Wireless communication signal remote interference discrimination method and device, storage medium and terminal

Abstract

The application discloses a wireless communication signal remote interference discrimination method and device, a storage medium and a terminal, wherein the method comprises the following steps: acquiring parameter information of a to-be-detected signal, and performing normalization processing on the parameter information of the to-be-detected signal to obtain to-be-detected parameter information; inputting the to-be-detected parameter information into all interference discrimination neural networks in an interference discrimination neural network group respectively, so as to acquire interference labels output by each interference discrimination neural network; performing voting on all possible discrimination results based on the interference labels output by each interference discrimination neural network and the weights of the interference discrimination neural networks, and taking the possible discrimination result with the most votes as an interference discrimination result of the to-be-detected signal. The method of the application combines a neural network with evolutionary calculation, and realizes fast and efficient wireless communication signal remote interference discrimination with relatively high accuracy and reliability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, in particular to a wireless communication signal remote interference discrimination method and device, a storage medium and a terminal. BACKGROUND

[0002] In the process of wireless signal transmission between base stations, if the transmitted signal is interfered by atmospheric waveguide, the transmitted signal will not propagate along the expected path, and the transmitted signal of the remote base station will be received by other non-expected same-frequency base stations. At this time, the guard period (GP) set in advance by the base station cannot filter out the interference signal, thereby causing the normally transmitted signal to be interfered. The main reason for the same-frequency interference is that the occurrence of atmospheric waveguide makes the signal that should have rapidly attenuated propagate a long distance in the waveguide layer with low attenuation. In relatively open areas such as rural areas, suburban areas and coastal areas, the distance between base stations is relatively far, and the transmission of wireless signals is easily interfered by atmospheric waveguide, thereby causing interference to the normal communication of people at the receiving end far away.

[0003] The wireless electromagnetic wave signal from the remote end mainly has an impact on the uplink signal sent by the handheld mobile device to the base station, that is, the uplink signal contains lower energy than the remote signal, so the remote signal is easy to drown out the uplink signal.

[0004] In the prior art, non-integrated classification algorithms and integrated classification algorithms are usually used for atmospheric waveguide interference discrimination. The non-integrated classification algorithm has a faster learning speed, but its implementation effect is not ideal on real complex problems. The integrated learning realizes the learning of different features of the data set through the combination of multiple algorithms, but most current integrated algorithms do not consider the classification problem of minority samples, and thus the classification and learning ability for minority samples are not ideal, and the accuracy is relatively low. SUMMARY

[0005] The technical problem to be solved by the present application is that the existing wireless communication signal remote interference discrimination method does not consider the classification problem of minority samples, and the classification and learning ability for minority samples are not ideal, and the accuracy is relatively low.

[0006] In order to solve the above technical problem, the present application provides a wireless communication signal remote interference discrimination method, comprising:

[0007] obtaining parameter information of a to-be-detected signal, and performing normalization processing on the parameter information of the to-be-detected signal to obtain to-be-detected parameter information, the parameter information of the to-be-detected signal including meteorological information when the to-be-detected signal is received, distance information between a to-be-detected signal sending end and a base station receiving end, and network side information of the to-be-detected signal base station receiving end.

[0008] inputting the parameter information to be detected into all interference discrimination neural networks in the interference discrimination neural network group respectively to obtain interference labels output by each interference discrimination neural network;

[0009] voting all possible discrimination results based on the interference labels output by each interference discrimination neural network and the weights of the interference discrimination neural networks, and taking the possible discrimination result with the most votes as the interference discrimination result of the signal to be detected.

[0010] Preferably, the interference discrimination neural network group acquisition process comprises:

[0011] acquiring a set of training signal parameter information, and processing all training signal parameter information in the set of training signal parameter information respectively to obtain a set of training parameter information, and dividing all training parameter information in the set of training parameter information into a training subset and a detection subset, wherein the training signal parameter information comprises meteorological information when the training signal is received, distance information between a training signal sending end and a base station receiving end, network side information of the base station receiving end of the training signal, and path loss of the training signal;

[0012] constructing a plurality of preset neural networks, and training all the preset neural networks by using all the training parameter information in the training subset to obtain a plurality of trained neural networks, wherein the preset neural network comprises an input layer, an output layer and a plurality of hidden layers, neurons between all layers are connected in a one-way full interconnection manner, and a nonlinear activation function is arranged on neurons of all the hidden layers and neurons of the output layer, and the number of hidden layers and / or the number of neurons of the hidden layers are different between any two preset neural networks;

[0013] detecting all the trained neural networks by using all the training parameter information in the detection subset, calculating interference judgment accuracy rates of each trained neural network, and selecting a preset number of trained neural networks as temporary neural networks in descending order of the interference judgment accuracy rates;

[0014] setting a crossover operation occurrence probability and a mutation operation occurrence probability in evolution calculation, and performing evolution calculation on all the temporary neural networks by using the crossover operation or the mutation operation to obtain a plurality of interference discrimination neural networks, each interference discrimination neural network has a corresponding weight, and all the interference discrimination neural networks form an interference discrimination neural network group.

