CN112003663A - Cognitive radio frequency spectrum sensing method based on LSTM neural network - Google Patents

Cognitive radio frequency spectrum sensing method based on LSTM neural network Download PDF

Info

Publication number
CN112003663A
CN112003663A CN202010708961.3A CN202010708961A CN112003663A CN 112003663 A CN112003663 A CN 112003663A CN 202010708961 A CN202010708961 A CN 202010708961A CN 112003663 A CN112003663 A CN 112003663A
Authority
CN
China
Prior art keywords
neural network
lstm neural
cognitive radio
spectrum sensing
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.)
Pending
Application number
CN202010708961.3A
Other languages
Chinese (zh)
Inventor
贾敏
黄齐禹
王学东
郭庆
顾学迈
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.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
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 Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN202010708961.3A priority Critical patent/CN112003663A/en
Publication of CN112003663A publication Critical patent/CN112003663A/en
Pending legal-status Critical Current

Links

Images

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W72/00—Local resource management
    • H04W72/04—Wireless resource allocation
    • H04W72/044—Wireless resource allocation based on the type of the allocated resource
    • H04W72/0453—Resources in frequency domain, e.g. a carrier in FDMA
    • 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)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Electromagnetism (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a cognitive radio frequency spectrum sensing method based on an LSTM neural network, relates to the field of radio monitoring and frequency spectrum management, and aims to solve the problem of long sensing time of a single radio frequency spectrum. The invention is suitable for cognitive radio frequency spectrum sensing.

