WO2023097636A1

WO2023097636A1 - Data processing method and apparatus

Info

Publication number: WO2023097636A1
Application number: PCT/CN2021/135255
Authority: WO
Inventors: 刘文东; 田文强
Original assignee: Oppo广东移动通信有限公司
Priority date: 2021-12-03
Filing date: 2021-12-03
Publication date: 2023-06-08
Also published as: CN117916732A

Abstract

A data processing method and apparatus. The method comprises: acquiring channel data of a virtual channel that is generated by a channel generator of a generative adversarial network (S610); extracting a first channel statistical feature of the virtual channel on the basis of the channel data of the virtual channel (S620); extracting a second channel statistical feature of a real channel on the basis of channel data of the real channel (S630); and determining the difference between the first channel statistical feature and the second channel statistical feature (S640). The quality of channel data of a virtual channel that is generated by a channel generator in a generative adversarial network is measured by means of comparing a first channel statistical feature of the virtual channel with a second channel statistical feature of a real channel to determine the difference therebetween. Compared with directly comparing channel data of the real channel with the channel data of the virtual channel, the measurement mode is beneficial for more accurately measuring the quality of the channel data of the virtual channel, and is beneficial for improving the accuracy of a channel model established on the basis of the channel data of the virtual channel.

Description

Method and device for data processing

technical field

The present application relates to the technical field of communications, and more specifically, to a data processing method and device.

Background technique

Currently, channel generators in generative adversarial networks (GAN) are used to generate channel data for virtual channels as the basis for channel modeling. Specifically, the channel generator of GAN and the channel discriminator of GAN can be trained alternately, so that the channel generator can generate the channel data of the virtual channel close to the channel data of the real channel, and use the channel data of the virtual channel to distinguish the real channel The channel data is expanded to provide a data basis for building a channel model. Therefore, the similarity between the channel data of the virtual channel generated by the channel generator and the channel data of the real channel (that is, the data quality of the channel data of the virtual channel) directly affects whether the established channel model can accurately describe or characterize the real channel. . However, there is currently no effective way to evaluate the quality of channel data for virtual channels.

Contents of the invention

The present application provides a data processing method and device for evaluating the quality of channel data of a virtual channel, so as to improve the accuracy of a channel model established based on the channel data of the virtual channel.

In a first aspect, a data processing method is provided, including: acquiring channel data of a virtual channel generated by a channel generator generating an adversarial network; extracting a first channel statistical feature of the virtual channel based on the channel data of the virtual channel ; extracting a second channel statistical feature of the real channel based on channel data of the real channel; determining a difference between the first channel statistical feature and the second channel statistical feature.

In a second aspect, a data processing method is provided, including: obtaining channel data of a virtual channel generated by a channel generator of an adversarial generation network; based on a first difference between the channel data of the virtual channel and the channel data of a real channel , determines whether to save the channel builder.

In a third aspect, a data processing device is provided, including: an acquisition unit, configured to acquire channel data of a virtual channel generated by a channel generator generating an adversarial network; a processing unit, configured to extract the channel data based on the virtual channel The first channel statistical feature of the virtual channel; the processing unit is also used to extract the second channel statistical feature of the real channel based on the channel data of the real channel; the processing unit is also used to determine the first channel The difference between the statistical characteristics and the second channel statistical characteristics.

According to the fourth aspect, there is provided a data processing device, which is characterized in that it includes: an acquisition unit, configured to acquire channel data of a virtual channel generated by a channel generator of an adversarial generation network; a processing unit, configured to obtain channel data based on the virtual channel A first difference between the channel data and the channel data of the real channel determines whether to save the channel generator.

In a fifth aspect, there is provided a data processing device, including a processor and a memory, the memory is used to store one or more computer programs, and the processor is used to call the computer programs in the memory to make the data processing device execute Part or all of the steps in the above method.

In a sixth aspect, a network device is provided, including a processor, a memory, and a communication interface, the memory is used to store one or more computer programs, and the processor is used to invoke the computer programs in the memory to make the network device Perform some or all of the steps in the above methods.

In a seventh aspect, the embodiment of the present application provides a communication system, where the system includes the above-mentioned terminal and/or network device. In another possible design, the system may further include other devices that interact with the terminal or network device in the solutions provided by the embodiments of the present application.

In an eighth aspect, the embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and the computer program causes a terminal to execute part or all of the steps in the above method.

In the ninth aspect, the embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause the terminal to execute the above method Some or all of the steps in . In some implementations, the computer program product can be a software installation package.

In a tenth aspect, the embodiment of the present application provides a chip, the chip includes a memory and a processor, and the processor can call and run a computer program from the memory to implement some or all of the steps described in the above method.

In the embodiment of the present application, by comparing the difference between the first channel statistical characteristics of the virtual channel and the second channel statistical characteristics of the real channel, the quality of the channel data of the virtual channel generated by the channel generator in the generated confrontation network is measured, Compared with directly comparing the channel data of the real channel with the channel data of the virtual channel, the measurement method in the embodiment of the present application is conducive to more accurately measuring the quality of the channel data of the virtual channel, and is conducive to improving the establishment of channel data based on the virtual channel. Accuracy of the channel model.

Description of drawings

FIG. 1 is a wireless communication system 100 applied in an embodiment of the present application.

FIG. 2 is a schematic diagram of a GAN architecture applicable to an embodiment of the present application.

Fig. 3 is a schematic diagram of the architecture of the GAN of the embodiment of the present application.

Fig. 4 is a schematic diagram of a channel generator applicable to an embodiment of the present application.

Fig. 5 is a schematic diagram of a channel discriminator according to an embodiment of the present application.

FIG. 6 is a flowchart of a data processing method according to an embodiment of the present application.

Fig. 7 is a schematic diagram of a channel evaluator according to an embodiment of the present application.

Fig. 8 is a schematic diagram of the GAN training process of the embodiment of the present application.

FIG. 9 is a schematic diagram of a GAN training process according to another embodiment of the present application.

Fig. 10 is a schematic diagram of the statistical average result of the power delay spectrum of the virtual channel after the first round of training.

Fig. 11 is a schematic diagram of a single sample result of the power antenna spectrum of a virtual channel after the first round of training.

Fig. 12 shows a schematic diagram of the statistical average result of the power delay spectrum of the virtual channel after the 1200th round of training.

Fig. 13 shows a schematic diagram of a single sample result of the power antenna spectrum of a virtual channel after the 1200th round of training.

Fig. 14 shows a schematic diagram of the statistical average result of the power delay spectrum of a real channel.

Fig. 15 shows a schematic diagram of a single sample result of the power antenna spectrum of a real channel.

FIG. 16 is a schematic diagram of a data processing device according to an embodiment of the present application.

Fig. 17 is a schematic diagram of a data processing device according to another embodiment of the present application.

Fig. 18 is a schematic structural diagram of a data device according to an embodiment of the present application.

Detailed ways

The technical solution in this application will be described below with reference to the accompanying drawings. For ease of understanding, the system and the terms involved in this embodiment of the present application are firstly introduced below with reference to FIG. 1 to FIG. 4 .

FIG. 1 is a wireless communication system 100 applied in an embodiment of the present application. The wireless communication system 100 may include a network device 110 and a terminal device 120 . The network device 110 may be a device that communicates with the terminal device 120 . The network device 110 can provide communication coverage for a specific geographical area, and can communicate with the terminal device 120 located in the coverage area.

Figure 1 exemplarily shows one network device and two terminals. Optionally, the wireless communication system 100 may include multiple network devices and each network device may include other numbers of terminal devices within the coverage area. The embodiment does not limit this.

Optionally, the wireless communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in this embodiment of the present application.

It should be understood that the technical solutions of the embodiments of the present application can be applied to various communication systems, for example: the fifth generation (5th generation, 5G) system or new radio (new radio, NR), long term evolution (long term evolution, LTE) system , LTE frequency division duplex (frequency division duplex, FDD) system, LTE time division duplex (time division duplex, TDD), etc. The technical solutions provided in this application can also be applied to future communication systems, such as the sixth generation mobile communication system, and satellite communication systems, and so on.

The terminal equipment in the embodiment of the present application may also be called user equipment (user equipment, UE), access terminal, subscriber unit, subscriber station, mobile station, mobile station (mobile station, MS), mobile terminal (mobile terminal, MT) ), remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user device. The terminal device in the embodiment of the present application may be a device that provides voice and/or data connectivity to users, and can be used to connect people, objects and machines, such as handheld devices with wireless connection functions, vehicle-mounted devices, and the like. The terminal device in the embodiment of the present application can be mobile phone (mobile phone), tablet computer (Pad), notebook computer, palmtop computer, mobile internet device (mobile internet device, MID), wearable device, virtual reality (virtual reality, VR) equipment, augmented reality (augmented reality, AR) equipment, wireless terminals in industrial control, wireless terminals in self driving, wireless terminals in remote medical surgery, smart Wireless terminals in smart grid, wireless terminals in transportation safety, wireless terminals in smart city, wireless terminals in smart home, etc. Optionally, UE can be used to act as a base station. For example, a UE may act as a scheduling entity that provides sidelink signals between UEs in V2X or D2D, etc. For example, a cell phone and an automobile communicate with each other using sidelink signals. Communication between cellular phones and smart home devices without relaying communication signals through base stations.

The network device in this embodiment of the present application may be a device for communicating with a terminal device, and the network device may also be called an access network device or a wireless access network device, for example, the network device may be a base station. The network device in this embodiment of the present application may refer to a radio access network (radio access network, RAN) node (or device) that connects a terminal device to a wireless network. The base station can broadly cover various names in the following, or replace with the following names, such as: Node B (NodeB), evolved base station (evolved NodeB, eNB), next generation base station (next generation NodeB, gNB), relay station, Access point, transmission point (transmitting and receiving point, TRP), transmission point (transmitting point, TP), primary station MeNB, secondary station SeNB, multi-standard wireless (MSR) node, home base station, network controller, access node , wireless node, access point (access point, AP), transmission node, transceiver node, base band unit (base band unit, BBU), remote radio unit (Remote Radio Unit, RRU), active antenna unit (active antenna unit) , AAU), radio head (remote radio head, RRH), central unit (central unit, CU), distributed unit (distributed unit, DU), positioning nodes, etc. A base station may be a macro base station, a micro base station, a relay node, a donor node, or the like, or a combination thereof. A base station may also refer to a communication module, modem or chip used to be set in the aforementioned equipment or device. The base station can also be a mobile switching center, a device that undertakes the function of a base station in D2D, vehicle-to-everything (V2X), machine-to-machine (M2M) communication, and a device in a 6G network. Network-side equipment, equipment that assumes base station functions in future communication systems, etc. Base stations can support networks of the same or different access technologies. The embodiment of the present application does not limit the specific technology and specific device form adopted by the network device.