[0015] Preferably, the processing of the training signal parameter information comprises:

[0016] normalize meteorological information when the training signal is received, distance information between the training signal sending end and the base station receiving end, and network side information of the training signal base station receiving end to obtain signal parameter information;

[0017] tag path loss of the training signal in a preset manner to obtain an interference label;

[0018] The training parameter information includes the signal parameter information and the interference label.

[0019] Preferably, the step of tagging path loss of the training signal in a preset manner to obtain an interference label comprises:

[0020] determining whether the path loss of the training signal is greater than a preset loss value, and if yes, determining that the interference label of the training signal is 1, otherwise, determining that the interference label of the training signal is 0, wherein the interference label of 1 indicates that the current signal is subjected to remote interference, and the interference label of 0 indicates that the current signal is not subjected to remote interference.

[0021] Preferably, each layer of the preset neural network calculates a gradient value based on a back propagation algorithm, and adjusts connection weights between layers in the preset neural network according to a gradient descent principle.

[0022] Preferably, a bias neuron is added to each of the input layer and the hidden layers of the preset neural network.

[0023] Preferably, in the evolutionary calculation process, a fitness function G-mean is used as an index for measuring learning ability of the training neural network.

[0024] To solve the above technical problems, the application further provides a wireless communication signal remote interference discrimination device, comprising a data acquisition module, a data operation module and an interference discrimination result determination module.

[0025] The data acquisition module is configured to acquire parameter information of a to-be-detected signal, and normalize the parameter information of the to-be-detected signal to obtain to-be-detected parameter information, wherein the parameter information of the to-be-detected signal includes meteorological information when the to-be-detected signal is received, distance information between a to-be-detected signal sending end and a base station receiving end, and network side information of the to-be-detected signal base station receiving end.

[0026] The data operation module is configured to input the to-be-detected parameter information into all interference discrimination neural networks of an interference discrimination neural network group, so as to obtain an interference label output by each of the interference discrimination neural networks.

[0027] The interference discrimination result determination module is configured to vote all possible discrimination results based on the interference labels output by each interference discrimination neural network and the weights of the interference discrimination neural network, and take the possible discrimination result with the most votes as the interference discrimination result of the to-be-detected signal.

[0028] To solve the above technical problems, the application further provides a storage medium, which stores a computer program, and the program is executed by a processor to implement the wireless communication signal remote interference discrimination method.

[0029] To solve the above technical problems, the application further provides a terminal, which comprises a processor and a memory in communication connection with the processor.

[0030] The memory is configured to store a computer program, and the processor is configured to execute the computer program stored in the memory to enable the terminal to implement the wireless communication signal remote interference discrimination method.

[0031] Compared with the prior art, one or more of the above solutions can have the following advantages or beneficial effects:

[0032] The wireless communication signal remote interference discrimination method provided by the application comprises the following steps: constructing a plurality of neural networks for interference discrimination; training the neural networks based on meteorological information when a signal is received, distance information of a sending terminal and a receiving terminal, network side information of the receiving terminal and path loss as data basis; and performing evolutionary calculation on several neural networks with higher accuracy to improve the learning ability of the neural networks, and then improving the accuracy and reliability of the neural networks in discriminating whether a signal is interfered by a remote interference.

[0033] Other features and advantages of the application will be described in the following description, and become apparent from the description, or learned from the practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0034] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application, and are used together with the application examples to explain the application, and do not constitute a limitation on the application. In the drawings:

[0035] Figure 1 A flowchart of a wireless communication signal remote interference discrimination method according to an embodiment of the application is shown;

[0036] Figure 2 A part of a process diagram of a wireless communication signal remote interference discrimination method according to an embodiment of the application is shown.

[0037] Figure 3 Fig. 1 shows a schematic diagram of the process of cross operation on two temporary neural networks in the embodiment one of the present application;

[0038] Figure 4 Fig. 2 shows a schematic diagram of the process of mutation operation on temporary neural networks in the embodiment one of the present application;

[0039] Figure 5 Fig. 3 shows a schematic diagram of the structure of the wireless communication signal remote interference discrimination device in the embodiment two of the present application;

[0040] Figure 6 Fig. 4 shows a schematic diagram of the structure of the terminal in the embodiment four of the present application. DETAILED DESCRIPTION

[0041] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and embodiments, so that the technical problems can be solved by applying technical means, and the realization process of technical effects can be fully understood and implemented. It should be noted that, as long as there is no conflict, each embodiment in the present application and each feature in each embodiment can be combined with each other, and the technical solutions formed thereby are within the protection scope of the present application.

