CN114866166B - CNN-based Wi-Fi subcarrier cross-protocol interference identification method - Google Patents
CNN-based Wi-Fi subcarrier cross-protocol interference identification method Download PDFInfo
- Publication number
- CN114866166B CN114866166B CN202210219478.8A CN202210219478A CN114866166B CN 114866166 B CN114866166 B CN 114866166B CN 202210219478 A CN202210219478 A CN 202210219478A CN 114866166 B CN114866166 B CN 114866166B
- Authority
- CN
- China
- Prior art keywords
- subcarrier
- cnn
- subcarriers
- layer
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
-
- 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
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Quality & Reliability (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Mobile Radio Communication Systems (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The invention discloses a Wi-Fi subcarrier interference identification method based on CNN, which specifically comprises the following steps: step 1, collecting and classifying channel characteristic information of Wi-Fi subcarriers in a heterogeneous wireless network; step 2, preprocessing and dividing the obtained data set into a training set and a test set; step 3, learning and training the training set obtained in the step 2 through a CNN model to obtain a trained subcarrier classification model; the CNN model comprises an input layer, an intermediate layer L and an output layer; the middle layer comprises two convolutional layers, a nonlinear fitting function, a full connection layer and an activation function; and 4, carrying out interference identification on the subcarriers in the test set according to the classification model obtained by training in the step 3, and positioning the interfered subcarriers and the non-interfered subcarriers. Experimental results and performance analysis show that compared with the prior art, the method provided by the invention has the advantages that the precision, the generalization capability and the like are improved.
Description
Technical Field
The invention relates to the field of heterogeneous wireless network communication, in particular to a Wi-Fi subcarrier cross-protocol interference identification method based on a convolutional neural network.
Background
The heterogeneous wireless network is a complex communication unit formed by fusing multiple existing communication technologies in an open frequency band of the same ISM, and the phenomenon that multiple incompatible wireless communication technologies share the same spectrum resource at the same time and space is the coexistence of the heterogeneous wireless network. Under the era of high-speed development of artificial intelligence, application scenes (intelligent transportation, internet of vehicles and the like) of coexistence of heterogeneous wireless networks generally exist, cross-protocol co-frequency interference caused by the coexistence phenomenon of the heterogeneous wireless networks reduces the utilization rate of frequency spectrums and wastes effective bandwidth, which is an important challenge facing efficient data transmission in wireless communication. Generally, wireless devices sharing spectrum resources may use a carrier sense multiple access with collision avoidance (CSMA/CA) protocol to implement data transmission, monitor a channel status through a CCA manner, avoid collision by using a binary exponential backoff algorithm, and obtain a larger available bandwidth share of other devices at the cost of sacrificing a certain device. However, CSMA/CA relies on random backtracking, significantly reducing network throughput as more devices attempt to access the channel. And the method can not be applied to a high-power interference scene because the high-power consumption equipment can not detect the existence of the low-power consumption equipment, and the signal gain of the low-power consumption equipment is far less than that of a high-power signal, so that the high-power consumption equipment monopolizes a shared frequency spectrum, and the low-power consumption equipment is starved. In a heterogeneous wireless network coexistence environment, if heterogeneous devices adopting different communication protocols cannot share limited spectrum resources, interference between signals will be caused, which causes packet loss, signal distortion and communication delay increase, and further causes a decrease in data transmission success rate and spectrum utilization rate.
The interference of data transmission in heterogeneous wireless networks is mainly due to the incompatibility of the underlying architectures of different communication protocols. Since various wireless communications do not consider sharing spectrum resources in the same time domain across protocols at the beginning of design, devices conforming to different protocol standards compete for resources in a shared frequency band, and co-frequency interference is generated. So far, the interference problem generated by simultaneous communication of multiple protocols in heterogeneous wireless networks has received wide attention from many scientific researchers and industries in the field of domestic and foreign communications, and a series of solutions are proposed, which mainly include three categories, namely an interference avoidance scheme, an interference elimination scheme and a cross-protocol data transmission scheme. In any solution, the primary premise is to accurately identify cross-protocol interference, and then different interference solutions can be selected. For example: the 2.4GHz frequency band is the global shared ISM frequency band, and Wi-Fi, bluetooth, zigBee and other wireless communication technologies can work in the frequency band. However, since different communication modes conform to different MAC layer protocols and standards, for example, wi-Fi conforms to ieee802.11b/g/n protocol, and ZigBee conforms to ieee802.15.4 protocol, due to incompatibility of modulation/demodulation in physical layer, multiple devices in the same heterogeneous wireless network can generate cross-protocol interference when operating in the same frequency band, which affects overall performance of the heterogeneous wireless network. In the actual communication process, the influence of the interference of the narrow-band ZigBee signal on different Wi-Fi subcarriers is different, two types of interfered subcarriers and non-interfered subcarriers exist, and the performance difference of the subcarriers can bring more flexibility for Wi-Fi communication.