Description

Cognitive radio frequency spectrum sensing method based on LSTM neural network
Technical Field
The present invention relates to the field of radio monitoring and spectrum management.
Background
In recent years, the growth of wireless communication devices and the increase in demand for wireless communication have made radio frequency, an unrenewable resource, increasingly strained, and spectrum resources have increased to strategic resources in each country due to scarcity. Currently, when managing and allocating wireless spectrum resources, a fixed spectrum resource allocation manner is generally adopted. Under the fixed spectrum resource allocation mode, the whole radio spectrum resource is divided into two parts of authorized spectrum and unauthorized spectrum, and the spectrum allocation system has the phenomenon of unbalanced resource use: on the one hand, the licensed spectrum occupies most of spectrum resources, but in some cases, the licensed user (primary user) does not frequently use its licensed frequency band, and most of the licensed spectrum is not used by the primary user in some time period or some place, but is in an unused state, and as a result, the utilization rate of the licensed spectrum is very low. Some research results on the measurement of spectrum resources by the Federal Communications Commission (FCC) in the united states have shown that the average usage of the measured spectrum is 15% -85% in most regions and for most of the measured time periods. On the other hand, the unlicensed spectrum reserved for unlicensed users occupies only a small portion of the total resources of the spectrum. With the rapid development of wireless communication technology, especially the rapid development of some new wireless service services, such as wireless local area network, wireless personal area network, wireless metropolitan area network, etc., more and more services and users depend on unlicensed spectrum, the unlicensed frequency band becomes saturated more and more, and the problem of spectrum resource shortage becomes more and more serious. It is obvious that the low usage rate of the licensed spectrum is mainly caused by the inflexible fixed spectrum allocation strategy, which is a contradiction that the unlicensed users cannot use the corresponding frequency bands.
In order to solve the above-mentioned outstanding contradiction between the shortage of spectrum resources and the inflexible allocation strategy, which needs to be optimized and adjusted, the proposal of software radio makes it possible to dynamically configure communication devices. Based on the rapid development of machine learning technology and the emergence of software Radio, the concept of Cognitive Radio (CR) was creatively proposed by doctor Mitola. In a Cognitive Radio System (CRS), an unauthorized user (Cognitive user) can sense the frequency spectrum environment around the user, the method monitors and analyzes the Spectrum use condition of the authorized frequency band, can intelligently adjust parameters such as the Spectrum range and the access mode of the authorized user (primary user) on the premise of not generating intolerable interference to the communication of the authorized user, fully utilizes the Spectrum Hole (Spectrum Hole) shown in figure 5, namely the time slot or frequency band which is not used by the primary user in certain time range or region to perform dynamic Spectrum access to realize the communication of the business, and, on the authorized frequency band, the master user enjoys the right to access the frequency band in preference to the cognitive user, and when the unauthorized user detects that the authorized user accesses the authorized frequency band, the master user must actively quit the use of the frequency band. A technology in which a CR can operate on an available frequency band by changing its own parameters is considered as an important technology for improving the utilization rate of current wireless spectrum resources and solving the problem of shortage of spectrum resources.
The four functions of spectrum sensing, spectrum decision, spectrum sharing and spectrum switching in the cognitive radio technology can enable a cognitive user to sense an authorized spectrum and perform opportunistic access to a spectrum cavity so as to perform service communication. However, the implementation of these four functions has some drawbacks, which severely limit the improvement of the communication performance of the cognitive radio system, and mainly include the following four aspects:
(1) when sensing broadband spectrum, generally, a cognitive user needs to scan and sense the whole frequency range, which causes huge sensing processing delay and energy loss.
(2) In an actual situation, due to the limitation of hardware equipment conditions and communication conditions of the cognitive user, a large amount of time is consumed in the spectrum sensing process, so that the time of a data transmission stage in a sensing period is reduced, the sensing accuracy is influenced, the spectrum decision accuracy is further reduced, and the whole spectrum resource utilization rate of a communication system is seriously influenced.
(3) In spectrum sharing, a cognitive user may need to perform communication Service requests with different Quality of Service (QoS) requirements at different times, and in order to ensure fairness and reasonableness of resource allocation, higher time delay may be caused when performing spectrum allocation.
(4) In a passive spectrum switching mode, a cognitive user performs spectrum switching only after perceiving that a master user initiates a service request, and in the passive spectrum switching mode, the cognitive user cannot avoid transmission conflicts with the master user.
Disclosure of Invention
The invention provides a cognitive radio frequency spectrum sensing method based on an LSTM neural network, aiming at solving the problem of long sensing time of a single radio frequency spectrum at present.
The cognitive radio frequency spectrum sensing method based on the LSTM neural network is characterized by comprising the following steps: it comprises the following steps:
step one, establishing an LSTM neural network model;
acquiring channel state information to obtain a channel state data set;
processing the data in the channel state data set obtained in the step two to obtain a processed data set;
step four, training the LSTM neural network model established in the step one by using the processed data set obtained in the step three to obtain a trained LSTM neural network model;
step five, adopting the trained LSTM neural network model obtained in the step four to sense the cognitive radio frequency spectrum,
and finishing one cognitive radio frequency spectrum sensing based on the LSTM neural network, returning to the step one, and performing the next cognitive radio frequency spectrum sensing based on the LSTM neural network.
The invention has the following beneficial effects: after a large number of simulation experiments, the present invention, as shown in the simulations and results of fig. 8 to fig. 15, fully proves that the sensing time of a single radio frequency spectrum is greatly shortened.
Drawings
FIG. 1 is an RNN structural schematic;
FIG. 2 is a schematic diagram of an RNN model structure;
FIG. 3 is a schematic diagram of an RNN deployment architecture;
FIG. 4 is a schematic view of the LSTM structure;
FIG. 5 is a schematic diagram of channel occupancy simulation;
FIG. 6 is a schematic diagram of spectrum holes and spectrum access;
FIG. 7 is a schematic representation of channel occupancy simulation of experimental data;
FIG. 8 is a simulation diagram of the variation of the prediction accuracy of three models with the length of an input sequence;
FIG. 9 is a schematic diagram of comparison simulation of predicted values and true values of an LSTM model;
FIG. 10 is a schematic diagram of a comparison simulation of a predicted value and a true value of an SVM model;
FIG. 11 is a schematic diagram of comparison simulation of the predicted value and the true value of the MLP model;
FIG. 12 is a simulation diagram of the MLP model performance index varying with the length of the input sequence;
FIG. 13 is a simulation diagram of the LSTM model performance index varying with the length of the input sequence;
FIG. 14 is a simulation diagram of the LSTM model performance index varying with the length of the input sequence;
FIG. 15 is a schematic diagram of a data set generation process;
Detailed Description
The specific embodiment one, cognitive radio frequency spectrum perception method based on LSTM neural network, its characteristic is: it comprises the following steps:
step one, establishing an LSTM neural network model;
acquiring channel state information to obtain a channel state data set;
processing the data in the channel state data set obtained in the step two to obtain a processed data set;
step four, training the LSTM neural network model established in the step one by using the processed data set obtained in the step three to obtain a trained LSTM neural network model;
and step five, sensing the cognitive radio frequency spectrum by adopting the trained LSTM neural network model obtained in the step four, finishing one-time cognitive radio frequency spectrum sensing based on the LSTM neural network, returning to the step one, and performing the next cognitive radio frequency spectrum sensing based on the LSTM neural network.
The method also comprises a step six of obtaining the prediction accuracy of the cognitive radio frequency spectrum sensing based on the LSTM neural network in the step five, which comprises the following specific steps: it is examined by the acc value:
the acc value is the ratio of the sum of time slots which are correctly predicted to be idle and time slots which are correctly predicted to be busy of the channel to the sum of all time slots on the channel:
Figure BDA0002595852710000041
in the third step, the data in the channel state data set obtained in the second step is processed, specifically:
thirdly, preprocessing the data in the data set; acquiring useful signal information;
step two, acquiring the occupancy rate condition of the channel according to the useful signal information acquired in the step one;
in the seventh step, the step of obtaining the prediction error prediction probability of the cognitive radio spectrum sensing based on the LSTM neural network in the fifth step specifically includes:
according to the formula:
Figure BDA0002595852710000042
and (4) obtaining.
5. The method for cognitive radio spectrum sensing based on the LSTM neural network according to claim 2, further comprising a step eight of obtaining a probability of error prediction of the cognitive radio spectrum sensing based on the LSTM neural network in the step five, specifically: it is based on the formula:
Figure BDA0002595852710000043
and (4) obtaining.
The expanded structure of the RNN model is shown in fig. 1. Wherein x istRepresenting the input of the input layer at time t, stIs the output of the hidden layer, ytIs the output of the output layer. stIs not only dependent on xtAnd also get and st-1In this regard, it is a feature of RNN structure. Therefore, when we predict ytWhen, not only x is to be usedtAlso use xt-1The information of (1). The unit structure of the RNN model and the calculation process in the training stage are as follows:
training of recurrent neural networks
The RNN network training method differs from the traditional neural network training method in that the training parameters (W, U, V) in the RNN are shared. And the RNN uses a gradient descent method in which the output of each step is related to both the current output and the memory state of the previous steps. The back propagation training algorithm [53] used in RNN expands the neural network along time, redefining the connections in the network.
After the RNN network is developed, it can be seen that each layer has the same weight matrix (W, U, V), and the update process of the weight matrix is as follows:
let x (t) be the time series of inputs, u (t) be the intermediate calculation results, and the final output result be
Figure BDA0002595852710000051
The true values of the training samples are denoted by y (t), where:
x(t)＝(x1(t),x2(t),...xk(t))′