[0042] The atmospheric waveguide can capture the radio electromagnetic wave signal, so that the signal propagates in the atmospheric waveguide layer. At this time, the trajectory of the signal begins to bend towards the earth's surface and continuously propagates forward. The signal propagating in the atmospheric waveguide layer is different from the signal propagating in free space, and the signal in the waveguide layer has very low attenuation and can propagate a long distance, at which time the far-end co-frequency base station will receive this interference signal. For the far-end cell base station, people hold mobile devices to send the information needed to be transmitted to the base station, and then the base station transmits the information. The downlink signal sent by the base station to the mobile device belongs to a signal with high intensity, and such signal contains higher energy than the radio electromagnetic wave signal from the far end, so it is not easy to be interfered by the far-end signal. However, for the uplink signal sent by the handheld mobile device to the base station, the uplink signal contains lower energy than the far-end signal, so the far-end signal can easily drown the uplink signal. In order to minimize the possibility of atmospheric waveguide drowning normal signals and improve the reliability of communication, it is necessary to discriminate the remote interference of wireless communication signal and adjust the base station in time.

[0043] To solve this problem, three types of interference methods based on artificial intelligence (AI) are introduced here. In the prior art, non-integrated classification algorithms and integrated classification algorithms are usually used for wireless communication signal remote interference discrimination. Typical representatives of non-integrated classification algorithms include k-nearest neighbor algorithm (kNN), support vector machine (SVM), naive Bayes (NB), etc. Non-integrated classification algorithms place all sample points in a high-dimensional space, and perform interference discrimination by distance matching and finding a hyperplane. These methods find the most suitable matching point or matching surface for classification through simple mathematical calculations, so the learning speed is very fast and suitable for large data classification problems.

[0044] Although the non-integrated classification algorithm has a fast learning speed, it does not achieve ideal results in real complex problems. Currently, people have proposed an ensemble learning method to deal with more complex classification problems. Ensemble learning combines multiple algorithms to solve complex classification problems. Typical algorithms of ensemble learning include random forest (RF), bagging, boosting, stacking, etc. Ensemble learning learns different features of the data set through the combination of multiple algorithms, greatly improving the accuracy of classification. However, the current classification problem faces a new challenge, which is the classification problem of unbalanced data sets, such as cancer diagnosis, machine fault discrimination, spam message interception, and wireless communication signal remote interference discrimination. Most current ensemble algorithms do not consider the classification problem of minority samples, so the classification and learning ability of minority samples are not ideal, and the accuracy is relatively low.

[0045] We assume that the input of the wireless communication signal remote interference discrimination process is the real-time data that needs to be judged, and the output is the result of judging whether wireless communication signal remote interference will occur. When the path loss adopts the ratio of the actual antenna axial equivalent omnidirectional radiation power to the equivalent omnidirectional receiving power of the same polarization, the calculation expression of the path loss can be expressed as:

[0046]

[0047] wherein represents the signal field strength corresponding to the distance position, represents the path loss corresponding to the distance position, the wavelength of the transmitted signal is , the horizontal distance of propagation is x, and the vertical height of the transmitting antenna is z. Therefore, it can be known that the path loss of the wireless signal is affected by the horizontal transmission distance of the transmitted signal, the frequency of the transmitted signal, and the signal field strength. The calculation of the field strength is the key to calculating the path loss.

[0048] The field strength expression of the initial position of the transmitting antenna is:

[0049]

[0050]

[0051] Wherein B represents the antenna beam function, θ represents the antenna elevation angle, and p represents the beam. The initial position field strength of the transmitting antenna can be linked with the field strength by the forward parabolic equation, and the expression of the forward parabolic equation is as follows:

[0052]

[0053] By eliminating the partial derivative on the field strength, the expression of the step-by-step solution containing the initial field strength and the field strength is obtained as follows:

[0054]

[0055] Wherein k represents the wave number, n represents the refractive index, a is the earth radius, and h represents the height above the ground. The calculation expression of the refractive index is as follows:

[0056]

[0057] Wherein T represents the temperature, represents the pressure, and e represents the water vapor pressure. Therefore, the corresponding refractive index can be obtained by knowing the corresponding meteorological parameters. If more complex and more accurate refractive index calculation is required, meteorological data such as air temperature, air pressure, relative humidity, water vapor pressure, wind speed and precipitation at the corresponding position can also be obtained. The propagation distance of the signal is directly calculated by the longitude and latitude coordinates between the base stations. Therefore, the related factors for judging whether the base station is subjected to remote interference of the wireless communication signal include the signal wavelength, the longitude information, the latitude information, the meteorological parameters, the antenna height and the antenna elevation angle.

[0058] Example One

[0059] To solve the technical problems in the prior art, an embodiment of the present application provides a wireless communication signal remote interference discrimination method.

[0060] Figure 1 A flowchart of the wireless communication signal remote interference discrimination method of the embodiment one of the present application is shown; Figure 2 A part of the process diagram of the wireless communication signal remote interference discrimination method of the embodiment one of the present application is shown; reference Figure 1 and Figure 2 The wireless communication signal remote interference discrimination method of the embodiment of the present application comprises the following steps.