Base stations can be fixed or mobile. For example, a helicopter or drone can be configured to act as a mobile base station, and one or more cells can move according to the location of the mobile base station. In other examples, a helicopter or drone may be configured to serve as a device in communication with another base station.

In some deployments, the network device in this embodiment of the present application may refer to a CU or a DU, or, the network device includes a CU and a DU. A gNB may also include an AAU.

Network equipment and terminal equipment can be deployed on land, including indoors or outdoors, hand-held or vehicle-mounted; they can also be deployed on water; they can also be deployed on aircraft, balloons and satellites in the air. In the embodiment of the present application, the scenarios where the network device and the terminal device are located are not limited.

It should be understood that all or part of the functions of the communication device in this application may also be realized by software functions running on hardware, or by virtualization functions instantiated on a platform (such as a cloud platform).

Channel real mining

The use and understanding of channels (for example, wireless channels) is the basis for building communication systems, and the most direct way to study the channel state of channels is to collect channel data of actual channels (also known as "real channels"), and based on The collected channel data is used to determine the channel state information of the actual channel to assist the design of the communication system.

In some implementation manners, channel state information of an actual channel may be obtained by performing channel estimation on a channel used for signal transmission between a signal transmitter and a signal receiver. In other implementations, when the transmitter communicates with the receiver, a specific receiver (also known as a "third-party receiver") can be used to collect the signal transmitted by the transmitter (such as a network device of a cellular network), so as to obtain Channel state information of the actual channel.

Traditional Channel Modeling

In the process of actual channel acquisition, considering the difficulty and cost of actual acquisition, it is difficult to implement actual acquisition of channel states in all communication scenarios. For example, when the communication scenario is underwater communication, it will increase the difficulty and cost of collecting channel state information. Therefore, traditional channel modeling can only use limited channel data to extract the channel statistical characteristics of the real channel on the basis of actual channel sampling, and build a channel model based on the channel statistical characteristics of the real channel. However, due to limited channel data, the channel model established by the traditional channel modeling method cannot describe or describe the real channel well.

In some implementation manners, the above-mentioned channel statistical features may be channel features related to transmission in channel features, for example, large-scale parameters of the channel, small-scale parameters of the channel, multipath information of the channel, time-delay power spectral density of the channel, The transmission launch angle of the channel, the arrival angle of the channel, etc.

Generative adversarial network (GAN)

As the most promising branch of unsupervised learning methods in recent years, GAN has attracted widespread attention in the field of artificial intelligence (AI). The following describes the architecture of the GAN applicable to the embodiment of the present application with reference to FIG. 2 . The GAN shown in FIG. 2 includes a generator (generator, G) 210 and a discriminator (discriminator, D) 220 . The generator 210 may output dummy data (also referred to as "fake data") based on the input latent variables to simulate the distribution of real data. The discriminator 220 outputs a discriminative result based on the input data, and the discriminative result is used to discriminate whether the input data is real data from a real-world database or virtual data generated by the generator 210 .

In the process of training the GAN, the generator 210 and the discriminator 220 need to be trained simultaneously. The training process of the generator 210 is to maximize the error probability of the discriminator 220, and the training process of the discriminator 220 is to minimize the judgment error probability under the premise of a fixed generator. Through the mutual game between the generator 210 and the discriminator 220, under the premise of stable convergence of GAN, the virtual data output by the generator 210 can simulate real data, or in other words, the virtual data output by the generator 210 can be closer to real data. To achieve the effect of real ones. Therefore, the virtual data generated by the trained generator 210 can be used to supplement real data to form a larger-scale data set.

In this way, based on the purpose that the virtual data generated by the generator 210 can supplement the real data, the generator 210 can be applied to the channel modeling process to generate virtual channel data that is sufficiently fake to make up for the lack of real channel data introduced above. Insufficient quantity leads to the defect that the channel model cannot characterize or describe the real channel.

In the following, in conjunction with Figure 3 to Figure 5, the application of GAN to generate channel data will be introduced as an example. Fig. 3 is a schematic diagram of the architecture of the GAN of the embodiment of the present application. The GAN architecture shown in FIG. 3 includes a generator 310 and a discriminator 320 . It should be noted that, in the scenario of generating channel data, the generator 310 can also be called a "channel generator", and the discriminator 320 can also be called a "channel discriminator".

As mentioned above, in the process of training the GAN, the channel generator 310 and the channel discriminator 320 need to be trained simultaneously. The training target of the channel generator 320 is to generate the channel data of the virtual channel similar to the channel data of the real channel. The training goal of the channel discriminator 330 is to improve the accuracy of identifying whether the input channel data is channel data of a virtual channel or channel data of a real channel.

The channel generator 310 described above generates channel data of a virtual channel based on input data, and inputs the channel data of the virtual channel to the channel discriminator 320 .

The embodiment of the present application does not limit the dimensions of the above-mentioned input data. For example, the dimension of the above input data may be one-dimensional, two-dimensional or a preset higher dimension. For another example, the dimension of the input data may also be the same as the dimension of the channel data of the real channel. Of course, the above input data may also be generated by cutting and/or concatenating the above several data in different dimensions. The dimensions of the above-mentioned input data may include, for example, one or more parameters among the transmission delay of the channel, the number of transmitting antennas of the transmitting channel, the number of receiving antennas of the receiving channel, and the width of the frequency domain granularity in the channel. The frequency domain granularity (also called frequency domain unit) may include subcarriers, resource blocks (resource block, RB), subbands, and the like.

The embodiment of the present application does not limit the length of the above input data. The dimension of the input data may refer to the length (or width) of the input data in a certain dimension. For example, the length of the input data may be an arbitrarily set integer value, and for another example, the length of the input data may also be consistent with a certain dimension of the channel data of the real channel.

The embodiment of the present application does not limit the content of the input data. In some implementation manners, the content of the input data may include one or more of noise, random sequence, and conditional distribution of noise or random number sequence containing channel statistical information.

The above noise distribution p(z) may be a superposition of one or more probability distributions. The probability distribution may include random distributions such as Gaussian distribution, uniform distribution, Poisson distribution, and negative exponential distribution, for example.

The aforementioned random number sequence may be a positive integer random number sequence within a certain range, for example, an integer random sampling of [0,10]. The random number sequence can also be a decimal random number sequence within a certain range, for example, an integer random sampling of [0,10] is divided by 10. Certainly, the random number sequence may also be a pseudorandom number sequence constructed by a pseudorandom number sequence generator.

The noise or random number sequence conditional distribution containing channel statistical information may be, for example, the inner product of the correlation matrix of the channel data and the noise vector in the channel data ensemble of the real channel. For another example, the conditional distribution of noise or random number sequence containing channel statistical information may be in the form of noise distribution constructed according to the variance and/or mean value of the channel data in the channel data ensemble of the real channel. The above noise or random number sequence conditional distribution containing channel statistical information can be input as a kind of conditional distribution containing prior information.

The channel discriminator 320 is used for discriminating the input channel data to determine the discriminating result.

The above-mentioned channel data input to the channel discriminator 320 may be channel data of a virtual channel or channel data of a real channel. In some implementation manners, the channel data of the above-mentioned real channels may be directly acquired from the channel data ensemble of real channels of real channels. Of course, in other implementations, if the amount of data contained in the full set of channel data of the real channel is large, batch sampling can also be used to extract a batch of real channel data from the full set of channel data of the real channel, and then from this batch The channel data of the real channel is acquired from the real channel data as the input of the channel discriminator 320 .

Usually, in order to improve the identification accuracy of the channel discriminator 320, the channel data of the above-mentioned real channel and the channel data of the virtual channel are usually relatively close, for example, the channel data of the real channel and the channel data of the virtual channel can be consistent in dimension and length . For ease of understanding, the following uses a real channel as an example to introduce the representation of channel data. The identification manner of the channel data of the virtual channel may refer to the representation manner of the channel data of the real channel. For the sake of brevity, no further details are given below.

Assume that K represents the number of users contained in the full set of channel data of the real channel. T represents the number of time slots used by each user when sampling channel data, and B represents the batch size during batch training. N _t represents the number of transmit antennas used in the transmit channel, N _r represents the number of receive antennas used in the receive channel, N _d represents the number of delay paths in the channel (or in other words, the number of different transmission delays contained in the channel ), N _f represents the number of frequency-domain granularities contained in the channel.

Correspondingly, the set size of the full set H of channel data of the above real channel is K×T. For the time-domain channel, the elements contained in the channel data set H of the real channel are denoted as channel

Among them, the collection

Represents a set of real numbers, N _t , N _r , and N _d are respectively the first dimension, the second dimension, and the third dimension of the elements in the channel data set H of the real channel, and 2 indicates that the complex channel is split by the real part and the imaginary part, or The split of magnitude and phase acts as the fourth dimension. For frequency domain channels, the elements contained in the full set of channel data H of real channels can be denoted as channel

as, among them, collection

Represents a set of real numbers, N _t , N _r , and N _f are respectively the first dimension, the second dimension, and the third dimension of the elements in the channel data set H of the real channel, and 2 indicates that the complex channel is split by the real part and the imaginary part, or The split of magnitude and phase acts as the fourth dimension.

It should be noted that the above is only an example of a representation of elements in the full set of channel data for real channels, and elements in the full set of channel data for real channels may also use other representations. For example, a new first dimension can be formed based on the dimensions N _t and N _r of the above elements, namely

or

to reduce the number of dimensions of the elements. In addition, the dimension of an element may also include other dimensions, such as angle dimension, polarization dimension, number of symbols in time domain, and so on.