However, the existing channel estimation technology is to perform coarse-grained channel estimation and analysis on the Wi-Fi channel as a whole, and is not refined to the subcarrier level, and the transmission scheme cannot be determined by analyzing the channel characteristic information of all subcarriers and then classifying the subcarriers in the communication process.
Currently, several commonly used channel estimation methods have been proposed in succession:
(1) Channel estimation based on training sequences. The known training sequence is sent, initial channel estimation is carried out at a receiving end, and when useful information data are sent, a judgment is updated by using the initial channel estimation result, so that real-time channel estimation is completed.
(2) And (4) blind channel estimation. The channel estimation is carried out by utilizing the channel structure information and the statistical characteristics of the transmission information symbols, and the method does not need a training sequence and only obtains the channel state information by carrying out correlation processing on the received signals.
(3) And (4) semi-blind channel estimation. No or only a short training sequence is needed. The channel estimation method combines the advantages of the blind estimation method and the training sequence estimation method.
The channel estimation and analysis methods have advantages and disadvantages respectively, are respectively suitable for different specific application scenes, and may have different disadvantages aiming at different situations, but the methods do not pay attention to the channel state information of a WiFi single subcarrier. In a heterogeneous wireless network, after cross-protocol interference, channel states of different subcarriers are different, and if all subcarriers are viewed at the same time, the communication performance may be reduced. Therefore, channel estimation needs to be performed from a subcarrier level, and an effective classification model is adopted to perform interference identification on Wi-Fi subcarriers, so as to quickly locate interfered subcarriers and non-interfered subcarriers.
Disclosure of Invention
In order to solve the problems, the invention provides a Wi-Fi subcarrier interference identification method based on a convolutional neural network. The method comprises the steps of constructing a training set and a testing set after channel characteristic information is preprocessed, further adopting a proposed Convolutional Neural Network (CNN), generating a classification model of subcarrier interference identification through training, and finally realizing the interference identification of the subcarrier in the heterogeneous wireless network through the classification model. The subcarrier interference identification problem is equivalent to learning the conditional probability distribution of subcarrier classification, and the subcarrier is classified and decided by using the maximum conditional probability.
In order to realize the task, the invention adopts the following technical scheme:
a Wi-Fi subcarrier interference identification method based on CNN specifically comprises the following steps:
step 1, collecting and classifying channel characteristic information of Wi-Fi subcarriers in a heterogeneous wireless network;
step 3, learning and training the training set obtained in the step 2 through a CNN model to obtain a trained subcarrier classification model; the CNN model comprises an input layer, an intermediate layer L and an output layer; the middle layer comprises two convolutional layers, a nonlinear fitting function, a full connection layer and an activation function;
and 4, carrying out interference identification on the subcarriers in the test set according to the classification model obtained by training in the step 3, and positioning the interfered subcarriers and the non-interfered subcarriers.
Further, the specific operation of step 1 is as follows:
in time slot T, the transmitting end adopts subcarrier S i Sending K data packets to a receiving node, for subcarrier S i The receiving end feeds back the sub-carrier channel information to the sending end, and the sending end equipment collects the sub-carrier S fed back by the receiving end i After acquiring the channel characteristic information of all the subcarriers, the sending end divides the categories of the n subcarriers.
Further, the step 2 specifically includes the following sub-steps:
step 21, performing data filling on the channel characteristic information to obtain a data set after the data filling; the specific operation is as follows:
abnormal value detection is carried out on values deviating from a normal range in K times of sampling of subcarriers by adopting a boxed graph analysis method, the abnormal values are regarded as missing values, and filling processing is carried out on the missing values by adopting a K-nearest neighbor algorithm;
and step 22, performing normalization processing on the data set filled with the data obtained in the step 21 by using a linear function to obtain a preprocessed data set.
And step 23, dividing the preprocessed data set into a training set and a test set.
Further, in the step 3, the function of the intermediate layer is as follows:
a) The two layers of convolution layers are used for carrying out deep level feature sensing and extraction and locally sensing the channel features of the sub-carriers;
b) Connecting all the characteristics through the full-connection layer, and sending an output value to a classifier;
c) Introducing dropout operation in a full connection layer, randomly deleting part of neurons in a neural network in one cycle, then performing training and optimization processes of the network in the cycle, repeating the process in the next cycle, measuring the quality of model prediction by using a loss function, and leading the loss value to be in a descending trend along with the training until the training is finished;
further, in step 3, the two convolutional layers adopt a 3 × 3 convolutional kernel.