y(t)＝(y1(t),y2(t),...yk(t))′
t＝1,2,...,T (3-1)
the final output error of the model can be expressed as:
Figure BDA0002595852710000052
the output of the model can be obtained by a forward propagation algorithm, and then the error is propagated backwards along the time series, and the error term at each time point can be expressed as:
Figure BDA0002595852710000053
Figure BDA0002595852710000054
in the formulaj(T) is the error at the moment of the output unit T,i(T) is an intermediate state ui(T) an error in the time of day,j(T) andithe error of (T) is calculated as follows:
Figure BDA0002595852710000055
Figure BDA0002595852710000056
by using the back propagation algorithm, the update process of the weight matrix is as follows:
Figure BDA0002595852710000057
Figure BDA0002595852710000058
Figure BDA0002595852710000059
Figure BDA00025958527100000510
wherein, WijIs a weight matrix, U, between each neuron in the hidden layerijIs a weight matrix between the input layer and the hidden layer, VijIs a weight matrix between the hidden layer and the output layer, Vij' is the weight matrix passed by the output layer to the hidden layer.
Compared with the traditional neural network, the RNN has better performance when processing time-related sequence data, but the long-term dependence problem still exists in the RNN training algorithm, mainly the problem that gradient disappears when back propagation is performed, so that the RNN network model cannot train a longer-time sequence, and longer-time information cannot be captured by the RNN network. This problem also exists in spectral prediction, and therefore to solve this problem, the present invention proposes a prediction method based on the LSTM model.
LSTM network model introduction and Structure
The long-short term memory network is a variant of the recurrent neural network, and is mainly used for solving the problem of long-term dependence on time. For example, in the problems of picture description, speech recognition, natural language processing and the like, the LSTM model shows a good effect and is widely applied to various fields.
The chain structure of the repeating modules constitutes the LSTM model, but unlike the RNN model, the problem of gradient disappearance is solved by introducing memory cells and three gate structures, an input gate, a forgetting gate and an output gate. The gate structure contains a Sigmoid function layer, which can compress the value between 0 and 1, which is helpful for updating or removing information, because any number multiplied by 0 is 0, and the information is removed. Similarly, any number multiplied by 1 will yield itself, and this information is perfectly preserved. Thus, the network can know which data needs to be forgotten and which data needs to be stored. Due to the gate structures in the LSTM model, the memory cells existing in each layer can only store long-term information, and the problem of gradient disappearance can be solved.
FIG. 4 is a schematic diagram of the LSTM structure of the present invention, wherein xtThe input data at time t is, in this application, the channel state information data at time t. Sigma is the Sigmoid activation function layer,
Figure BDA0002595852710000061
calculation representing multiplication of elements, sigma
Figure BDA0002595852710000062
The combination constitutes a gate structure, denoted forgetting gate, input gate and output gate from left to right, respectively. The forgetting gate determines which related information in the previous memory unit needs to be reserved; the input gate determines which information in the current input is important and needs to be added to the memory unit; the output gate determines what the next hidden state should be.
Training of LSTM models
The training method of the LSTM model covers the calculation processes of a forward calculation stage and a backward propagation stage, and the specific calculation process is as follows:
(1) and (3) forward calculation:
1) forget the door: the input vector at the time t and the output vector of the output layer at the time t-1 jointly determine the output state of the forgetting gate at the time t, and the calculation method is shown in the formulas 4-15:
ft＝σ(Wf·(yt-1,xt)+bf) (3-11)
in the formula:
ft-the output vector at time t;
sigma-sigmoid activation function;
Wf-a weight matrix, formed by the matrix WfyAnd matrix WfxSplicing to obtain the finished product;
yt-1,xt-the output vector at time t-1 of the output layer and the input vector at time t of the input layer, respectively;
bf-a bias term.
Wherein, WfCan be expressed as:
Figure BDA0002595852710000071
2) an input gate: the input gate can be represented by formulas 3-13, and the meaning represented by each symbol in the formulas is similar to that of the forgetting gate, and is not described herein.
it＝σ(Wi·(yt-1,xt)+bi) (3-13)
3) A memory unit: the state of the memory cell at the current time can be calculated in two steps, first using the output at time t-1 and the input at time t to determine the state of the current input cell
Figure BDA0002595852710000072
The calculation is carried out, and the calculation expression of the calculation can be expressed as:
Figure BDA0002595852710000073
then for the memory cell state c at the current timetAnd (3) calculating:
Figure BDA0002595852710000074
the symbol in the formula represents a multiplication operation between the respective elements. Cell state input at LSTM current time
Figure BDA0002595852710000075
And long term memory ct-1The junctions are joined by the memory cells.
4) An output gate:
ot＝σ(Wo·(yt-1,xt)+bo) (3-16)
yt＝ot*tanh(ct) (3-17)
ytrepresenting the final output of the LSTM model at the current time, i.e. the spectrum at time tAnd (5) predicting the value. As can be seen from equations 3-17, the output of the output gate and the state of the memory cell at the current time together determine ytThe value of (c).
(2) And (3) error back propagation process:
similar to the RNN model, the calculation of the error back propagation of the LSTM model also utilizes the value of an error term, and the weight is updated by a gradient descent method according to the error term obtained by calculation. In the training process of the LSTM model, the weight values of the model to be learned are 4 groups, and the specific learning parameters are shown in a table 3-1:
TABLE 3-1 learning parameters
Figure BDA0002595852710000081
The derivation process of the weight matrix in the backward propagation has different roles, so the weight matrix is split as follows:
Wo＝[Woy Wox]
Wf＝[Wfy Wfx]
Wi＝[Wiy Wix]
Wc＝[Wcy Wcx] (3-18)
let the error (loss function) be E and the output at time t be ytThen the layer error term is output at time ttComprises the following steps:
Figure BDA0002595852710000082
for the four weight matrices described above, there are four weighted inputs, with time t corresponding to ft、it、ct、otLet their corresponding error terms be
Figure BDA0002595852710000083
Then:
Figure BDA0002595852710000091
similarly, the chain rule shows that:
Figure BDA0002595852710000092
Figure BDA0002595852710000093
Figure BDA0002595852710000094
the error propagation along the time direction is to calculate the error term at the moment of t-1t-1，t-1The calculation expression is:
Figure BDA0002595852710000095
the computational formula for the propagation of the error term to the upper layer can be expressed as:
Figure BDA0002595852710000096
wherein, W (y)t,xt) + b represents the weighted input for layer l-1 and f is the activation function for layer l-1.
Taking the output gate as an example, the gradient of each weighting matrix at the time t and the gradient of the bias term are calculated by using the calculated error term, and the calculation formula can be expressed as:
Figure BDA0002595852710000097
Figure BDA0002595852710000098
Figure BDA0002595852710000101
the final gradient is the sum of the gradients at the various time instants:
Figure BDA0002595852710000102
Figure BDA0002595852710000103
Figure BDA0002595852710000104
similarly, the weight matrix and the gradient of the bias term of the forgetting gate, the input gate and the memory unit can be obtained.
Single-channel spectrum prediction model design based on LSTM network
3.4.1 data set description for model training and testing
The experimental data used herein are derived from the wireless spectrum usage of the school of maastrichtt (Maastricht) at the University of Aachen industry, germany, measured with an Agilent E4440A spectrum analyzer for one week, measuring the frequency band covering the range of 20MHz to 1.52GHz, with a measurement resolution bandwidth of 200KHz, scanning the entire measurement frequency band each time, with an average scanning time of 1.8 s. Approximately 335000 scans are performed in a week, and the frequency point number of each scan sampling is 8192. Thus, there are 335000 × 8192 data points (approximately 9GB data size) in the entire data set.
(1) Data pre-processing
In the threshold value comparison method, once the threshold value is set to be smaller, some noise signals are mistaken for useful signals, and if the threshold value is higher, some useful signals are missed. How to reasonably determine the decision threshold value is a key step in data preprocessing.
In the current practical research, the threshold value is generally determined by researchers by using an empirical method, that is, a certain tolerance value is used as a threshold for signal determination on the basis of the background noise (or average noise) level. In view of the wide application of the method, and the advantages of high precision of the prediction result, convenience in calculation and the like, the method is also used when experimental data is preprocessed.
In order to provide a data sequence analyzed by the experiment, a threshold value processing method is used for converting a spectrum power value in a data set into a binary representation form of 0 and 1, wherein 1 represents that the channel is occupied by a main user at the moment t, 0 represents that the channel is in an idle state, and a cognitive user can perform spectrum access.
The following concepts are first clarified:
channel: in the data used in this experiment, the channel refers to a 200KHz bandwidth. Channel X refers to the frequency band [ (20+0.2(X-1)) MHz, (20+0.2X) MHz ], X > 0, and the channels are indexed sequentially.
Channel State Information (CSI): is a function of time and channel.
CSI (t, c) ═ 0 indicates that channel c is in an idle state in the t slot, and CSI (t, c) ═ 1 indicates the opposite state.
Converting the channel raw data into CSI data (0 or 1) is a binary hypothesis testing process, and for convenience of processing, through a simple threshold processing process: for channel c, the threshold is set to 5dB above the channel background noise. In the t slot, if the power value is smaller than the threshold value, CSI (t, c) is 0, otherwise, CSI (t, c) is 1. A sliding window in Time steps is then set for generating the data set of this experiment, and the data set generation process for multiple channels is shown in fig. 15.
In consideration of the fact that the collected spectrum data is huge in amount and is not beneficial to statistical analysis, data of 885MHz-915MHz (GSM900 uplink frequency band), 925M-960MHz (GSM900 downlink frequency band) and a TV frequency band are selected from the collected spectrum data for preprocessing and analysis. And then selecting data generated by ten channels in the frequency band of 885MHz-915MHz as experimental data of the experiment for training and testing the model.
The original data volume is very large and irregular, and such data is obviously not beneficial to statistical analysis and calculation of the overall occupation situation of the frequency spectrum, so that data preprocessing needs to be performed on the frequency spectrum metadata.
Spectral metadata needs to be preprocessed before statistical analysis of the wireless-spectrum data. Due to the occupation of the spectrum, there are only two states, namely an idle state and an occupied state. In general, researchers use a threshold comparison method to preprocess spectrum data, that is, if the signal strength in a channel at a certain time is greater than a preset threshold, the channel is considered to be occupied by a primary user and is denoted by 1. Otherwise, it is represented by 0.