The above describes GAN as a whole with reference to FIG. 2 and FIG. 3 , and the architecture of the channel generator 310 and channel discriminator 320 in GAN is introduced below in conjunction with FIG. 4 and FIG. 5 .

Fig. 4 is a schematic diagram of a channel generator applicable to an embodiment of the present application. The channel generator 310 shown in Figure 4 can be a neural network model, for example, can be a convolutional neural network (convolutional neural networks, CNN), a recurrent neural network (recurrent neural network, RNN), a deep neural network (deep neural networks, DNN) and so on.

Continuing to refer to FIG. 4 , the neural network in the channel generator 310 can be divided into three types according to the positions of different layers: an input layer 410 , a hidden layer 420 and an output layer 430 . Generally speaking, the first layer is the input layer 410 , the last layer is the output layer 430 , and the middle layer between the first layer and the last layer is the hidden layer 420 .

The input layer 410 is used for input data, wherein the input data may be, for example, the input data of the channel generator 410 . The hidden layer 420 is used to process the input data. The output layer 430 is used to output processed output data, for example, channel data of a virtual channel.

As shown in Figure 4, the neural network includes multiple layers, each layer includes multiple neurons, and the neurons between layers can be fully connected or partially connected. For connected neurons, the output of neurons in the previous layer can be used as the input of neurons in the next layer.

It should be noted that, in the embodiment of the present application, the structure of each layer in the neural network model in the channel generator 310 is not limited. For example, the above hidden layer 420 may include multiple convolutional layers. For another example, the above hidden layer 420 may include multiple fully connected layers. For another example, the hidden layer 420 may include a convolutional layer and a pooling layer arranged at intervals. In addition, the embodiment of the present application does not specifically limit the model parameters (for example, the number of neurons, the number of channels, the size of the convolution kernel, etc.) in the neural network model in the channel generator 310 .

Fig. 5 is a schematic diagram of a channel discriminator according to an embodiment of the present application. The channel discriminator 320 shown in FIG. 5 may be a neural network model, for example, CNN, RNN, DNN and so on.

Continuing to refer to FIG. 5 , the neural network in the channel discriminator 320 can be divided into three types according to the positions of different layers: an input layer 510 , a hidden layer 520 and an output layer 530 . Generally speaking, the first layer is the input layer 510 , the last layer is the output layer 530 , and the middle layer between the first layer and the last layer is the hidden layer 520 .

The input layer 510 is used for input data, wherein the input data may be, for example, the input data of the channel generator 510 . The hidden layer 520 is used to process the input data. The output layer 530 is used to output processed output data, for example, channel data of a virtual channel.

As shown in Figure 5, the neural network includes multiple layers, each layer includes multiple neurons, and the neurons between layers can be fully connected or partially connected. For connected neurons, the output of neurons in the previous layer can be used as the input of neurons in the next layer.

Usually, the output layer 530 can also be provided with a loss function, which is used to calculate the prediction error, or to evaluate the degree of difference between the output result of the neural network model (also known as the predicted value) and the ideal result (also known as the real value). . In the embodiment of the present application, the above loss function may be used to evaluate the difference between the channel data of the real channel and the channel data of the virtual channel.

In one implementation, the above loss function may be a cross-entropy loss function, wherein the label corresponding to the channel data of the real channel is 1, and the label corresponding to the channel data of the virtual channel is 0, and the output layer 530 of the channel discriminator 320 may use a sigmoid activation function, and use the cross-entropy loss function V(D,G) to optimize the channel generator 310 and channel discriminator 320, wherein,

p(z) represents the distribution of the input information of the channel generator 310, and p(H) represents the distribution of the channel data of the real channel.

It can be seen from the above loss function that during the training process, the training purpose of the channel discriminator 320D is to maximize the cross-entropy loss function, so that the channel discriminator 320 can better distinguish the channel data of the real channel and the channel of the virtual channel data. The training purpose of the channel generator 310G is to minimize the cross-entropy loss function, so that the channel data of the virtual channel generated by the channel generator 310 is closer to the channel data of the real channel, so that the identification result of the channel discriminator D is wrong.

In another implementation, bulldozer distance can be used as the loss function. Among them, the bulldozer distance can be expressed as:

in,

Indicates the shortest working distance required from the input information distribution p(z) to the channel data distribution p(H) of the real channel.

It can be seen from the above loss function that during the training process, the training purpose of the channel discriminator 320D is to make the distribution p(z) of the channel data G(z) of the generated virtual channel different from the distribution p(H) of the channel data of the real channel. ) distance (or in other words, maximize the bulldozer distance), so that the channel discriminator 320 can better distinguish the channel data of the real channel and the channel data of the virtual channel. On the contrary, the training purpose of the channel generator 310G is to minimize the distance between the distribution p(z) of the channel data G(z) of the generated virtual channel and the distribution p(H) of the channel data of the real channel (or in other words, make the bulldozer The distance is minimized), so that the channel data of the virtual channel generated by the channel generator 310 is closer to the channel data of the real channel, so that the identification result output by the channel discriminator 320D is wrong.

It should be noted that, in the embodiment of the present application, the structure of each layer in the neural network model in the channel discriminator 320 is not limited. For example, the above hidden layer 520 may include multiple convolutional layers. For another example, the hidden layer 520 may include multiple fully connected layers. For another example, the hidden layer 520 may include a convolutional layer and a pooling layer arranged at intervals. In addition, the embodiment of the present application does not specifically limit the model parameters (for example, the number of neurons, the number of channels, the size of the convolution kernel, etc.) in the neural network model in the channel discriminator 320 .

In addition, the channel discriminator 320 and the channel generator 310 may adopt the same model structure. Of course, the channel discriminator 320 and the channel generator 310 may also adopt different model structures, which is not limited in this embodiment of the present application.

Usually, the channel discriminator 320 is only used to discriminate the channel data of the real channel and the channel data of the virtual channel during the training phase of the GAN. In the actual deployment stage, only the channel generator 310 may be deployed without deploying the channel discriminator 320 . Of course, if it is desired that the GAN supports online training, the channel generator 310 and the channel discriminator 320 can also be used at the same time.

Based on the above introduction, it can be seen that by using GAN to alternately train the channel generator 310 and the channel discriminator 320, the channel generator 310 can generate the channel data of the virtual channel that is similar to the channel data of the real channel, and use the channel data of the virtual channel To expand the channel data of the real channel to provide a data basis for establishing a channel model. Therefore, the similarity between the channel data of the virtual channel generated by the channel generator and the channel data of the real channel (that is, the data quality of the channel data of the virtual channel) directly affects whether the established channel model can accurately describe or characterize the real channel. . However, there is currently no effective way to evaluate the quality of channel data for virtual channels.

Therefore, the present application provides a data processing method, by comparing the channel statistical characteristics of the virtual channel (also known as "first channel statistical characteristics") and the channel statistical characteristics of real channels (also known as "second channel statistical characteristics") The difference between them is used to evaluate the data quality of the channel data of the virtual channel generated by the channel generator.

For ease of understanding, the following six channel statistical features applicable to the embodiments of the present application are introduced first. In some implementation manners, the channel data is generally expressed in the form of a matrix, and correspondingly, the statistical features of the channel obtained through statistics based on the channel data may be expressed in the form of a correlation matrix. Of course, based on differences in statistical methods and channel data representation methods, the foregoing channel statistical features may also be represented in other forms, which is not limited in this embodiment of the present application.

The channel statistical feature 1 is used to indicate the statistical results of channel states of multiple delay paths of the channel corresponding to all users of the channel. Wherein, the statistical result may be an average value, a maximum value, a minimum value, etc. of the channel state, which is not limited in this embodiment of the present application.

For a virtual channel, the channel statistical feature 1 may indicate statistical results of channel states of multiple delay paths of the virtual channel corresponding to all users of the virtual channel.

As mentioned above, in some implementations, batch training can be used to train the GAN. Therefore, the channel data of the virtual channel generated by the channel generator is generated based on a batch of channel data of real channels. At this time, all users of the virtual channel can be understood as all users of this batch of real channels. Usually, in order to increase the diversity of the channel data of a batch of real channels, the number of users to which the channel data of a batch of real channels belongs is equal to the batch size.

For a real channel, the channel statistical feature 1 may indicate statistical results of channel states of multiple delay paths of the real channel corresponding to all users of the real channel.

The following uses the full channel correlation matrix to represent channel statistical features 1 as an example for introduction. It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the channel correlation matrix R _k, k of the dth delay path of the real channel corresponding to the kth user can be expressed as

Indicates the channel state of the real channel obtained by sampling the time slot t for the d-th delay path of the real channel corresponding to the k-th user,

Represents the conjugate transposition matrix of the matrix h _k,t,k, then the full channel correlation matrix R of the real channel indicates the mean value of the channel state of all delay paths of the real channel corresponding to all users of the real channel, that is

For a virtual channel, the full channel correlation matrix of the virtual channel

Indicates the mean value of the channel state of all delay paths of the virtual channel corresponding to all users of the virtual channel, namely

Represents the channel state of the virtual channel of the dth delay path of the virtual channel corresponding to the bth user (or the user corresponding to the bth training data in the batch training),

representation matrix

The conjugate transpose matrix of .

Channel statistical feature 2, considering that the channel states of different delay paths may be different, therefore, channel statistical feature 2 can be used to indicate the channel state of a certain delay path (also known as "target delay path") corresponding to all users of the channel As a statistical result, the channel statistical feature 2 above may also be called a delay-specific (delay-specific) channel statistical feature.

For the virtual channel, the channel statistical feature 2 may indicate the statistical result of the channel state of all users of the virtual channel corresponding to the target delay path of the virtual channel.

For the real channel, the channel statistical feature 2 may indicate the statistical result of the channel state of the target delay path of the real channel corresponding to all users of the real channel.

