal detecting method based on deep learning |
CN110336631B (en) * | 2019-06-04 | 2020-10-13 | 浙江大学 | Signal detection method based on deep learning |
CN110289926A (en) * | 2019-07-22 | 2019-09-27 | 南通大学 | Frequency spectrum sensing method based on modulated signal Cyclic Autocorrelation Function asymmetric peak |
CN110830124A (en) * | 2019-11-21 | 2020-02-21 | 长春理工大学 | Spectrum sensing method based on quantum particle swarm optimization extreme learning machine |
Also Published As
Publication number | Publication date |
---|---|
CN108924847B (en) | 2022-11-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108924847A (en) | A kind of cognitive radio frequency spectrum prediction technique and equipment based on ANN | |
Liu et al. | Deep CM-CNN for spectrum sensing in cognitive radio | |
CN110166154B (en) | Software radio frequency spectrum monitoring and identifying method based on neural network | |
Ivanov et al. | Probabilistic spectrum sensing based on feature detection for 6G cognitive radio: A survey | |
CN106791696B (en) | Wireless video monitoring system and its image transfer method and device | |
CN107623554A (en) | A kind of cooperative frequency spectrum sensing method based on trust value, cognition wireless network | |
CN105978644B (en) | Star based on cuckoo searching algorithm ground cognitive system frequency spectrum access method | |
CN111510232B (en) | Vehicle networking combined spectrum sensing method based on neural network and application thereof | |
CN116866129A (en) | Wireless communication signal detection method | |
CN112787736A (en) | Long-short term memory cooperative spectrum sensing method based on covariance matrix | |
CN101969352A (en) | Fast variable polarization-based spectrum sensing method | |
CN110649982A (en) | Double-threshold energy detection method based on secondary user node selection | |
CN106230544A (en) | The monitoring identification of a kind of automobile remote-control interference signal and localization method | |
Wang et al. | Specific emitter identification based on deep adversarial domain adaptation | |
CN108494511A (en) | A kind of dynamic arrival frequency spectrum sensing method based on absolute value accumulation | |
CN111417124A (en) | Method for sensing frequency spectrum in cognitive wireless network environment | |
CN105429913B (en) | More level detections of feature based value and recognition methods | |
CN104394543B (en) | Joint spectrum cognitive method based on Adaboost algorithm | |
Liu et al. | A novel wireless interference identification and scheduling method based on convolutional neural network | |
Zhang et al. | A spectrum sensing algorithm for OFDM signal based on deep learning and covariance matrix graph | |
Wang et al. | Indoor fingerprinting localization based on fine-grained CSI using principal component analysis | |
Zhao et al. | Cyclic spectrum based intelligent modulation recognition with machine learning | |
CN108174379A (en) | The malicious user recognition methods screened based on support vector machines and threshold value and device | |
Xiao et al. | Research on wireless spectrum sensing technology based on machine learning | |
Shrote et al. | An Efficient Cooperative Spectrum Sensing Method Using Renyi Entropy Weighted Optimal Likelihood Ratio for CRN |
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 |