Claims (5)

1. The cognitive radio frequency spectrum sensing method based on the LSTM neural network is characterized by comprising the following steps: it comprises the following steps:
step one, establishing an LSTM neural network model;
acquiring channel state information to obtain a channel state data set;
processing the data in the channel state data set obtained in the step two to obtain a processed data set;
step four, training the LSTM neural network model established in the step one by using the processed data set obtained in the step three to obtain a trained LSTM neural network model;
step five, adopting the trained LSTM neural network model obtained in the step four to sense the cognitive radio frequency spectrum,
and finishing one cognitive radio frequency spectrum sensing based on the LSTM neural network, returning to the step one, and performing the next cognitive radio frequency spectrum sensing based on the LSTM neural network.
2. The method for cognitive radio spectrum sensing based on the LSTM neural network according to claim 1, further comprising a sixth step of obtaining a prediction accuracy of the cognitive radio spectrum sensing based on the LSTM neural network in the fifth step, specifically: it is examined by the acc value:
the acc value is the ratio of the sum of the time slots correctly predicted to be idle and the time slots correctly predicted to be busy of the channel to the sum of all the time slots on the channel:
Figure FDA0002595852700000011
3. the cognitive radio spectrum sensing method based on the LSTM neural network according to claim 1, wherein in step three, the data in the channel state data set obtained in step two is processed, specifically:
thirdly, preprocessing the data in the data set; acquiring useful signal information;
step two, acquiring the occupancy rate condition of the channel according to the useful signal information acquired in the step one;
4. the method for cognitive radio spectrum sensing based on the LSTM neural network according to claim 2, further comprising a seventh step of learning the prediction error prediction probability of the cognitive radio spectrum sensing based on the LSTM neural network in the fifth step, specifically:
according to the formula:
Figure FDA0002595852700000012
and (4) obtaining.
5. The method for cognitive radio spectrum sensing based on the LSTM neural network according to claim 2, further comprising a step eight of obtaining a probability of a wrong prediction of the cognitive radio spectrum sensing based on the LSTM neural network in the step five, specifically: it is based on the formula:
Figure FDA0002595852700000013
CN202010708961.3A 2020-07-22 2020-07-22 Cognitive radio frequency spectrum sensing method based on LSTM neural network Pending CN112003663A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010708961.3A CN112003663A (en) 2020-07-22 2020-07-22 Cognitive radio frequency spectrum sensing method based on LSTM neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010708961.3A CN112003663A (en) 2020-07-22 2020-07-22 Cognitive radio frequency spectrum sensing method based on LSTM neural network