The following uses a correlation matrix to represent channel statistical features 2 as an example for introduction. Wherein, the correlation matrix representing the channel statistical feature 2 may be called a "delay-specific channel correlation matrix". It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the channel correlation matrix R _k ,k of the d-th delay path of the real channel corresponding to the k-th user (that is, the target delay path above) can be expressed as

h _k,t,k represent the channel state of the real channel obtained by sampling the time slot t for the d-th delay path of the real channel corresponding to the k-th user,

represents the conjugate transposition matrix of the matrix h _k,t,k, then the delay-specific channel correlation matrix R _k of the real channel indicates the mean value of the channel state of the d-th delay path corresponding to all users of the real channel, that is

For virtual channels, the delay-specific channel correlation matrix of virtual channels

Indicates the channel state statistical results of the target delay paths corresponding to all users of the virtual channel, namely

You can refer to the above channel statistical characteristics 1

introduction.

Channel statistical feature 3, considering that the channel states of different users may be different, therefore, channel statistical feature 3 can be used to indicate the channel state of the channel state of all delay paths corresponding to a certain user (also known as "first user") of the channel Statistical results. Therefore, the above channel statistical feature 3 may also be called a user-specific (UE-specific) channel statistical feature.

For a virtual channel, the channel statistical feature 3 may indicate statistical results of channel states of multiple delay paths of the virtual channel corresponding to the first user of the virtual channel.

As mentioned above, in some implementations, batch training can be used to train GAN. Therefore, the channel data of the virtual channel generated by the channel generator is generated based on a batch of channel data of real channels. At this point, the users of the virtual channel can be understood as the users of this batch of real channels.

In addition, the multiple delay paths of the above-mentioned virtual channel may be all delay paths in the virtual channel. Of course, the above-mentioned multiple delay paths may also be part of the delay paths in the virtual channel. Multiple delay paths with power greater than a preset value.

For a real channel, the channel statistical feature 3 may indicate statistical results of channel states of multiple delay paths of the real channel corresponding to the first user of the real channel.

The multiple delay paths of the above-mentioned real channel can be all the delay paths in the real channel. Of course, the above-mentioned multiple delay paths can also be part of the delay paths in the real channel. For example, it can be that the transmission power in the real channel is greater than Multiple delay paths with preset values.

It should be noted that the first user of the virtual channel and the first user of the real channel may be the same user or different users, which is not limited in this embodiment of the present application.

The following uses a correlation matrix to represent channel statistical features 3 as an example for introduction. Wherein, the correlation matrix representing the channel statistical feature 3 may be called "user-specific channel correlation matrix". It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the user-specific channel correlation matrix R _k of the real channel indicates the mean value of the channel state of the N _k delay paths of the real channel corresponding to the kth user, namely

For R _k,k, please refer to the introduction of R _k,k in the channel statistical characteristics 1 above.

For a virtual channel, the user-specific channel correlation matrix of the virtual channel

Indicates the mean value of the channel state of the N _k delay paths of the virtual channel corresponding to the bth user, namely

You can refer to the above channel statistical characteristics 1

introduction.

Channel statistical feature 4, considering that the channel state is related to users and transmission delays, therefore, the channel statistical feature 4 can be used to indicate a certain delay path (also known as " The statistical result of the channel state of the target delay path"), therefore, the above-mentioned channel statistical feature 4 can also be called user delay-specific (UE-delay-specific) channel statistical feature.

For the virtual channel, the channel statistical feature 4 may indicate the statistical result of the channel state of the target delay path of the virtual channel corresponding to the first user of the virtual channel.

For the real channel, the channel statistical feature 4 may indicate the statistical result of the channel state of the target delay path corresponding to the first user of the real channel.

The following uses a correlation matrix to represent channel statistical features 4 as an example for introduction. Wherein, the correlation matrix representing the channel statistical feature 4 may be called "a channel correlation matrix specific to user delay". It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, assuming that the target delay path is the dth delay path, the user delay-specific channel correlation matrix R _{k,k of the} real channel indicates the dth delay path of the real channel corresponding to the kth user For statistical results of the channel state of , R _k,k can refer to the introduction of R _k,k in channel statistical characteristics 1 above.

For a virtual channel, assuming that the target delay path is the dth delay path, the user delay-specific channel correlation matrix of the virtual channel

Indicates the statistical result of the channel state of the dth delay path of the virtual channel corresponding to the bth user,

You can refer to the above channel statistical characteristics 1

introduction.

Channel statistical feature 5, considering that the channel feature is related to one or more of the antenna of the transmitting channel, the antenna of the receiving channel, and the antenna pair corresponding to the channel, therefore, the channel statistical feature 5 can be used to indicate that it is related to a certain antenna or a certain The channel state of the channel corresponding to the antenna pair. Wherein, the channel corresponding to the antenna pair may be understood as a channel transmitted and received through the antenna pair. The channel corresponding to the antenna may be understood as a channel transmitted or received by the antenna, and in this case, the antenna may be a transmitting antenna or a receiving antenna.

For a virtual channel, the channel statistical feature 5 may indicate a channel state of the virtual channel corresponding to the first antenna. Alternatively, the channel statistical feature 5 may indicate the channel status of the virtual channel corresponding to the first antenna pair. In this case, the channel statistical feature 5 may also be called "transmitting-receiving antenna-specific (Tx-Rx-specific) channel statistical feature".

When the above-mentioned first antenna is a receiving antenna, the virtual channel corresponding to the first antenna can be understood as a virtual channel received through the first antenna. At this time, the channel statistical feature 5 can also be called "receiving antenna-specific (Rx-specific) channel statistics feature". When the above-mentioned first antenna is a transmitting antenna, the virtual channel corresponding to the first antenna can be understood as a virtual channel transmitted through the first antenna. At this time, the channel statistical feature 5 can also be called "transmitting antenna-specific (Tx-specific) channel statistics feature".

In some implementation manners, the above-mentioned channel statistical feature 5 may represent the channel state of the virtual channel based on a combination of the first antenna and the delay path. For example, when the first antenna is a transmitting antenna, the channel statistical feature 5 may indicate a channel state of a delay path of a virtual channel transmitted through the first antenna. For another example, when the first antenna is a transmitting antenna, the channel statistical feature 5 may indicate the channel status of multiple delay paths (for example, all or part of the delay paths included in the virtual channel) of the virtual channel transmitted through the first antenna . For another example, when the first antenna is a receiving antenna, the channel statistical feature 5 may indicate a channel state of a delay path of a virtual channel received through the first antenna. For another example, when the first antenna is a receiving antenna, the channel statistical feature 5 may indicate channel states of multiple delay paths of the virtual channel received through the first antenna (for example, it may be an average value of the channel states).

In some other implementation manners, the above-mentioned channel statistical feature 5 may represent the channel state of the virtual channel based on the combination of the first antenna and the user. For example, when the first antenna is a transmitting antenna, the channel statistical feature 5 may indicate the channel state of the virtual channel corresponding to multiple users of the virtual channel transmitted through the first antenna (for example, some or all users of the virtual channel). For another example, when the first antenna is a transmitting antenna, the channel statistical feature 5 may indicate a channel state of a virtual channel corresponding to a certain user transmitted through the first antenna. For another example, when the first antenna is a receiving antenna, the channel statistical feature 5 may indicate a channel state of a virtual channel received by a certain user through the first antenna. For another example, when the first antenna is a receiving antenna, the channel statistical feature 5 may indicate multiple times of the virtual channel received by the first antenna corresponding to multiple users of the virtual channel (for example, some or all users of the virtual channel). The mean value of the channel state over the path.

In addition, the above-mentioned channel statistical feature 5 can also represent the channel state of the virtual channel based on the combination of the first antenna pair and the delay path. For a specific example, refer to the example of the combination of the first antenna and the delay path above. Of course, the channel statistical feature 5 above can also represent the channel state of the virtual channel based on the combination of the first antenna pair and the user. For a specific example, refer to the above example of the combination of the first antenna and the user. For the sake of brevity, no further details are given below.

For the real channel, the channel statistical feature 5 may indicate the channel state of the real channel corresponding to the second antenna. Alternatively, the channel statistical feature 5 may indicate the channel state of the real channel corresponding to the second antenna pair. In this case, the channel statistical feature 5 may also be called "transmitting-receiving antenna-specific (Tx-Rx-specific) channel statistical feature".

When the above-mentioned second antenna is a receiving antenna, the real channel corresponding to the second antenna can be understood as the real channel received through the second antenna. At this time, the channel statistical feature 5 can also be called "receiving antenna-specific (Rx-specific) channel statistics feature". When the above-mentioned second antenna is a transmitting antenna, the real channel corresponding to the second antenna can be understood as the real channel transmitted through the second antenna. At this time, the channel statistical feature 5 can also be called "transmitting antenna-specific (Tx-specific) channel statistics feature".

In some implementation manners, the above-mentioned channel statistical feature 5 may represent the channel state of the real channel based on the combination of the second antenna and the delay path. For example, when the second antenna is a transmitting antenna, the channel statistical feature 5 may indicate a channel state of a certain delay path of a real channel transmitted through the second antenna. For another example, when the second antenna is a transmitting antenna, the channel statistical feature 5 may indicate the channel state of multiple delay paths (for example, all or part of the delay paths included in the real channel) transmitted through the second antenna mean value. For another example, when the second antenna is a receiving antenna, the channel statistical feature 5 may indicate a channel state of a certain delay path of a real channel received through the second antenna. For another example, when the second antenna is a receiving antenna, the channel statistical feature 5 may indicate an average value of channel states of multiple delay paths of a real channel received through the second antenna.

In some other implementation manners, the above-mentioned channel statistical feature 5 may represent the channel state of the real channel based on the combination of the second antenna and the user. For example, when the second antenna is a transmitting antenna, the channel statistical feature 5 may indicate the average value of channel states corresponding to multiple users (for example, some or all users of the real channel) transmitted through the second antenna. For another example, when the second antenna is a transmitting antenna, the channel statistical feature 5 may indicate a channel state corresponding to a certain user of a real channel transmitted through the second antenna. For another example, when the second antenna is a receiving antenna, the channel statistical feature 5 may indicate the channel state of a real channel received by a certain user through the second antenna. For another example, when the second antenna is a receiving antenna, the channel statistical feature 5 may indicate multiple delay paths of the real channel received by multiple users of the real channel (for example, may be some or all users of the real channel) through the second antenna. The mean value of the channel state of .