[0061] In step S101, the parameter information of the to-be-detected signal is acquired, and the parameter information of the to-be-detected signal is normalized to obtain to-be-detected parameter information. The parameter information of the to-be-detected signal includes meteorological information when the to-be-detected signal is received, distance information between the to-be-detected signal sending end and the base station receiving end, and network side information of the to-be-detected signal base station receiving end.

[0062] Specifically, the parameter information of the to-be-detected signal is acquired, and the parameter information of the to-be-detected signal specifically includes meteorological information when the to-be-detected signal is received, distance information between the to-be-detected signal sending end and the base station receiving end, and network side information of the to-be-detected signal base station receiving end. The meteorological information when the to-be-detected signal is received can be searched from a meteorological data network. It should be noted that the data set formats disclosed by each meteorological website are different, and the types and implementation granularities of the collected meteorological data are also different. Therefore, the selection of the meteorological information acquisition channel can be adapted according to the granularities of the distance information and the network side information.

[0063] Since the remote interference signal of the wireless communication signal mainly has an influence on the uplink signal sent by the handheld mobile device to the base station, and the to-be-detected information is generally the uplink information of the mobile device, the sending end of the to-be-detected signal is generally the mobile device, and the receiving end of the to-be-detected signal is generally the base station receiving end. The distance information between the sending end of the detection signal and the base station receiving end is generally obtained by the exact longitude and latitude information of the base station provided by the public communication website or the operator, and then the distance length between the sending mobile terminal and the base station receiving end is calculated according to the longitude and latitude information.

[0064] One way to obtain the network side information of the to-be-detected signal base station receiving end is to directly apply to the operator. The channel parameters provided by the operator include: downlink channel bandwidth, working frequency band, EARFCN, local cell number, base station longitude and latitude, cell orientation, antenna height, electronic downtilt angle, mechanical downtilt angle, path loss, and measurement time. The network side information of the to-be-detected signal base station receiving end can be directly obtained from the channel parameters provided by the operator. At the same time, since most of the data in the channel parameters provided by the operator are non-continuous parameters, it is also necessary to obtain continuous parameters through one-hot code processing. Another way to obtain the channel parameters is to use the PETOOL series of simulation platforms. PETOOL simulation platform is currently recognized as the most experimental simulation platform in the world. Another way to obtain the network side information is to use a measurement device to measure, for example, a time domain millimeter wave channel probe of Keysight (KS). After setting the signal frequency, bandwidth, height of the transmitting antenna and receiving antenna, and other parameters of the time domain millimeter wave channel probe of Keysight (KS), the relevant information can be measured and obtained.

[0065] After obtaining the parameter information of the to-be-detected signal, the meteorological information when the to-be-detected signal is received, the distance information between the to-be-detected signal sending end and the base station receiving end, and the network side information of the to-be-detected signal base station receiving end in the to-be-detected signal parameter information are respectively normalized to obtain the to-be-detected parameter information of the to-be-detected signal. The normalization processing of the parameter information of the to-be-detected signal can effectively reduce the calculation pressure of the neural network model.

[0066] In step S102, the to-be-detected parameter information is input into all interference discrimination neural networks in the interference discrimination neural network group to obtain the interference labels output by each interference discrimination neural network.

[0067] Specifically, the specific obtaining process of the interference discrimination neural network group includes the following steps.

[0068] In step S201, a training signal parameter information set is obtained, and all training signal parameter information in the training signal parameter information set is processed to obtain a training parameter information set. All training parameter information in the training parameter information set is divided into a training sub-set and a detection sub-set. The training signal parameter information includes meteorological information when a training signal is received, distance information between a training signal sending end and a base station receiving end, network side information of a training signal base station receiving end, and path loss of the training signal.

[0069] Specifically, a large amount of training signal parameter information is obtained, and all the training signal parameter information is collected into a training signal parameter information set. The training signal parameter information includes meteorological information when a training signal is received, distance information between a training signal sending end and a base station receiving end, network side information of a training signal base station receiving end, and path loss of the training signal. The meteorological information when the training signal is received, the distance information between the training signal sending end and the base station receiving end, and the network side information of the training signal base station receiving end are obtained in the same way as the meteorological information when the to-be-detected signal is received, the distance information between the to-be-detected signal sending end and the base station receiving end, and the network side information of the to-be-detected signal base station receiving end in step S101.

[0070] Further, the process of processing the training signal parameter information in the training signal parameter information set includes: respectively normalizing the meteorological information when the training signal is received, the distance information between the training signal sending end and the base station receiving end, and the network side information of the training signal base station receiving end to obtain signal parameter information; marking the path loss of the training signal in a predetermined manner to obtain an interference label; and taking the obtained signal parameter information and interference label as the training parameter information of the current training signal.

[0071] Further, the marking the path loss of the training signal in a preset manner to obtain the interference label specifically comprises: judging whether the path loss of the training signal is greater than a preset loss value, if yes, determining that the interference label of the training signal is 1, otherwise, determining that the interference label of the training signal is 0, wherein setting the interference label as 1 indicates that the current signal is interfered by a remote interference, and setting the interference label as 0 indicates that the current signal is not interfered by a remote interference. Preferably, the preset loss value can be selected as -100 dB. Meanwhile, the preset loss value can also be selected as other reasonable values.