Further, in the step 3, a BP algorithm is adopted for updating and optimizing, learning and training of the network are completed, and prediction errors are minimized; wherein, the optimization process has two stages:
1) Forward calculation is carried out, a final output value is obtained by sequentially calculating from front to back according to input, and the difference between the current output and a target is calculated, namely a loss function is calculated;
2) Reverse updating minimizes loss function by random gradient descent method, and updates weight value by transfer error value;
further, in the step 3, for the intermediate layers L and L + i, a nonlinear function is used to perform nonlinear mapping on the output of the intermediate layer L; the input at L < th > level is denoted as x (L) Weight is represented as u (L) ；
Then input x of layer L +1 (L+1) Expressed as:
x (L+1) ＝f(u (L) x (L) )
weight of intermediate layer L = (u) (1) ，u (2) ，…u (L) ) Input = (x) (1) ，x (2) ，…x (L) ) And f (·) represents a nonlinear function.
Further, in the step 3, the full connection layer obtains the conditional probabilities for implementing subcarrier classification through an activation function softmax, and the sum of the conditional probabilities of different classes is 1; the conditional probability distribution is as follows:
the softmax activation function is expressed as follows:
wherein x is (0) Represents input layer data, u represents a weight, j =0 or 1 represents a subcarrier class that may be output, exp (y (j | u, x) (0) ) Represents the value given when the output layer outputs a subcarrier class of j.
Further, in step 3, the cross entropy loss function expression used by the CNN model is:
where n represents the number of input samples, k represents the number of classes, S i Representing the true class of the input sample i, j representing the prediction class, 1 {. Cndot.) represents that the function is equal to 1 only if the expression in brackets is true, and 0 otherwise.
Further, in step 3, the objective function of the CNN model is:
G(u，x (0) )＝-klog(P(k＝1|u，x (0) ))-(1-k)log(P(k＝0|u，x (0) ))
＝-log(P K (k))
wherein k represents the number of classes, x (0) Denotes input layer data, u denotes weight, P (k =1 calc, x) (0) ) Representing the probability of outputting class 1, P K (k) Representing a conditional probability distribution of the prediction class.
Compared with the prior art, the invention has the following technical characteristics:
(1) The research on the Wi-Fi channel state is refined to the subcarrier level, and the fine-grained analysis and classification are performed on the subcarrier channel characteristic information after the acquisition, so that the method is favorable for more efficiently utilizing frequency spectrum resources in a frequency domain.
(2) The CNN technology is applied to a subcarrier interference identification scene for the first time, channel characteristic information of all subcarriers is comprehensively considered in a model training process, interference identification of the subcarriers is realized, interfered subcarriers and non-interfered subcarriers are quickly positioned, and the CNN technology has high accuracy and generalization capability.
(3) The method is based on a real scene to collect the channel characteristic information of the subcarriers, construct a training set and a test set, and prove the feasibility of the proposed CNN model for subcarrier identification. It is important and valuable to use data collected in real scenes for training and testing.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is an exemplary diagram of subcarrier EVM data;
fig. 3 shows Wi-Fi subcarrier energy monitoring before and after interference, where (a) is before interference and (b) is after interference;
FIG. 4 is a graph showing the classification accuracy of a training data set under a sub-carrier interference recognition CNN model;
FIG. 5 shows the classification accuracy of a test data set under a sub-carrier interference recognition CNN model;
FIG. 6 shows the classification accuracy of a test data set under different configurations of a sub-carrier interference recognition CNN model;
FIG. 7 is a comparison of classification accuracy of the CNN model and the SVM model.
The invention is further explained below with reference to the drawings and the detailed description.
Detailed Description
Related art terms related to the present invention:
1. CNN: convolutional Neural Networks (Convolutional Neural Networks).
2. Subcarrier channel characteristic information: when Wi-Fi equipment in a heterogeneous wireless network is interfered by heterogeneous equipment (such as ZigBee), a subcarrier S is used in a certain fixed time slot T i (0＜i＜n, where i represents a subcarrier index, and n represents a number of subcarriers), all relevant information reflecting a channel state of the subcarriers, such as an Error rate, an energy detection, an amplitude, an EVM (Error Vector Magnitude), and the like.
In order to quickly position the interfered subcarrier and the non-interfered subcarrier, the invention provides a subcarrier interference identification method based on CNN.
The CNN-based subcarrier interference identification method comprehensively considers the real subcarrier channel characteristic information and obtains a subcarrier classification model through CNN training. Firstly, acquiring channel characteristic information of Wi-Fi subcarriers in a heterogeneous wireless network, determining the interference condition of the subcarriers according to the channel characteristic information of the subcarriers, and generating a training set and a test set after data preprocessing; secondly, a classification model is generated through CNN autonomous learning training of an input layer, an intermediate layer L (a convolutional layer, a nonlinear fitting function, a full connection layer and an activation function) and an output layer of the training set; and finally, carrying out interference identification on the subcarriers by utilizing a subcarrier classification model generated by training, namely carrying out maximum conditional probability decision according to conditional probability values of different classes output by the subcarriers, and dividing the subcarriers into different classes.