In addition, the above-mentioned channel statistical feature 5 can also represent the channel state of the real channel based on the combination of the second antenna pair and the delay path. For a specific example, refer to the example of combining the second antenna and the delay path above. Of course, the channel statistical feature 5 above can also represent the channel state of the real channel based on the combination of the second antenna pair and the user. For a specific example, refer to the above example of the combination of the second antenna and the user. For the sake of brevity, no further details are given below.

It should be noted that, the foregoing first antenna and the second antenna may be the same antenna or different antennas, which is not limited in this embodiment of the present application. Similarly, the first antenna pair and the second antenna pair may be the same or different antenna pairs.

In the following, the first antenna and the second antenna are used as receiving antennas, and the corresponding correlation matrix represents the channel statistical feature 5 as an example for introduction. Wherein, the correlation matrix representing the channel statistical feature 5 may be referred to as a "receiving antenna-specific channel correlation matrix". It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the channel correlation matrix R _X specific to the receiving antenna of the real channel indicates the mean value of the channel state of multiple delay paths of the real channel received by the Xth receiving antenna, namely

For a virtual channel, the channel correlation matrix specific to the receiving antenna of the virtual channel

Indicates the mean value of the channel state of multiple delay paths of the virtual channel received by the Yth receiving antenna, namely

It should be noted that when the first antenna and the second antenna are transmitting antennas or antenna pairs, the correlation matrix of the channel statistical feature 5 is similar to the channel correlation matrix dedicated to the receiving antenna above, and will not be described in detail below for brevity.

In the following, the first antenna and the second antenna are used as the transmitting antennas, and a correlation matrix is used to represent the channel statistical feature 5 as an example for introduction. Wherein, the correlation matrix representing the channel statistical feature 5 may be referred to as a "transmitting antenna-specific channel correlation matrix". It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the channel correlation matrix specific to the transmit antenna of the real channel

Indicates the channel state of the d-th delay path of the real channel received through the R-th transmit antenna, namely

Indicates the channel state of the d-th delay path of the virtual channel received by the R'th receiving antenna, that is

It should be noted that when the first antenna and the second antenna are receiving antennas or antenna pairs, the correlation matrix of the channel statistical feature 5 is similar to the above-mentioned channel correlation matrix dedicated to the transmitting antenna, and will not be described in detail below for brevity.

The channel statistical feature 6, considering that the channel feature is related to the frequency domain granularity of the channel, therefore, the channel statistical feature 6 may indicate the channel state of the channel divided according to the first frequency domain granularity. Wherein, the frequency domain granularity may be any frequency unit, for example, it may be a carrier, a subcarrier, an RB, a subband, and the like. In addition, the above-mentioned channel may be a frequency domain channel.

For the virtual channel, the channel statistical feature 6 may indicate the channel state of the virtual channel divided according to the first frequency domain granularity.

In some implementation manners, the above-mentioned channel statistical feature 6 may be combined with the user based on the first frequency domain granularity to represent the channel state of the virtual channel. For example, the channel statistical feature 6 may indicate the mean value of the channel state of the virtual channel corresponding to multiple users passing through the virtual channel (for example, some or all users of the virtual channel), where the virtual channel is divided according to the first frequency domain granularity . For another example, the channel statistical feature 6 may indicate a channel state of a virtual channel corresponding to a user of the virtual channel, where the virtual channel is divided according to the first frequency domain granularity.

For the real channel, the channel statistical feature 6 may indicate the channel state of the real channel divided according to the first frequency domain granularity.

In some implementation manners, the above-mentioned channel statistical feature 6 may be combined with the user based on the first frequency domain granularity to represent the channel state of the real channel. For example, the channel statistical feature 6 may indicate the mean value of the channel state of the real channel corresponding to multiple users passing through the real channel (for example, may be some or all users of the real channel), where the real channel is divided according to the first frequency domain granularity . For another example, the channel statistical feature 6 may indicate a channel state of a real channel corresponding to a certain user of the real channel, where the real channel is divided according to the first frequency domain granularity.

For ease of understanding, the following uses an example in which the first frequency domain granularity is RB and uses a correlation matrix to represent the channel statistical feature 6 as an example. It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the channel correlation matrix R _RB of the real channel indicates the mean value of the channel state of the real channel corresponding to the kth user, namely

For a virtual channel, the correlation matrix of the virtual channel

Indicates the mean value of the channel state of the virtual channel corresponding to the bth user of the virtual channel, namely

For ease of understanding, the following uses a subcarrier as the first frequency domain granularity and uses a correlation matrix to represent the channel statistical feature 6 as an example for introduction. It should be noted that the meanings of the letters in the formula have been introduced above, and will not be repeated here for brevity.

For a real channel, the correlation matrix R _F,d of the real channel indicates the statistical result of the channel state of the real channel corresponding to the kth user of the real channel, namely

For a virtual channel, the correlation matrix of the virtual channel

Indicates the statistical result of the channel state of the virtual channel corresponding to the bth user of the virtual channel,

It should be noted that there are many ways to express the correlation matrix of the above-mentioned channel statistical feature 6 based on different statistical contents of the channel statistical feature. Users are replaced by antennas. Correspondingly, the antennas in the above correlation matrix can be replaced by users. The general idea is similar. For the sake of brevity, details will not be described below.

The channel statistical features applicable to the embodiment of the present application are described above in combination with the channel statistical feature 1 to the channel statistical feature 6. The following describes the flow chart of the data processing method in the embodiment of the present application in conjunction with FIG. 6 . It should be understood that the method shown in FIG. 6 may be executed by a device having a data processing function, for example, a network device or a terminal device. The data processing method shown in FIG. 6 includes step S610 and step S640.

In step S610, the channel data of the virtual channel generated by the channel generator generating the adversarial network is acquired.

In step S620, the first channel statistical feature of the virtual channel is extracted based on the channel data of the virtual channel.

In other words, statistics are performed on the channel data of the virtual channel to obtain the first channel statistical feature. Wherein, the first channel statistical feature may be any one of the channel statistical features 1-6 introduced above. The first channel statistical feature may also be a combination of any of the above channel statistical features 1-6. Of course, this embodiment of the present application may also use other channel statistical features, for example, the angle of arrival of the channel, and the like.

In step S630, the second channel statistical feature of the real channel is extracted based on the channel data of the real channel.

In other words, statistics are performed on the channel data of the real channel to obtain the second channel statistical feature. Wherein, the second channel statistical feature may be any one of the channel statistical features 1-6 introduced above. The second channel statistical feature may also be a combination of any of the above channel statistical features 1-6. Of course, this embodiment of the present application may also use other channel statistical features, for example, the angle of arrival of the channel, and the like.

Usually, in order to improve the accuracy of describing the real channel by the second channel statistical feature, the second channel statistical feature may be calculated based on the channel data of all real channels in the channel data ensemble of the real channel. Of course, in order to reduce the amount of calculation for obtaining the second channel statistical features, some real channel data may be selected from the full set of real channel channel data to count the second channel statistical features, which is not limited in this embodiment of the present application.

In step S640, the difference between the first channel statistical feature and the second channel statistical feature is determined.

The difference between the first channel statistical feature and the second channel statistical feature may be used to indicate the difference between the channel data of the virtual channel and the channel data of the real channel. In other words, the above-mentioned difference between the first channel statistical feature and the second channel statistical feature may be replaced by the difference between the channel data of the virtual channel and the channel data of the real channel.

As mentioned above, the channel discriminator only helps to train the channel generator in the training phase. In the field of communication, the quality of the channel generator (or the quality of the channel data generated by the channel generator for virtual channels) is particularly important. However, at present, both the channel generator and the channel discriminator are evaluated as a whole by the loss function during the training process, that is, the result of the loss function can only indicate the performance of the channel generator and the channel discriminator as a whole, and cannot be evaluated separately The quality of the channel generator. In this way, when the loss function reaches the optimization goal, the quality of the channel generator may still be low, and then channel modeling is performed based on the channel data of the virtual channel output by the channel generator, which will cause the established channel model to be unable to describe or Characterize the true channel.

Therefore, in order to avoid the above problems, the embodiment of the present application also provides a data processing method, by comparing the difference between the channel data of the virtual channel and the channel data of the real channel, to evaluate the quality of the channel generator independently, that is, to obtain the countermeasure generating the channel data of the virtual channel generated by the channel generator of the network; and determining whether to save the channel generator based on the difference between the channel data of the virtual channel and the channel data of the real channel (also called "the first difference").

In the embodiment of the present application, the quality of the channel generator is evaluated by comparing the difference between the channel data of the real channel and the channel data of the virtual channel, so as to improve the accuracy of channel modeling. When evaluating the training results of GAN based on the loss function as a whole, the quality of the channel generator cannot be evaluated separately.

In order to further improve the accuracy of evaluating the quality of the channel generator, the channel statistical characteristics (also known as "first channel statistical characteristics") of the channel data of the virtual channel generated by the channel generator can be compared with the channel statistical characteristics of the channel data of the real channel (also known as "second channel statistical characteristics") to evaluate the quality of the channel generator alone. In some implementation manners, the above-mentioned scheme for evaluating the quality of a channel generator based on channel statistical characteristics can also be used in combination with the scheme in FIG. 7 , that is, the method described in FIG. 6 can also include: The difference between features that determine whether to save the channel generator. Therefore, the above-mentioned solution for evaluating the quality of the channel generator may be called a "channel evaluation solution", and correspondingly, the above-mentioned device for evaluating the channel generator may be called a "channel evaluator". Of course, the above-mentioned solution for evaluating the quality of the channel generator based on channel statistical characteristics can also be used alone, which is not limited in this embodiment of the present application.