[0072] After obtaining the training parameter information set, all training parameter information in the training parameter information set is divided into a training subset and a detection subset. For example, any 80% of the training parameter information in the training parameter information set can be set as the training subset, and the remaining 20% of the training parameter information set can be set as the detection subset. Similarly, the number of training parameter information in the training subset and the detection subset can also be set as other reasonable values.

[0073] Step S202, a plurality of preset neural networks are constructed, and all preset neural networks are trained by all training parameter information in the training subset to obtain a plurality of trained neural networks, wherein the preset neural network comprises an input layer, an output layer and a plurality of hidden layers, the neurons between all layers are connected in a one-way full interconnection manner, and a nonlinear activation function is arranged on the neurons of all hidden layers and the neurons of the output layer, the number of hidden layers and / or the number of neurons of the hidden layers of any two preset neural networks are different.

[0074] Specifically, a plurality of preset neural networks are constructed, and the preset neural network needs to meet the following conditions: comprising an input layer, an output layer and a plurality of hidden layers, the structure of the plurality of hidden layers can better represent the complex nonlinear relationship between the atmospheric waveguide variables, and improve the accuracy of the calculation result; meanwhile, the neurons between all layers are connected in a one-way full interconnection manner, which can mobilize all neurons to calculate and improve the accuracy of the output result; finally, a nonlinear activation function is arranged on the neurons of all hidden layers and the neurons of the output layer, which can better express the nonlinear relationship between the calculation parameters. It should be noted that the number of hidden layers and / or the number of neurons of the hidden layers of any two of the above preset neural networks are different, that is, the structures of all constructed preset neural networks are different, which is specifically reflected in the number of hidden layers or the number of neurons of the hidden layers.

[0075] Further optimization, the import of preset neural network training parameter information adopts batch or row-by-row input of training information, so as not to excessively consume computing resources, so that the neural network can be better applied to low-config hardware devices for calculation. And each layer of the preset neural network calculates the gradient value based on the back propagation algorithm, and adjusts the connection weight between the layers in the preset neural network according to the gradient descent principle. In order to simplify the calculation, the preset neural network increases a bias neuron in the input layer and each hidden layer, which has no corresponding input value and no activation function, and its output value is equal to -1, so it is only named as a neuron, not a real neuron. And in order to prevent the occurrence of overfitting phenomenon, the preset neural network also needs to set a proper tolerance to ensure that the model has good universality and generalization ability under a certain precision loss.

[0076] After constructing multiple pre-trial neural networks, all training parameter information in the training subset is used to train all preset neural networks respectively to obtain a training completed neural network corresponding to each preset neural network.

[0077] In step S203, all training parameter information in the detection subset is used to detect all training completed neural networks, and the interference judgment accuracy of each training completed neural network is calculated. The preset number of training completed neural networks are selected as temporary neural networks in the order of decreasing interference judgment accuracy.

[0078] Specifically, the detection process of a single training completed neural network by all training parameter information in the detection subset includes: taking the signal parameter information in all training parameter information in the detection subset as input, and inputting into the current training completed neural network respectively to obtain the predicted interference label corresponding to each signal parameter information. Then, the predicted interference label corresponding to each signal parameter information is compared with the real interference label corresponding thereto to obtain the interference judgment accuracy of the current training completed neural network. The interference judgment accuracy of all training completed neural networks is obtained by the above method. Then, all training completed neural networks are sorted in the order of decreasing interference judgment accuracy, and the preset number of training completed neural networks are selected from front to back as temporary neural networks.

[0079] In step S204, the occurrence probability of the crossover operation and the occurrence probability of the mutation operation in the evolution calculation are set, and the evolution calculation is performed on all temporary neural networks by the crossover operation or the mutation operation to obtain multiple interference discrimination neural networks, each of which has a corresponding weight, and all interference discrimination neural networks form an interference discrimination neural network group.

[0080] Specifically, after obtaining a preset number of temporary neural networks, evolutionary computation is introduced to improve the learning ability of the neural networks. Further, according to the computing capability of the current hardware environment, the probability of occurrence of the crossover operation and the probability of occurrence of the mutation operation in the evolutionary computation are set, and then all the temporary neural networks are subjected to the crossover operation based on the probability of occurrence of the crossover operation or subjected to the mutation operation based on the probability of occurrence of the mutation operation, so as to obtain a plurality of interference discrimination neural networks, each of which has a corresponding weight, and then all the interference discrimination neural networks form an interference discrimination neural network group. Figure 3 and Figure 4 The process diagrams of the mutation operation of the temporary neural networks and the process diagram of the crossover operation of two temporary neural networks are shown respectively. In order to evolve the model towards a more accurate classification direction, a suitable fitness function is selected as an index to measure the learning ability of the neural network. Further, in order to ensure the effectiveness of the crossover and mutation and at the same time ensure that the hardware device can bear the corresponding computing intensity, the number of hidden layers and the number of neurons need to be selected appropriately when the preset neural network is set.