The invention provides a Wi-Fi subcarrier interference identification method based on CNN, which comprises the following steps:
step 1, collecting and classifying channel characteristic information of Wi-Fi subcarriers in a heterogeneous wireless network. The specific operation is as follows:
in time slot T, the transmitting end adopts subcarrier S i Sending K data packets to a receiving node, for subcarrier S i The sending process of each data packet adopts a closed loop feedback thought, the receiving end feeds back subcarrier channel information to the sending end, and the sending end equipment collects the subcarrier S fed back by the receiving end through an upper computer i After acquiring the channel characteristic information of all the subcarriers, the sending end visually analyzes the channel characteristic information and adds label parameters to the corresponding subcarriers, namely, the classes of the n subcarriers are divided.
and step 21, performing data filling on the channel characteristic information to obtain a data set after the data filling.
This is because due to the existence of cross-protocol interference in the heterogeneous wireless network, when a closed-loop idea is adopted to collect channel characteristic information of some subcarriers, a phenomenon of partial data exception or missing often occurs. For example: in the information acquisition process of the ith time (i < k, k is the sampling time), part of data is distributed as follows: 0.5 160, 12,4,3,25 \ 8230; such data sequences cannot be used directly as data sets.
Specifically, for the abnormal value detection of the value deviating from the normal range in the k times of sampling of the subcarrier, the boxed graph analysis method is adopted in this embodiment, that is, a standard for identifying the abnormal value is provided, the value exceeding the range defined by the Upper Bound (UB) and the Lower Bound (LB) and the random code value are identified as the abnormal value, and the abnormal value is treated as the missing value and is processed by the missing value processing method.
Before the missing value is processed, the mechanism and form of data missing are determined, and the abnormal value (missing value) is completely random missing, random missing or non-random missing. In this embodiment, the data loss belongs to random loss, and for k times of sampling of subcarriers, the data loss situation is related to the subcarrier interfered situation, and abnormal values and missing values are more likely to occur in the interfered subcarriers in the closed-loop feedback process. If missing records are discarded, a large amount of information is lost, resulting in systematic differences between incompletely observed data and completely observed data. For missing value processing, a missing value is filled according to the distribution of the feature data in the rest samples in the sample set.
Specifically, a K neighbor algorithm is adopted to process the missing value, and the missing value is filled according to the average value of the values of the missing value Di in other K pieces of channel characteristic information;
and step 22, carrying out normalization processing on the data set filled with the data obtained in the step 21 to obtain a preprocessed data set.
For the presence of singular sample data in the acquired channel characteristic information, for example, discrete values 0.5 and 150 in the acquired subcarrier channel characteristic information, such discrete data will cause the network convergence speed to become slow, and therefore, the discrete data needs to be normalized to between [0,1] according to a certain rule. In this embodiment, a linear function normalization method is adopted to perform scaling processing, as shown in the following formula:
wherein D is channel characteristic information, D max Maximum value of the channel characteristic information, D min Is a minimum value, D, of the channel characteristic information norm The values are normalized.
And step 23, dividing the preprocessed data set into a training set and a test set.
Step 3, learning and training the training set obtained in the step 2 through a CNN model to obtain a trained subcarrier classification model; the network model comprises an input layer, an intermediate layer (comprising two convolutional layers, a nonlinear fitting function, a full connection layer and an activation function) and an output layer. Wherein, the intermediate layer has the following functions:
a) The two layers of convolution layers are used for carrying out deep level feature sensing and extraction and locally sensing the channel features of the sub-carriers;
b) All features are connected through a full link layer and the output value is fed to the classifier (i.e., the activation function).
c) Introducing dropout operation (conventionally set to be 0.5) in a full connection layer, randomly deleting part of neurons in a neural network in one cycle, then carrying out the training and optimizing process of the network in the cycle, repeating the process in the next cycle, measuring the quality of model prediction by using a loss function, and ideally, leading the loss value to be in a descending trend along with the training until the training is finished. Through Dropout operation, a neural unit and a randomly selected neuron can be commanded to work together, joint adaptability among neuron nodes is weakened, the overfitting condition of a network can be effectively prevented, and the generalization capability of the network is enhanced.