In some implementation manners, the quality of the channel generator may be evaluated based on the difference _i after the i-th round of training process and the difference _i+1 after the i+1-th round of training process, where i is a positive integer. When the above difference _i+1 is higher than the difference _i , it can be considered that the quality of the channel generator _i+1 obtained through the training process of the i+1th round is relatively high. At this time, the model saving instruction can be output to instruct to save the channel generator _i+1 . Conversely, when the above difference _i is higher than the difference _i+1 , the quality of the channel generator _i+1 obtained through the training process of the i+1th round is poor. At this time, the channel generator can continue to be trained, that is Channel generator _i+1 is not saved. Certainly, in some other implementation manners, the quality of the channel generator may be evaluated by comparing the foregoing difference with a preset threshold. When the above difference is higher than the preset threshold, it can be considered that the quality of the channel generator is high, and at this time, the training process of the channel generator can be stopped. For example, a model save indication may be output to instruct the save channel generator. On the contrary, when the above-mentioned difference is lower than the preset threshold, it can be considered that the quality of the channel generator is low, and at this time, the training of the channel generator can be continued, that is, the channel generator is not saved. There are many ways to determine whether to include the channel generator based on the above difference, which is not limited in this embodiment of the present application.

For ease of understanding, the channel evaluator 700 in the embodiment of the present application is introduced below with reference to FIG. 7 by taking determining the quality of a channel generator based on the difference in each round of training process as an example. It should be noted that, for terms related to the above in FIG. 7 , reference may be made to the introduction above, and for the sake of brevity, details are not repeated here. The channel estimator 700 shown in FIG. 7 includes a discriminator 701 .

In step S710, the channel estimation discriminator 700 extracts a second channel statistical feature from the ensemble of channel data of the real channel.

In step S720, the channel evaluator 700 extracts a first channel statistical feature from the channel data of the virtual channel.

In step S730, the first channel statistical feature and the second channel statistical feature are input to the discriminator 701 respectively.

The above-mentioned discriminator is used to calculate the difference between the first channel statistical feature and the second channel statistical feature after each round of training process, and compare the difference corresponding to each round of training process, and determine whether to store the channel generator model based on the comparison result. Specifically, when the above-mentioned difference _i+1 is higher than or equal to the difference _i , step S740 is executed. On the contrary, when the above-mentioned difference _i+1 is lower than the difference _i , the current round of comparison process ends.

In step S740, a model saving instruction is output to instruct the channel generator to be saved.

For ease of understanding, the process of using the channel estimator 700 in the GAN training process is introduced below with reference to FIG. 8 . Fig. 8 is a schematic diagram of the GAN training process of the embodiment of the present application. It should be understood that units with the same function in FIG. 7 and FIG. 8 use the same number. The method shown in FIG. 8 includes steps S810 to S890.

In step S810, channel data of real channels are extracted from the ensemble of channel data of real channels.

In some implementation manners, if the channel data set of real channels has a large amount of data, batch sampling may be used to extract a batch of channel data of real channels from the full set of channel data of real channels.

In step S820, the input information is input into the channel generator to obtain the channel data of the virtual channel.

In step S830, input the channel data into the channel discriminator to obtain the discriminant result.

The above channel data may be channel data of a virtual channel or may also be channel data of a real channel.

In step S840, the channel data of the real channel, the channel data of the virtual channel and the identification result are input into the loss function to obtain the result of the statistical loss function.

In step S850, the second channel statistical feature is extracted from the channel data of the real channel.

In step S860, the first channel statistical feature is extracted from the channel data of the virtual channel.

In step S880, the first channel statistical feature and the second channel statistical feature are respectively input into the channel evaluator 700.

The functions of the channel evaluator 700 can be referred to in FIG. 7 . Specifically, when the above-mentioned difference _i+1 is higher than or equal to the difference _i , step S890 is performed. On the contrary, when the above-mentioned difference _i+1 is lower than the difference _i , the current round of comparison process ends.

In step S890, a model saving instruction is output to instruct the channel generator to be saved.

The channel statistical features applicable to this embodiment of the present application are introduced above in combination with channel statistical features 1-6. For ease of understanding, the discriminators applicable to the embodiments of the present application are respectively introduced below based on the above-mentioned channel statistical characteristics 1-6. It should be noted that there are many kinds of discriminators applicable to the embodiments of the present application, and are not limited to the discriminators described below.

The discriminator 1 designed based on the channel statistical feature 1 is used to calculate the statistical results of the channel state of the multiple delay paths contained in the virtual channel corresponding to all users of the virtual channel, and the multiple delay paths contained in the real channel corresponding to all users of the real channel The difference between the statistical results of the channel state of the delay paths.

In some implementations, if the above statistical results use the full channel correlation matrix R of the real channel and the full channel correlation matrix of the virtual channel respectively

Indicates that the discriminator 1 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

Represents the full channel correlation matrix of the i-th round of virtual channels.

when

When , update the model of the channel generator G to be the training model of the i+1 round. Conversely, when

When , the model of the channel generator G is still saved as the training model of the i-th round.

It should be noted that, the norm of the above matrix can usually be set as F-norm (Frobenius norm). Of course, the norm of the above matrix can also be set to 0, 1, 2, infinite norm, etc. according to different requirements.

The discriminator 2 designed based on the channel statistical characteristics 2 is used to calculate the statistical results of the transmission delay required by all users of the virtual channel to transmit data through the target delay path of the virtual channel, which is different from that of all users of the real channel through the real channel. The difference between the statistical results of the transmission delay required for data transmission along the target delay path.

In some implementations, if the above statistical results use the delay-specific channel correlation matrix R _d of the real channel and the delay-specific channel correlation matrix of the virtual channel respectively

Indicates that the discriminator 2 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

Denotes the delay-specific channel correlation matrix of the i-th round of virtual channels.

when

It should be noted that the norm of the above matrix can usually be set as the F-norm. Of course, the norm of the above matrix can also be set to 0, 1, 2, infinite norm, etc. according to different requirements.

The discriminator 3 designed based on the channel statistical feature 3 is used to calculate the statistical results of the channel state of the multiple delay paths of the virtual channel corresponding to the first user of the virtual channel, and the multiple of the real channel corresponding to the first user of the real channel The difference between the statistical results of the channel state of the delay paths.

In some implementations, if the above statistical results use the user-specific channel correlation matrix R _k of the real channel and the user-specific channel correlation matrix of the virtual channel respectively

Indicates that the discriminator 3 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

Indicates the user-specific channel correlation matrix of the i-th round of virtual channels.

when

It should be noted,

Indicates the degree of similarity of the selected real channel state (that is, the channel state of the real channel corresponding to the kth user) that is closest to the channel state of the virtual channel corresponding to the bth user in the full set of channel data of the real channel. In some implementation manners, the channel state of the real channel corresponding to the kth user may be selected after traversing the full set of channel data of the real channel based on the channel state of the virtual channel corresponding to the bth user. Therefore, the kth user and the bth user may be the same user or different users.

In addition, the norm of the above matrix can usually be set as the F-norm. Of course, the norm of the above matrix can also be set to 0, 1, 2, infinite norm, etc. according to different requirements.

The discriminator 4 designed based on the channel statistical feature 4 is used to calculate the statistical results of the channel state of the target delay path of the virtual channel corresponding to the first user of the virtual channel, and the target time of the real channel corresponding to the first user of the real channel The difference between the statistical results of the extended path and the channel state.

In some implementations, if the above statistical results use the channel correlation matrix R k,d specific to the user delay of the real channel and the channel correlation matrix R _k,d specific to the user delay of the virtual channel respectively

Indicates that the discriminator 4 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

Indicates the user delay-specific channel correlation matrix of the i-th round of virtual channels.

when

It should be noted,

Indicates the real channel state closest to the channel state of the d-th delay path of the selected virtual channel corresponding to the b-th user in the full set of channel data of the real channel (that is, the real channel state of the real channel corresponding to the k-th user The degree of similarity between channel states of d delay paths). In some implementations, the channel state of the d-th delay path of the real channel corresponding to the k-th user may be based on the channel state of the d-th delay path of the virtual channel corresponding to the b-th user traversing the real channel Selected after the full set of channel data. Therefore, the kth user and the bth user may be the same user or different users. Similarly, the dth delay path of the real channel and the dth delay path of the virtual channel may be channel The delay paths with the same arrival order in the channel can also be the delay paths with different arrival orders in the channel.

The discriminator 5 designed based on the channel statistical feature 5 is used to calculate the difference between the channel state of the virtual channel corresponding to the first antenna and the channel state of the real channel corresponding to the second antenna. Alternatively, the discriminator 5 may also be used to calculate the difference between the channel state of the virtual channel corresponding to the first antenna pair and the channel state of the real channel corresponding to the second antenna pair.

It should be noted that the corresponding discriminators can be set based on the different meanings of the channel statistical features 5. In general, it can be understood as the difference between the channel statistical features 5 of the hypothetical round i virtual channel and the channel statistical features 5 of the real channel is the difference _i , the discriminator can be used to calculate the above difference _i , and, when the difference _i ≤ difference _i+1 , update the model of the channel generator G to be the training model of the i+1 round. Conversely, when difference _i > difference _i+1 , the model of channel generator G is still saved as the training model for the i-th round.

The discriminator 5 is introduced below by taking the channel statistical feature 5 representing the transmission delay of a channel received by a certain receiving antenna as an example. The principle of the discriminator 5 designed based on other forms of channel statistical features 5 is similar to that of the discriminator 5 described below, and for the sake of brevity, details will not be described below.

In some implementations, if the above channel statistical feature 5 respectively uses the channel correlation matrix R X specific to the receiving antenna of the real channel and the channel correlation matrix R _X specific to the receiving antenna of the virtual channel

Indicates that the discriminator 5 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

when

The discriminator 6 designed based on the channel statistical feature 6 is used to calculate the difference between the channel state of the virtual channel divided according to the first frequency domain granularity and the channel state of the real channel divided according to the first frequency domain granularity.

It should be noted that the corresponding discriminator 6 can be set based on the different meanings of the channel statistical features 6. In general, it can be understood as the difference between the channel statistical features 6 of the i-th round virtual channel and the channel statistical features 6 of the real channel. If the difference is difference _i , the discriminator can be used to calculate the above difference _i , and, when the difference _i ≤ difference _i+1 , update the model of the channel generator G to be the training model of the i+1 round. Conversely, when difference _i > difference _i+1 , the model of channel generator G is still saved as the training model for the i-th round.