[0081] Further, the purpose of the crossover operation is to improve the global search ability of the algorithm, and the main operation of the crossover is shown in Figure 3 . The crossover operation is to select a random number of layer structures from two temporary neural networks, and the two layer structures are exchanged with each other to generate a new neural network. It should be noted that at least one layer structure is selected, and the layer structure is selected layer by layer, and the layer structure interval cannot be skipped. Finally, a suitable fitness function is selected to determine whether the newly generated neural network has good effect, and if the effect is better than the original neural network, it is retained, otherwise it is discarded and re-iterated. The selection of the fitness function on the classification problem is usually G-mean, which well balances the accuracy of the minority samples in the data set, and its expression is:

[0082]

[0083] where TP represents true positive, FN represents false negative, TN represents true negative, and FP represents false positive.

[0084] The purpose of the mutation operation is to improve the local random search ability of the algorithm, so as to accelerate the convergence of the algorithm to the optimal solution, and the schematic diagram of the mutation operation is shown in Figure 4 . The mutation operation mainly has three ways, including randomly adding a hidden layer, randomly deleting a hidden layer, and randomly changing the number of neurons in a hidden layer. After the initial neural network is subjected to the mutation operation, a fitness function is also used to determine whether to retain the newly generated neural network.

[0085] After the interference discrimination neural network group is determined, the to-be-detected parameter information is input into all interference discrimination neural networks in the interference discrimination neural network group to obtain interference labels output by each interference discrimination neural network.

[0086] In step S103, all possible discrimination results are voted based on the interference labels output by each interference discrimination neural network and the weights of the interference discrimination neural networks, and the possible discrimination result with the most votes is taken as the interference discrimination result of the to-be-detected signal.

[0087] Specifically, all possible results of whether the to-be-detected signal is subjected to remote interference are voted based on the interference labels output by all interference discrimination neural networks and the weights of the interference discrimination neural networks, and the possible discrimination result with the most votes is taken as the interference discrimination result of the to-be-detected signal.

[0088] To further illustrate the advantages of the wireless communication signal remote interference discrimination method according to the embodiments of the present application, the following actual examples are used for illustration.

[0089] Through cooperation with China Mobile operators, the distance data and network side data of Jiangsu all-weather base stations are used to train and analyze the effectiveness of the wireless communication signal remote interference discrimination method in this detailed example, that is, one of the two base stations is taken as a sending end and the other is taken as a receiving end, and the specific information of the base stations is shown in Table 1. While obtaining the base station data, the weather data from the weather fitting software CFSv2 of the National Oceanic and Atmospheric Administration of the United States is used to match the weather information on the base station side.

[0090] Table 1, base station distance information and network side information

[0091]

[0092] The weather fitting software CFSv2 uses the most advanced global data assimilation system and perfect database today, and performs quality control and assimilation processing on observation data from various sources (ground, ship, radio sounding, wind sounding balloon, airplane, satellite, etc.), to obtain a complete reanalysis data set. It not only contains many elements, wide range, and long time period, but also is a comprehensive data set. Through it, the weather information near the base station can be obtained, and the weather information exists in multiple time granularities, including hourly, six-hourly, weekly, and monthly weather data information. According to the distance data and network side data and the refractive index factor of atmospheric waveguide, the hourly weather information is selected, and the temperature, humidity, and pressure are selected as specific weather parameters, and the specific details of the weather information are shown in Table 2.

[0093] Table 2, weather information

[0094]

[0095] To demonstrate the effectiveness of the proposed method, we selected representative classification algorithms for comparison. To ensure the rationality of the numerical results, we used data that did not overlap with the training set. To better evaluate the learning ability of the various algorithms for noise data, we maintained a consistent number of iterations. Detailed configurations for the various algorithms are shown in Table 3.

[0096] Table 3. Algorithm parameter configuration

[0097]

[0098] By comparing the model's predictions with actual data, we evaluate the classification capabilities of various algorithms for wireless communication signal interference detection. Precision and recall are used to measure the algorithm's classification ability. It's important to note that a higher number of algorithm iterations generally results in better learning outcomes. However, this also significantly increases hardware computing time, and excessive iterations result in ineffective learning. Therefore, choosing a reasonable number of iterations is crucial for more intuitively demonstrating the algorithm's interference detection capabilities.

[0099] Table 4 shows the interference discrimination capabilities of various algorithms on multiple datasets with varying data volumes. Comparison revealed that the neural network based on evolutionary computation (i.e., the interference discrimination neural network group in this application) significantly outperformed other algorithms in terms of recall while maintaining accuracy. The recall rate was approximately 10% higher than that of other algorithms, demonstrating that the present invention's remote interference discrimination method for wireless communication signals is an effective method for identifying interference with wireless communication signals.