Publications (1)

Publication Number Publication Date
CN112003663A true CN112003663A (en) 2020-11-27

Family

ID=73467261

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010708961.3A Pending CN112003663A (en) 2020-07-22 2020-07-22 Cognitive radio frequency spectrum sensing method based on LSTM neural network

Country Status (1)

Country Link
CN (1) CN112003663A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112564881A (en) * 2020-12-01 2021-03-26 国网安徽省电力有限公司 5G communication self-adaptive transmission method based on long-time multi-threshold channel state prediction
CN113114400A (en) * 2021-04-14 2021-07-13 中南大学 Signal frequency spectrum hole sensing method based on time sequence attention mechanism and LSTM model
CN113283576A (en) * 2021-01-18 2021-08-20 上海应用技术大学 Spectrum sensing method for optimizing LSTM based on whale algorithm
CN113315593A (en) * 2021-05-20 2021-08-27 南京工业大学 Frequency spectrum sensing algorithm based on FLOM covariance matrix and LSTM neural network
CN113347638A (en) * 2021-05-14 2021-09-03 东北大学 Dynamic cooperative spectrum prediction and sensing method for cognitive industry Internet of things
CN113346969A (en) * 2021-06-07 2021-09-03 中山大学 Frequency spectrum sensing method and system based on gated cyclic unit
CN114567398A (en) * 2022-04-11 2022-05-31 浙江工业大学 Frequency spectrum sensing method based on convolution long-time and short-time memory neural network

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160099010A1 (en) * 2014-10-03 2016-04-07 Google Inc. Convolutional, long short-term memory, fully connected deep neural networks
CN109194423A (en) * 2018-08-13 2019-01-11 中国人民解放军陆军工程大学 The single-frequency point spectrum prediction method of shot and long term memory models based on optimization

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20160099010A1 (en) * 2014-10-03 2016-04-07 Google Inc. Convolutional, long short-term memory, fully connected deep neural networks
CN109194423A (en) * 2018-08-13 2019-01-11 中国人民解放军陆军工程大学 The single-frequency point spectrum prediction method of shot and long term memory models based on optimization