In the following, the discriminator 6 will be introduced by taking the channel statistical feature 6 as an example representing the transmission delay of a channel divided by the RB as the frequency domain granularity. The discriminator 6 designed based on other forms of channel statistical features 6 is similar in principle to the discriminator 6 described below, and for the sake of brevity, details are not described below.

In some implementations, if the above channel statistical feature 6 respectively uses the channel correlation matrix R _RB of the real channel and the correlation matrix of the virtual channel

Indicates that the discriminator 6 can be expressed as

Among them, ||·|| _l represents the k-norm of the matrix,

Indicates the channel correlation matrix of the i-th round of virtual channels.

when

As introduced above, the essence of the training process of GAN is that the channel generator 310 and the channel discriminator 320 are in the process of confrontation game. In the alternate training process of the channel generator 310 and the channel discriminator 320, the training objective of the loss function is just opposite, which makes the process of training GAN very unstable, and even GAN is difficult to converge. In addition, the loss functions used in the current GAN training process are all general loss functions, which further reduces the convergence speed of GAN.

Therefore, in order to improve the convergence speed of the GAN, the embodiment of the present application also provides a data processing method, that is, to train the generative adversarial network according to the difference between the statistical characteristics of the first channel and the statistical characteristics of the second channel.

In some implementations, the above-mentioned difference between the first channel statistics and the second channel statistics can be combined with a traditional loss function (for example, a cross-entropy loss function or a loss function based on bulldozer distance) to form a joint loss function to train the GAN. Of course, in some other manners, the difference between the first channel statistical feature and the second channel statistical feature may also be directly used as a loss function to train the GAN, which is not specifically limited in this embodiment of the present application.

In the embodiment of the present application, GAN is trained based on the difference between the first channel statistical characteristics and the second channel statistical characteristics, so as to focus the output of the loss function on comparing the channel statistical characteristics of the virtual channel and the channel statistical characteristics of the real channel In order to better guide the training process of GAN, it is beneficial to improve the convergence speed of GAN. It avoids the traditional process of training GAN based on the ordinary loss function. The output of the loss function is the difference between the channel data of the virtual channel and the channel data of the real channel, resulting in a slow convergence speed of GAN.

In some implementations, the above joint loss function includes two parts of the first loss function L ₁ (H, z) and the second loss function L ₂ (H, z), namely

Wherein, the first loss function L ₁ (H, z) is a traditional loss function adopted by the channel discriminator (such as a cross-entropy loss function or a loss function based on bulldozer distance); the second loss function L ₂ (H, z) Indicates the loss function designed based on the channel statistical characteristics provided by the embodiment of the present application, therefore, it is also called the statistical loss function.

It should be noted that the weight coefficient λ in the joint loss function may be a fixed value, such as λ=0.3. Of course, the weight coefficient λ can also be a value that changes with the training process. For example, in the initial training result, set λ=0.7 so that the channel generator can quickly obtain the initial model weight with the assistance of channel statistics, and then reduce it by The weight coefficient (for example, λ=0.2) is used to reduce the proportion of channel statistical information assisting GAN training, so that the training of the channel discriminator and channel generator is more stable.

In addition, the foregoing second loss function may be designed based on different channel statistical characteristics, which is not limited in this embodiment of the present application. In some implementation manners, the expression manner of the above-mentioned second loss function may be the same as that of the discriminator. For example, when the channel statistical feature 1 is used as the channel statistical feature of the virtual channel and the real channel, the expression of the second loss function may be the same as that of the discriminator 1 . For another example, when the channel statistical feature 2 is used as the channel statistical feature of the virtual channel and the real channel, the expression of the second loss function may be the same as that of the discriminator 2 . For another example, when the channel statistical feature 3 is used as the channel statistical feature of the virtual channel and the real channel, the expression of the second loss function may be the same as that of the discriminator 3 . For another example, when the channel statistical feature 4 is used as the channel statistical feature of the virtual channel and the real channel, the expression of the second loss function may be the same as that of the discriminator 4 . For another example, when the channel statistical feature 5 is used as the channel statistical feature of the virtual channel and the real channel, the representation of the second loss function may be the same as that of the discriminator 5 . For another example, when the channel statistical feature 6 is used as the channel statistical feature of the virtual channel and the real channel, the expression of the second loss function may be the same as that of the discriminator 6 .

For ease of understanding, the following uses channel statistical feature 1 as an example to introduce the second loss function in the embodiment of the present application. Suppose the full-channel correlation matrix of the real channel is R, and the full-channel correlation matrix of the virtual channel is

And the second loss function constructed by choosing the F-norm can be expressed as:

The following describes the process of using the joint loss function in the GAN training of the embodiment of the present application with reference to FIG. 9 . The method shown in FIG. 9 includes steps S910 to S970.

In step S910, the channel data of the real channel is extracted from the full set of channel data of the real channel.

In step S920, the input information is input into the channel generator to obtain the channel data of the virtual channel.

In step S930, input the channel data into the channel discriminator to obtain the discriminant result.

In step S940, the second channel statistical feature is extracted from the channel data of the real channel.

In step S950, a first channel statistical feature is extracted from the channel data of the virtual channel.

In step S960, the first channel statistical feature and the second channel statistical feature are input into the statistical loss function to obtain a result of the statistical loss function.

In step S970, the result of the statistical loss function and the identification result are input into the first loss function to obtain the result of the first loss function.

The result of the first loss function above is used to guide the training process of the GAN network.

In order to verify the training results of the channel generator by the method of the embodiment of the present application, the following describes the channel data quality of the virtual channel of the channel generator trained by the training method provided by the embodiment of the present application with reference to FIG. 10 to FIG. 15 . Wherein, FIG. 10 shows the statistical average result of the power delay spectrum of the virtual channel after the first round of training. Figure 11 shows a single sample result of the power antenna spectrum of the virtual channel after the first round of training. Fig. 12 shows the statistical average result of the power delay spectrum of the virtual channel after the 1200th round of training. Figure 13 shows a single sample result of the power antenna spectrum of the virtual channel after the 1200th round of training. Fig. 14 shows the statistical averaging result of the power delay profile of a real channel. Figure 15 shows a single sample result of the power antenna spectrum for a real channel.

Referring to Figures 10 to 15, it can be clearly seen that the channel data of the virtual channel cannot fit the distribution of the real channel in the first round of training, but after the 1200th round of training, using the embodiment of the present application to provide The channel data of the virtual channel output by the channel generator selected by the model evaluator can better fit the distribution of the real channel.

The method embodiment of the present application is described in detail above with reference to FIG. 1 to FIG. 15 , and the device embodiment of the present application is described in detail below in conjunction with FIG. 16 to FIG. 18 . It should be understood that the descriptions of the method embodiments correspond to the descriptions of the device embodiments, therefore, for parts not described in detail, reference may be made to the foregoing method embodiments.

FIG. 16 is a schematic diagram of a data processing device according to an embodiment of the present application. The data processing device 1600 shown in FIG. 16 includes an acquisition unit 1610 and a processing unit 1620 .

The obtaining unit 1610 is used to obtain the channel data of the virtual channel generated by the channel generator generating the confrontation network;

A processing unit 1620, configured to extract a first channel statistical feature of the virtual channel based on the channel data of the virtual channel;

The processing unit 1620 is further configured to extract second channel statistical features of the real channel based on the channel data of the real channel;

The processing unit 1620 is further configured to determine a difference between the first channel statistical feature and the second channel statistical feature.

In some optional implementation manners, the processing unit is further configured to train the generative adversarial network according to the difference between the first channel statistical feature and the second channel statistical feature, and the generating Adversarial networks contain channel discriminators.

In some optional implementation manners, the processing unit is further configured to: determine whether to save the channel generator according to a difference between the first channel statistical feature and the second channel statistical feature.

In some optional implementation manners, the first channel statistical feature is used to indicate statistical results of channel states of multiple delay paths contained in the virtual channel corresponding to all users of the virtual channel; and/or, The second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths contained in the real channel corresponding to all users of the real channel.

In some optional implementation manners, the first channel statistical feature is used to indicate the statistical result of the channel state of the target delay path of the virtual channel corresponding to all users of the virtual channel; and/or, the The second channel statistical feature is used to indicate a statistical result of a channel state of a target delay path of the real channel corresponding to all users of the real channel.

In some optional implementation manners, the first channel statistical feature is used to indicate statistical results of channel states of multiple delay paths in the virtual channel corresponding to the first user of the virtual channel; and/or , the second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths in the real channel corresponding to the first user of the real channel.

In some optional implementation manners, the first channel statistical feature is used to indicate the statistical result of the channel state of the target delay path of the virtual channel corresponding to the second user of the virtual channel; and/or, the The second channel statistical feature is used to indicate the statistical result of the channel state of the target delay path of the real channel corresponding to the second user of the real channel.

In some optional implementation manners, the first channel statistical feature is used to indicate the channel state of the virtual channel corresponding to the first antenna or the first antenna pair; and/or, the second channel statistical feature is used to Indicating the channel state of the real channel corresponding to the second antenna or the second antenna pair.

In some optional implementation manners, the first antenna is a receiving antenna or a transmitting antenna, and/or, the second antenna is a receiving antenna or a transmitting antenna.

In some optional implementation manners, the first channel statistical feature is used to indicate the channel state of the virtual channel divided according to the first frequency domain granularity; and/or, the second channel statistical feature is used to indicate the channel state according to The channel state of the real channel divided by the first frequency domain granularity.

Fig. 17 is a schematic diagram of a data processing device according to another embodiment of the present application. The apparatus 1700 shown in FIG. 17 includes an acquisition unit 1710 and a processing unit 1720 .

An acquisition unit 1710, configured to acquire channel data of the virtual channel generated by the channel generator of the confrontation generation network;

The processing unit 1720 is configured to determine whether to save the channel generator based on the first difference between the channel data of the virtual channel and the channel data of the real channel.

In some optional implementation manners, the processing unit is further configured to: extract the first channel statistical feature of the virtual channel based on the channel data of the virtual channel; extract the first channel statistical feature of the real channel based on the channel data of the real channel second channel statistics; and determining a second difference between the first channel statistics and the second channel statistics, wherein the second difference is indicative of the first difference.