[0100] Table 4. Interference discrimination ability of the algorithm on data sets with different data volumes

[0101]

[0102] Table 5 shows the interference discrimination capabilities of various algorithms on datasets with varying imbalance ratios. A comparison reveals that the neural network based on evolutionary computation (i.e., the interference discrimination neural network group in this application) significantly outperforms other algorithms in terms of recall while maintaining accuracy, with a recall rate approximately 15% higher than that of other algorithms. Furthermore, the recall rate of the neural network based on evolutionary computation decreases the slowest as the imbalance ratio increases. This demonstrates that even when a dataset contains relatively little interfered data, the wireless communication signal remote interference discrimination method of the present invention can effectively learn the characteristics of the interfered data and make accurate judgments.

[0103] Table 5. Interference discrimination ability of the algorithm on datasets with different imbalance ratios

[0104]

[0105] The wireless communication signal remote interference discrimination method provided by the embodiment of the present application constructs multiple neural networks for interference discrimination, and respectively trains the neural networks based on meteorological information when a signal is received, distance information of a sending end and a receiving end, network side information of the receiving end, and path loss as data basis, then selects several neural networks with higher accuracy for evolution calculation, so as to improve the learning ability of the neural networks, and further improve the accuracy and reliability of the neural networks in discriminating whether a signal is interfered by wireless communication signal remote interference. The method of the present application combines the neural networks with evolution calculation, and realizes fast and efficient wireless communication signal remote interference discrimination with relatively high accuracy and reliability.

[0106] Example Two

[0107] To solve the above technical problems in the prior art, the embodiment of the present application further provides a wireless communication signal remote interference discrimination device in a wireless communication network.

[0108] Figure 5 The structure schematic diagram of the wireless communication signal remote interference discrimination device according to the second embodiment of the present application is shown; as shown in Figure 5 The wireless communication signal remote interference discrimination device according to the embodiment of the present application includes a data acquisition module, a data operation module, and an interference discrimination result determination module.

[0109] The data acquisition module is used for acquiring parameter information of a to-be-detected signal, and performing normalization processing on the parameter information of the to-be-detected signal to obtain to-be-detected parameter information. The parameter information of the to-be-detected signal includes meteorological information when the to-be-detected signal is received, distance information of a sending end of the to-be-detected signal and a base station receiving end, and network side information of a base station receiving end of the to-be-detected signal.

[0110] The data operation module is used for inputting the to-be-detected parameter information into all interference discrimination neural networks of an interference discrimination neural network group, so as to obtain interference labels output by each interference discrimination neural network.

[0111] The interference discrimination result determination module is used for voting all possible discrimination results based on the interference labels output by each interference discrimination neural network and the weights of the interference discrimination neural networks, and taking the possible discrimination result with the most votes as the interference discrimination result of the to-be-detected signal.

[0112] The wireless communication signal remote interference discrimination device provided by the embodiment of the present application comprises a plurality of neural networks for interference discrimination, and the weather information when the signal is received, the distance information of the sending end and the receiving end, the network side information of the receiving end and the path loss are taken as data basis to train the neural networks respectively, and then several neural networks with higher accuracy are selected to perform evolution calculation, so as to improve the learning ability of the neural network, and further improve the accuracy and reliability of the neural network in discriminating whether the signal is interfered by remote interference.

[0113] Example Three

[0114] To solve the above technical problems in the prior art, the embodiment of the present application further provides a storage medium storing a computer program, which can realize all steps in the wireless communication signal remote interference discrimination method of the embodiment one when executed by a processor.

[0115] The specific steps of the wireless communication signal remote interference discrimination method and the beneficial effects obtained by using the readable storage medium provided by the embodiment of the present application are the same as those of the embodiment one, and will not be repeated here.

[0116] It should be noted that the storage medium includes ROM, RAM, magnetic disc or optical disc and various media capable of storing program codes.

[0117] Example Four

[0118] To solve the above technical problems in the prior art, the embodiment of the present application further provides a terminal.

[0119] Figure 6 The structure schematic diagram of the terminal of the embodiment four of the present application is shown, and the terminal comprises a processor and a memory connected with each other. Figure 6 The terminal of the present embodiment comprises a processor and a memory connected with each other; the memory is used for storing a computer program, and the processor is used for executing the computer program stored in the memory, so that the terminal can realize all steps in the CO and CO2 emission proportion coefficient estimation method of the embodiment one when executed.

[0120] The specific steps of the CO and CO2 emission proportion coefficient estimation method and the beneficial effects obtained by using the terminal provided by the embodiment of the present application are the same as those of the embodiment one, and will not be repeated here.

[0121] It should be noted that the memory can include a random access memory (RAM) and can also include a non-volatile memory, such as at least one disk memory. Similarly, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., and can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component.

[0122] Although the disclosed embodiments of the present application are as above, the above description is only for the purpose of facilitating the understanding of the present application and is not intended to limit the present application. Any modification and change in the form and details of the embodiments of the present application can be made by any person skilled in the art without departing from the spirit and scope of the present application, and the protection scope of the present application shall be subject to the scope defined by the appended claims.