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
卢光跃等: "基于LSTM神经网络的频谱感知算法", 《信号处理》 *
周虎等: "基于深度LSTM神经网络的在线消费评论情感分类研究", 《中华医学图书情报杂志》 *
杨旭等: "LSTM神经网络在太阳F_(10.7)射电流量中期预报中的应用", 《空间科学学报》 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112564881A (en) * 2020-12-01 2021-03-26 国网安徽省电力有限公司 5G communication self-adaptive transmission method based on long-time multi-threshold channel state prediction
CN113283576A (en) * 2021-01-18 2021-08-20 上海应用技术大学 Spectrum sensing method for optimizing LSTM based on whale algorithm
CN113283576B (en) * 2021-01-18 2024-02-06 上海应用技术大学 Spectrum sensing method for optimizing LSTM (least squares) based on whale algorithm
CN113114400A (en) * 2021-04-14 2021-07-13 中南大学 Signal frequency spectrum hole sensing method based on time sequence attention mechanism and LSTM model
CN113114400B (en) * 2021-04-14 2022-01-28 中南大学 Signal frequency spectrum hole sensing method based on time sequence attention mechanism and LSTM model
CN113347638A (en) * 2021-05-14 2021-09-03 东北大学 Dynamic cooperative spectrum prediction and sensing method for cognitive industry Internet of things
CN113347638B (en) * 2021-05-14 2024-04-30 东北大学 Cognitive industrial Internet of things dynamic cooperation spectrum prediction and perception method
CN113315593A (en) * 2021-05-20 2021-08-27 南京工业大学 Frequency spectrum sensing algorithm based on FLOM covariance matrix and LSTM neural network
CN113346969A (en) * 2021-06-07 2021-09-03 中山大学 Frequency spectrum sensing method and system based on gated cyclic unit
CN114567398A (en) * 2022-04-11 2022-05-31 浙江工业大学 Frequency spectrum sensing method based on convolution long-time and short-time memory neural network
CN114567398B (en) * 2022-04-11 2024-04-23 浙江工业大学 Frequency spectrum sensing method based on convolution long short-time memory neural network

Similar Documents

Publication Publication Date Title
CN112003663A (en) Cognitive radio frequency spectrum sensing method based on LSTM neural network
CN110138475B (en) Self-adaptive threshold channel occupancy state prediction method based on LSTM neural network
CN110505644B (en) User task unloading and resource allocation joint optimization method
CN108123828B (en) Ultra-dense network resource allocation method based on access user mobility prediction
CN112383369A (en) Cognitive radio multi-channel spectrum sensing method based on CNN-LSTM network model
Zhang et al. Spectrum allocation in cognitive radio networks using swarm intelligence
Zhang et al. Deep learning based user association in heterogeneous wireless networks
CN109412661B (en) User clustering method under large-scale MIMO system
Mosavat-Jahromi et al. Prediction and modeling of spectrum occupancy for dynamic spectrum access systems
CN110932807A (en) Spectrum sensing method of MIMO (multiple input multiple output) system under non-Gaussian noise
CN114723057A (en) Neural network collaborative reasoning method for multi-access edge computing system
CN106972900B (en) Based on generalized T2Blind spectrum sensing method for statistics
Li et al. MTF 2 N: Multi-Channel Temporal-Frequency Fusion Network for Spectrum Prediction
Gracla et al. Learning resource scheduling with high priority users using deep deterministic policy gradients
Zuo et al. Spectrum prediction for frequency bands with high burstiness: Analysis and method
CN112702133A (en) Main user channel state prediction method based on dynamic Bayesian hidden Markov model
Valadão et al. Cooperative spectrum sensing system using residual convolutional neural network
Usha et al. Dynamic spectrum sensing in cognitive radio networks using ML model
CN108183757B (en) Segmented approximate convex decomposition method for perception duration and resource allocation joint optimization
CN112243283B (en) Cell-Free Massive MIMO network clustering calculation method based on successful transmission probability
Chirov et al. Spectrum Occupancy Prediction Algorithm Using Artificial Neural Networks
CN108055699B (en) A method of perception duration and resource allocation combined optimization
CN117641598B (en) Consumer electronic resource allocation method and system based on split robust learning
CN111162853B (en) Cooperative broadband compressed spectrum detection method based on supervised learning in cognitive wireless network
Wang et al. Network association for cognitive communication and radar co-systems: A POMDP formulation

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: 20201127

RJ01 Rejection of invention patent application after publication