Specifically, the two convolutional layers in step 3 may locally sense the channel characteristics of the subcarriers, and the key is the process of determining the size of the convolutional kernel and optimizing the weight, which specifically includes:
by combining the known channel characteristic information and the characteristic that Wi-Fi subcarriers are subjected to cross-protocol interference, when a CNN model is constructed, a convolution kernel which accords with the characteristics of sample data and cross-protocol interference needs to be designed on the basis of conventional experience, the process of identifying the subcarriers by the CNN model is accelerated, and a 3x3 convolution kernel is adopted. In the problem of subcarrier interference identification in heterogeneous networks, the 3x3 convolution kernel is sufficient to capture the variation of subcarrier channel characteristics relative to a large convolution kernel (5x5,7x7, \8230;) in terms of the effect of the convolution itself. Assuming that the number of convolution kernels is n, 15 x5 convolution kernel parameter is 25n, 23 x3 convolution kernels have parameter of 18n, 23 x3 convolution kernels have the same field as 15 x5 convolution kernel, and the former can effectively reduce the computation complexity and parameter; moreover, 2 convolutional layers with the convolutional kernel size of 3x3 are stacked to have more nonlinear transformation than 1 convolutional layer with the convolutional kernel size of 5x5, the former can use a nonlinear activation function twice, and the latter only has one time (the feature diversity of the former is increased), so that the network capacity is larger, the learning capacity of the CNN on the features is stronger, and further the distinguishing capacity on different types of subcarriers is stronger; from the perspective of model compression, on the premise of the same receptive field, the small convolution kernel stacking is adopted to obtain a deeper network with fewer parameters, and the fitting capability of the network is improved.
For the weight optimization process, the BP algorithm is adopted for updating, the learning and training of the network are completed, and the prediction error is minimized. The optimization process mainly comprises two stages:
1) The forward calculation calculates the final output value according to the input from front to back in sequence, and calculates the difference between the current output and the target, namely calculates the loss function.
2) And the reverse updating utilizes a random gradient descent method to minimize a loss function, and parameters such as a weight value and the like are updated through a transfer error value.
Specifically, for the intermediate layers L and L +1 in step 3, a nonlinear function is used to perform nonlinear mapping on the output of the intermediate layer L, so as to enhance the classification capability of the network on complex features;
non-linear processing is required between the CNN network intermediate layers L and L +1, wherein the input of the L-th layer can be represented as x (L) The weight is represented as u (L) The present embodiment employs an activation function softsign.
Input x of the L +1 layer (L+1) Can be expressed as:
x (L+1) ＝f(u (L) x (L) )
here, the weight of each layer of the intermediate layer may be represented as u = (u) (1) ，u (2) ，…u (L) ) The input may be expressed as x = (x) (1) ，x (2) ，…x (L) ) And f (·) represents a nonlinear function.
Assuming that the output characteristics are not non-linearly processed in the CNN intermediate layer, the output layer and the input layer may be expressed as a simple linear relationship, and the CNN is equivalent to a network model simplified to only one layer, where the relationship may be expressed as follows:
x (O) ＝u (L) u (L-1) …u (1) x (0)
wherein x is (0) Representing input data, x (O) And (4) representing output data, and if a nonlinear processing process is not added in the middle layer, complex features cannot be learned for classification.
Specifically, the full connectivity layer in step 3 needs to obtain conditional probabilities for realizing subcarrier classification through an activation function softmax, and the sum of the conditional probabilities of different classes is 1;
because of the sub-carrier classification into interferenceThe 2 classes disturbed/not disturbed, so that each vector in the output data is of length 2 by softmax, where each data has a value between 0 and | describing the likelihood that the input data belongs to the class it represents relative to the other classes. Assuming here that the classifier predicts the subcarrier class S e {0,1}, which is a discrete random variable, S =0 and S =1 denote the interfered subcarrier and the non-interfered subcarrier, respectively, and vice versa. In this case, the prediction probability P follows a Bernoulli distribution, and its probability mass function can be expressed as P S (s)＝ P s (1-P) 1-s And the conditional probability distribution can be written as the following equation:
further, the expression of the softmax activation function adopted in step 3 is as follows:
wherein x is (0) Represents input layer data, u represents a weight, j =0 or 1 represents a subcarrier class that may be output, exp (y (j | u, x) (0) ) Represents the value given when the output layer outputs a subcarrier class of j. Further, the conditional probability P of the k classification problem can be derived as:
specifically, the cross entropy loss function expression used by the CNN model in step 3 is:
where n represents the number of input samples, k represents the number of classes, S i Representing the true class of the input sample i, and j representing the predicted class1 {. Means } that the function is equal to 1 only if the expression in parentheses is true, and is equal to 0 otherwise.
On the basis of the loss function, an objective function of the CNN model is established, wherein the objective function describes the relation between the conditional probability distribution and the loss function:
G(u，x (0) )＝-klog(P(k＝1|u，x (0) ))-(1-k)log(P(k＝0|u，x (0) ))
＝-log(P K (k))
where k denotes the number of classes (k =2 in the present embodiment), x (0) Represents (input layer data), u represents weight, P (k =1 calc u (0) ) Representing the probability, P, of outputting class 1 K (k) Representing a conditional probability distribution of the prediction classes.