Claims

1. A method for remote interference identification of wireless communication signals, applicable to atmospheric waveguide interference, comprising: Acquire parameter information of a signal to be detected, and normalize the parameter information of the signal to be detected to obtain the parameter information to be detected, wherein the parameter information of the signal to be detected includes meteorological information when the signal to be detected is received, distance information between a transmitting end of the signal to be detected and a receiving end of a base station, and network side information of the receiving end of the base station of the signal to be detected; Inputting the parameter information to be detected into all interference discrimination neural networks in the interference discrimination neural network group respectively to obtain interference labels output by each interference discrimination neural network; Based on the interference labels output by each interference discrimination neural network and the weight of the interference discrimination neural network, all possible discrimination results are voted, and the possible discrimination result with the largest number of votes is used as the interference discrimination result of the signal to be detected.

2. The method according to claim 1, characterized in that The interference discrimination neural network group acquisition process includes: Obtaining a training signal parameter information set, and processing all training signal parameter information in the training signal parameter information set respectively to obtain a training parameter information set, and dividing all training parameter information in the training parameter information set into a training subset and a detection subset, wherein the training signal parameter signal includes meteorological information when receiving the training signal, distance information between the training signal transmitting end and the base station receiving end, network side information of the training signal base station receiving end, and path loss of the training signal; Constructing multiple preset neural networks, and training all of the preset neural networks respectively using all of the training parameter information in the training subset to obtain multiple trained neural networks, wherein the preset neural networks include an input layer, an output layer, and multiple hidden layers, neurons between all layers are connected in a one-way fully interconnected manner, and nonlinear activation functions are set on neurons in all hidden layers and neurons in the output layer, and any two preset neural networks have different numbers of hidden layers and / or different numbers of neurons in the hidden layers; All the trained neural networks are detected using all the training parameter information in the detection subset, and the interference judgment accuracy of each trained neural network is calculated, and a preset number of trained neural networks are selected in descending order of the interference judgment accuracy as temporary neural networks; The probability of occurrence of a crossover operation and a mutation operation in the evolutionary calculation is set, and evolutionary calculation is performed on all the temporary neural networks through the crossover operation or the mutation operation to obtain multiple interference discrimination neural networks, and each of the interference discrimination neural networks has a corresponding weight, and all the interference discrimination neural networks form an interference discrimination neural network group.

3. The method according to claim 2, characterized in that Processing the training signal parameter information includes: Normalizing the meteorological information when the training signal is received, the distance information between the training signal transmitting end and the base station receiving end, and the network side information of the training signal base station receiving end to obtain signal parameter information; Marking the path loss of the training signal in a preset manner to obtain an interference label; The training parameter information includes the signal parameter information and the interference label.

4. The method according to claim 3, characterized in that Marking the path loss of the training signal in a preset manner to obtain an interference label includes: Determine whether the path loss of the training signal is greater than a preset loss value. If so, determine that the interference label of the training signal is 1; otherwise, determine that the interference label of the training signal is 0, where the interference label is 1, indicating that the current signal is subject to remote interference, and the interference label is 0, indicating that the current signal is not subject to remote interference.

5. The method according to claim 2, characterized in that Each layer of the preset neural network calculates a gradient value based on a back-propagation algorithm, and adjusts the connection weights between layers in the preset neural network according to the gradient descent principle.

6. The method according to claim 2, characterized in that A bias neuron is added to the input layer and each hidden layer of the preset neural network.

7. The method according to claim 2, characterized in that During the evolutionary computation process, the fitness function G-mean is used as an indicator to measure the learning ability of the neural network after the training.

8. A device for remotely detecting interference of wireless communication signals in a wireless communication network, characterized in that: Applicable to atmospheric waveguide interference, the device includes a data acquisition module, a data calculation module and an interference discrimination result determination module; The data acquisition module is used to obtain parameter information of the signal to be detected and normalize the parameter information of the signal to be detected to obtain the parameter information to be detected, wherein the parameter information of the signal to be detected includes meteorological information when the signal to be detected is received, distance information between the transmitting end of the signal to be detected and the receiving end of the base station, and network side information of the receiving end of the base station of the signal to be detected; The data operation module is used to input the parameter information to be detected into all interference discrimination neural networks of the interference discrimination neural network group respectively to obtain the interference label output by each interference discrimination neural network; The interference discrimination result determination module is used to vote on all possible discrimination results based on the interference labels output by each interference discrimination neural network and the weight of the interference discrimination neural network, and use the possible discrimination result with the most votes as the interference discrimination result of the signal to be detected.

9. A storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for remote interference determination of wireless communication signals according to any one of claims 1 to 7 is implemented.

10. A terminal, characterized in that: include: a processor and a memory, wherein the memory is communicatively connected to the processor; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes the wireless communication signal remote interference determination method according to any one of claims 1 to 7.

Images