And 4, carrying out interference identification on the sub-carrier to be detected according to the classification model obtained by training in the step 3, and positioning the interfered sub-carrier and the non-interfered sub-carrier.
Specifically, the accuracy rate of predicting the subcarrier category in step 4 can be expressed as;
where n denotes the number of samples, S i Representing the true class of the input sample i and j representing the predicted class.
And 4, predicting the sub-carrier class S e {0,1} of the test set obtained after the data in the step 2 is preprocessed through the classification model obtained through training in the step 3, and performing classification decision by utilizing the maximum conditional probability to position the interfered sub-carriers and the non-interfered sub-carriers.
In order to prove the feasibility and the effectiveness of the method for identifying the subcarrier interference, the method carries out the following experimental evaluation on a data sample set acquired by WARP equipment under the scene of a real heterogeneous wireless network.
1. Description of data set
The data set used in the experiment is collected by experimental equipment, the subcarrier Information of the needed Wi-Fi device end is measured in a heterogeneous network environment, the collected subcarrier Information is divided into a training data set and a testing data set in the experiment, the training data set and the testing data set comprise the throughput of subcarriers, SER (Symbol Error Rate), CSI (Channel State Information) amplitude, subcarrier energy Information and EVM (Error Vector manager), and the specific conditions of the data set are shown in Table 1. As shown in fig. 1 and 2, the captured partial channel information is shown, fig. 1 shows the EVM value and the distribution of the interfered data after the Wi-Fi subcarrier is subjected to cross-protocol interference, fig. 2 shows the channel energy changes before and after the Wi-Fi subcarrier is subjected to cross-protocol interference, and these information need to be able to truly and effectively reflect the subcarrier channel information.
Table 1 statistics of data sets
2. Analysis of the results of the experiment
Fig. 3 shows the classification performance of CNN models on the subcarriers of the training set as the number of training rounds increases. In the environment with relatively high SINR and relatively low SINR, the training turns with the classification accuracy reaching 100% are less in the environment with medium SINR. The reason is that under the high interference environment of the heterogeneous network, when the Wi-Fi subcarrier is interfered by the cross-protocol, avoidance and backspacing on a time domain occur, so that the network convergence speed is reduced, and under the low interference environment, the cross-protocol interference characteristic is weaker at the moment, and the network convergence speed is reduced due to environmental noise.
Fig. 4 shows the classification performance of the CNN model for the test set at different SINRs. It can be seen that, under the medium SINR, the classification accuracy of the test data set is the highest, which is caused by the obvious characteristics that the sub-carriers are interfered by the cross-protocol under the real environment, and under different SINRs, the CNN model is used to achieve the convergence effect, which proves the robustness of the CNN model for interference identification of the Wi-Fi sub-carriers.
Fig. 5 shows the performance variation of the CNN model when the number of convolution layers is increased. By increasing the number of convolution layers, the classification performance of the test set under different SINRs is compared. The experimental result shows that the classification performance after the number of convolution layers is increased is basically consistent with that under the original configuration, which shows that the classification performance cannot be obviously improved by increasing the number of convolution layers under the network model.
Fig. 6 shows a comparison of the performance of the CNN model and the SVM model for subcarrier interference identification under different SINRs. The Support Vector Machine (SVM) requires the selection of subcarrier features from a time domain or a frequency domain, and the method is simple in training and high in recognition speed. However, its performance depends largely on the choice of features and cannot adapt to environmental changes. This is because the SVM relies on feature extraction, and the manually extracted features cannot completely represent the channel features of the Wi-Fi subcarriers in the heterogeneous network coexistence environment. The interference recognition is carried out on the subcarriers through the CNN model, the classification accuracy of the subcarriers under different SINRs can be found to be obviously higher than that of an SVM, and the model has better generalization capability.
Claims (10)
1. A Wi-Fi subcarrier interference identification method based on CNN is characterized by comprising the following steps:
step 1, collecting channel characteristic information of Wi-Fi subcarriers in a heterogeneous wireless network and classifying the channel characteristic information;
step 2, performing data preprocessing on the channel characteristic information acquired in the step 1, wherein the data preprocessing comprises filling missing values and abnormal values, performing normalization processing on the channel characteristic information, and dividing a preprocessed data set into a training set and a test set;
step 3, learning and training the training set obtained in the step 2 through a CNN model to obtain a trained subcarrier classification model; the CNN model comprises an input layer, an intermediate layer and an output layer; the middle layer comprises two convolutional layers, a nonlinear fitting function, a full connection layer and an activation function;
and 4, carrying out interference identification on the subcarriers in the test set according to the classification model obtained by training in the step 3, and positioning the interfered subcarriers and the non-interfered subcarriers.
2. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein the specific operation of step 1 is as follows:
in time slot T, the transmitting end adopts subcarrier S i Sending K data packets to a receiving node, for subcarrier S i The receiving end feeds back the sub-carrier channel information to the sending end, and the sending end equipment acquires the sub-carrier S fed back by the receiving end i After acquiring the channel characteristic information of all the subcarriers, the sending end divides the categories of the n subcarriers.
3. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein the step 2 specifically includes the following substeps:
step 21, performing data filling on the channel characteristic information to obtain a data set after the data filling; the specific operation is as follows:
abnormal value detection is carried out on values deviating from a normal range in K times of sampling of subcarriers by adopting a boxed graph analysis method, the abnormal values are regarded as missing values, and filling processing is carried out on the missing values by adopting a K-nearest neighbor algorithm;
step 22, performing normalization processing on the data set filled with the data obtained in the step 21 by adopting a linear function to obtain a preprocessed data set;
and step 23, dividing the preprocessed data set into a training set and a test set.
4. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein in the step 3, the role of the intermediate layer is as follows:
a) The two layers of convolution layers are used for carrying out deep level feature sensing and extraction and locally sensing the channel features of the sub-carriers;
b) Connecting all the characteristics through the full-connection layer, and sending an output value to a classifier;
c) And introducing dropout operation in a full connection layer, randomly deleting part of neurons in the neural network in one cycle, then performing the training and optimizing process of the network in the cycle, repeating the process in the next cycle, measuring the quality predicted by the model by using a loss function, and leading the loss value to be in a descending trend along with the training until the training is finished.
5. The CNN-based Wi-Fi subcarrier interference identification method of claim 1, wherein in the step 3, the two convolutional layers employ a 3x3 convolutional kernel.
6. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein in step 3, updating optimization is performed by using a BP algorithm to complete learning and training of a network and minimize a prediction error; the process of weight optimization has two stages:
1) Forward calculation is carried out to obtain a final output value by sequentially calculating from front to back according to input, and the difference between the current output and a target is calculated, namely a loss function is calculated;
2) And the reverse updating utilizes a random gradient descent method to minimize a loss function, and updates the weight value through a transfer error value.
7. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein in the step 3, for the intermediate layers L and L +1, the output of the intermediate layer L is nonlinearly mapped by using a nonlinear function; the input at L < th > level is denoted as x (L) The weight is represented as u (L) ；
Then input x of layer L +1 (L+1) Expressed as:
x (L+1) ＝f(u (L) x (L) )
weight u = (u) of each layer of the intermediate layer (1) ,u (2) ,…u (L) ) Input x = (x) (1) ,x (2) ,…x (L) ) And f (·) denotes a nonlinear function.
8. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein in the step 3, the full connectivity layer obtains conditional probabilities for implementing subcarrier classification by activating function softmax, and the sum of the conditional probabilities of different classes is 1; the conditional probability distribution is as follows:
the softmax activation function is expressed as follows:
A(s＝j|u，x (0) )∈[0，1]
wherein x is (0) Represents input layer data, u represents a weight, j =0 or 1 represents a subcarrier class that may be output, y (j | u, x) (0) ) Which represents the value given when the output layer outputs a subcarrier class of j.
9. The method for identifying Wi-Fi subcarrier interference based on CNN according to claim 1, wherein in the step 3, the cross entropy loss function expression used by the CNN model is:
where n represents the number of input samples, k represents the number of classes, S i Representing the true class of the input sample i, j representing the prediction class, 1 {. Cndot.) represents that the function is equal to 1 only if the expression in brackets is true, and 0 otherwise.
10. The CNN-based Wi-Fi subcarrier interference identification method of claim 1, wherein in the step 3, an objective function of the CNN model:
G(u，x (0) )＝-klog(P(k＝1|u，x (0) ))-(1-k)log(P(k＝0|u，x (0) ))
＝-log(P K (k))
where k denotes the number of classes, x (0) Denotes input layer data, u denotes a weight, P (k =1 calc u (0) ) Representing the probability, P, of outputting class 1 K (k) Representing a conditional probability distribution of the prediction classes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210219478.8A CN114866166B (en) | 2022-03-08 | 2022-03-08 | CNN-based Wi-Fi subcarrier cross-protocol interference identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210219478.8A CN114866166B (en) | 2022-03-08 | 2022-03-08 | CNN-based Wi-Fi subcarrier cross-protocol interference identification method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114866166A CN114866166A (en) | 2022-08-05 |
| CN114866166B true CN114866166B (en) | 2023-01-10 |
Family
ID=82628025
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210219478.8A Active CN114866166B (en) | 2022-03-08 | 2022-03-08 | CNN-based Wi-Fi subcarrier cross-protocol interference identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114866166B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111556453A (en) * | 2020-04-27 | 2020-08-18 | 南京邮电大学 | Multi-scene indoor action recognition method based on channel state information and BilSTM |
| CN112163574A (en) * | 2020-11-23 | 2021-01-01 | 南京航天工业科技有限公司 | ETC interference signal transmitter identification method and system based on deep residual error network |
| WO2021043126A1 (en) * | 2019-09-02 | 2021-03-11 | Iq Works Limited | System and method for event recognition |
| CN112990026A (en) * | 2021-03-19 | 2021-06-18 | 西北大学 | Wireless signal perception model construction and perception method and system based on countermeasure training |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11665777B2 (en) * | 2018-09-28 | 2023-05-30 | Intel Corporation | System and method using collaborative learning of interference environment and network topology for autonomous spectrum sharing |
-
2022
- 2022-03-08 CN CN202210219478.8A patent/CN114866166B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021043126A1 (en) * | 2019-09-02 | 2021-03-11 | Iq Works Limited | System and method for event recognition |
| CN111556453A (en) * | 2020-04-27 | 2020-08-18 | 南京邮电大学 | Multi-scene indoor action recognition method based on channel state information and BilSTM |
| CN112163574A (en) * | 2020-11-23 | 2021-01-01 | 南京航天工业科技有限公司 | ETC interference signal transmitter identification method and system based on deep residual error network |
| CN112990026A (en) * | 2021-03-19 | 2021-06-18 | 西北大学 | Wireless signal perception model construction and perception method and system based on countermeasure training |
Non-Patent Citations (4)
| Title |
|---|
| Channel Selective Activity Recognition with WiFi: A Deep Learning Approach Exploring Wideband Information;Fangxin Wang等;《IEEE Transactions on Network Science and Engineering》;20180410;第181-192页 * |
| HEIR: Heterogeneous interference recognition for wireless sensor networks;Meng Hou等;《Proceeding of IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks 2014》;20141009;第1-9页 * |
| 一种通信干扰分类方法;Marc Lichtman等;《通信对抗》;20160315(第01期);第56-62页 * |
| 基于卷积神经网络联合多域特征提取的干扰识别算法;王鹏宇等;《信号处理》;20211025;第1-13页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114866166A (en) | 2022-08-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kulin et al. | End-to-end learning from spectrum data: A deep learning approach for wireless signal identification in spectrum monitoring applications | |
| Davaslioglu et al. | Generative adversarial learning for spectrum sensing | |
| Bitar et al. | Wireless technology identification using deep convolutional neural networks | |
| WO2018204917A1 (en) | Spectral sensing and allocation using deep machine learning | |
| Zhang et al. | Signal detection and classification in shared spectrum: A deep learning approach | |
| CN112347871B (en) | Interference signal modulation identification method for communication carrier monitoring system | |
| Shahid et al. | A convolutional neural network approach for classification of lpwan technologies: Sigfox, lora and ieee 802.15. 4g | |
| Camelo et al. | A semi-supervised learning approach towards automatic wireless technology recognition | |
| CN111585925A (en) | Robust real-time radio frequency signal modulation identification method based on deep learning | |
| Subray et al. | Towards enhancing spectrum sensing: Signal classification using autoencoders | |
| Testi et al. | Machine learning for user traffic classification in wireless systems | |
| Camelo et al. | A general approach for traffic classification in wireless networks using deep learning | |
| CN116915582A (en) | Diagnosis and analysis method and device for fault root cause of communication terminal | |
| KR20230091093A (en) | Data-Based Probabilistic Modeling of Wireless Channels Using Conditional Variational Auto-Encoders | |
| Kumar et al. | Analysis of spectrum sensing using deep learning algorithms: CNNs and RNNs | |
| CN114866166B (en) | CNN-based Wi-Fi subcarrier cross-protocol interference identification method | |
| US20220369115A1 (en) | Predicting a radio spectrum usage | |
| Ya et al. | Modulation recognition of digital signal based on deep auto-ancoder network | |
| Zhang et al. | Machine learning based protocol classification in unlicensed 5 GHz bands | |
| Zhang et al. | A mac protocol identification approach based on convolutional neural network | |
| Zhang et al. | A spectrum sensing algorithm for OFDM signal based on deep learning and covariance matrix graph | |
| Yang et al. | Improving medium access efficiency with intelligent spectrum learning | |
| Rahman et al. | Spectrum based wireless radio traffic classification using hybrid deep neural network | |
| Cheong et al. | Evaluating deep neural networks to classify modulated and coded radio signals | |
| CN114915526B (en) | Communication signal modulation identification method, device and system |
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 |