Fig. 18 is a schematic structural diagram of a data device according to an embodiment of the present application. The dashed line in Figure 18 indicates that the unit or module is optional. The apparatus 1800 may be used to implement the methods described in the foregoing method embodiments. Apparatus 1800 may be a chip, a terminal device or a network device.

Apparatus 1800 may include one or more processors 1810 . The processor 1810 may support the device 1800 to implement the methods described in the foregoing method embodiments. The processor 1810 may be a general purpose processor or a special purpose processor. For example, the processor may be a central processing unit (central processing unit, CPU). Alternatively, the processor can also be other general-purpose processors, digital signal processors (digital signal processors, DSPs), application specific integrated circuits (application specific integrated circuits, ASICs), off-the-shelf programmable gate arrays (field programmable gate arrays, FPGAs) Or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.

Apparatus 1800 may also include one or more memories 1820 . A program is stored in the memory 1820, and the program can be executed by the processor 1810, so that the processor 1810 executes the methods described in the foregoing method embodiments. The memory 1820 may be independent from the processor 1810 or may be integrated in the processor 1810 .

The apparatus 1800 may also include a transceiver 1830 . The processor 1810 can communicate with other devices or chips through the transceiver 1830 . For example, the processor 1810 may send and receive data with other devices or chips through the transceiver 1830 .

The embodiment of the present application also provides a computer-readable storage medium for storing programs. The computer-readable storage medium can be applied to the terminal or the network device provided in the embodiments of the present application, and the program enables the computer to execute the methods performed by the terminal or the network device in the various embodiments of the present application.

The embodiment of the present application also provides a computer program product. The computer program product includes programs. The computer program product can be applied to the terminal or the network device provided in the embodiments of the present application, and the program enables the computer to execute the methods performed by the terminal or the network device in the various embodiments of the present application.

The embodiment of the present application also provides a computer program. The computer program can be applied to the terminal or the network device provided in the embodiments of the present application, and the computer program enables the computer to execute the methods performed by the terminal or the network device in the various embodiments of the present application.

It should be understood that the terms "system" and "network" may be used interchangeably in this application. In addition, the terms used in the application are only used to explain the specific embodiments of the application, and are not intended to limit the application. The terms "first", "second", "third" and "fourth" in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order . Furthermore, the terms "include" and "have", as well as any variations thereof, are intended to cover a non-exclusive inclusion.

In the embodiments of the present application, the "indication" mentioned may be a direct indication, an indirect indication, or an association relationship. For example, A indicates B, which can mean that A directly indicates B, for example, B can be obtained through A; it can also indicate that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also indicate that there is an association between A and B relation.

In this embodiment of the application, "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B may also be determined according to A and/or other information.

In this embodiment of the application, the term "corresponding" may indicate that there is a direct or indirect correspondence between the two, or that there is an association between the two, or that it indicates and is instructed, configures and is configured, etc. relation.

In this embodiment of the application, "predefined" or "preconfigured" can be realized by pre-saving corresponding codes, tables or other methods that can be used to indicate relevant information in devices (for example, including terminal devices and network devices). The application does not limit its specific implementation. For example, pre-defined may refer to defined in the protocol.

In the embodiment of the present application, the "protocol" may refer to a standard protocol in the communication field, for example, may include the LTE protocol, the NR protocol, and related protocols applied to future communication systems, which is not limited in the present application.

The term "and/or" in the embodiment of the present application is only an association relationship describing associated objects, which means that there may be three relationships, for example, A and/or B, which can mean: A exists alone, and A and B exist at the same time , there are three cases of B alone. In addition, the character "/" in this article generally indicates that the contextual objects are an "or" relationship.

In various embodiments of the present application, the sequence numbers of the above-mentioned processes do not mean the order of execution, and the execution order of each process should be determined by its functions and internal logic, rather than the implementation process of the embodiments of the present application. constitute any limitation.

In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.

In the above embodiments, all or part of them may be implemented by software, hardware, firmware or any combination thereof. When implemented using software, it may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on the computer, the processes or functions according to the embodiments of the present application will be generated in whole or in part. The computer can be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website, computer, server or data center Transmission to another website site, computer, server or data center by wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be read by a computer, or a data storage device such as a server or a data center integrated with one or more available media. The available medium may be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital versatile disc (digital video disc, DVD)) or a semiconductor medium (for example, a solid state disk (solid state disk, SSD) )wait.

The above is only a specific implementation of the application, but the scope of protection of the application is not limited thereto. Anyone familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the application. Should be covered within the protection scope of this application. Therefore, the protection scope of the present application should be determined by the protection scope of the claims.

Claims

A data processing method, characterized in that, comprising:

Obtain channel data of a virtual channel generated by a channel generator generating an adversarial network;

extracting a first channel statistical feature of the virtual channel based on the channel data of the virtual channel;

extracting second channel statistical features of the real channel based on channel data of the real channel;

Differences between the first channel statistics and the second channel statistics are determined.
The method of claim 1, further comprising:

The generative adversarial network is trained according to the difference between the first channel statistical feature and the second channel statistical feature, and the generative adversarial network includes a channel discriminator.
The method of claim 1, further comprising:

Determine whether to save the channel generator according to the difference between the first channel statistical feature and the second channel statistical feature.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate channels of multiple delay paths of the virtual channel corresponding to all users of the virtual channel statistical results of the status; and/or,

The second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths of the real channel corresponding to all users of the real channel.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate the channel state of the target delay path of the virtual channel corresponding to all users of the virtual channel statistical results of ; and/or,

The second channel statistical feature is used to indicate the statistical result of the channel state of all users of the real channel corresponding to the target delay path of the real channel.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate the number of delay paths of the virtual channel corresponding to the first user of the virtual channel statistics on channel status; and/or,

The second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths of the real channel corresponding to the first user of the real channel.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate the channel of the target delay path of the virtual channel corresponding to the second user of the virtual channel statistical results of the status; and/or,

The second channel statistical feature is used to indicate the statistical result of the channel state of the target delay path of the real channel corresponding to the second user of the real channel.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate the channel state of the virtual channel corresponding to the first antenna or the first antenna pair; and/or ,

The second channel statistical feature is used to indicate the channel state of the real channel corresponding to the second antenna or the second antenna pair.
The method according to claim 8, wherein the first antenna is a receiving antenna or a transmitting antenna, and/or the second antenna is a receiving antenna or a transmitting antenna.
The method according to any one of claims 1-3, wherein the first channel statistical feature is used to indicate the channel state of the virtual channel divided according to the first frequency domain granularity; and/or

The second channel statistical feature is used to indicate the channel state of the real channel divided according to the first frequency domain granularity.
A data processing method, characterized in that, comprising:

Obtain channel data for the virtual channel generated by the channel generator of the adversarial generative network;

Based on a first difference between channel data of the virtual channel and channel data of a real channel, it is determined whether to save the channel generator.
The method according to claim 11, wherein, before determining whether to save the channel generator based on the difference between the channel data of the virtual channel and the channel data of the real channel,

The method also includes:

extracting a first channel statistical feature of the virtual channel based on the channel data of the virtual channel;

extracting second channel statistical features of the real channel based on channel data of the real channel;

A second difference between the first channel statistic and the second channel statistic is determined, wherein the second difference is indicative of the first difference.
A data processing device, characterized in that it comprises:

The obtaining unit is used to obtain the channel data of the virtual channel generated by the channel generator generating the confrontation network;

a processing unit, configured to extract a first channel statistical feature of the virtual channel based on the channel data of the virtual channel;

The processing unit is further configured to extract second channel statistical features of the real channel based on channel data of the real channel;

The processing unit is further configured to determine a difference between the first channel statistical feature and the second channel statistical feature.
The device according to claim 13, wherein the processing unit is further configured to:

The generative adversarial network is trained according to the difference between the first channel statistical feature and the second channel statistical feature, and the generative adversarial network includes a channel discriminator.
The device according to claim 13, wherein the processing unit is further configured to:

Determine whether to save the channel generator according to the difference between the first channel statistical feature and the second channel statistical feature.
The device according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate channels of multiple delay paths of the virtual channel corresponding to all users of the virtual channel statistical results of the status; and/or,

The second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths of the real channel corresponding to all users of the real channel.
The device according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate the channel state of the target delay path of the virtual channel corresponding to all users of the virtual channel statistical results of ; and/or,

The second channel statistical feature is used to indicate a statistical result of a channel state of a target delay path of all users of the real channel passing through the real channel.
The device according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate the number of delay paths of the virtual channel corresponding to the first user of the virtual channel statistics on channel status; and/or,

The second channel statistical feature is used to indicate statistical results of channel states of multiple delay paths of the real channel corresponding to the first user of the real channel.
The device according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate the channel of the target delay path of the virtual channel corresponding to the second user of the virtual channel statistical results of the status; and/or,

The second channel statistical feature is used to indicate the statistical result of the channel state of the target delay path of the real channel corresponding to the second user of the real channel.
The apparatus according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate the channel state of the virtual channel corresponding to the first antenna or the first antenna pair; and/or ,

The second channel statistical feature is used to indicate the channel state of the real channel corresponding to the second antenna or the second antenna pair.
The device according to claim 20, wherein the first antenna is a receiving antenna or a transmitting antenna, and/or the second antenna is a receiving antenna or a transmitting antenna.
The device according to any one of claims 13-15, wherein the first channel statistical feature is used to indicate the statistical result of the channel state of the virtual channel divided according to the first frequency domain granularity; and/ or

The second channel statistical feature is used to indicate the statistical result of the channel state of the real channel divided according to the first frequency domain granularity.
A data processing device, characterized in that it comprises:

The acquisition unit is used to acquire the channel data of the virtual channel generated by the channel generator of the confrontation generation network;

A processing unit, configured to determine whether to save the channel generator based on a first difference between the channel data of the virtual channel and the channel data of the real channel.
The device according to claim 23, wherein the processing unit is further configured to:

extracting a first channel statistical feature of the virtual channel based on the channel data of the virtual channel;

extracting second channel statistical features of a real channel based on channel data of the real channel; and

A second difference between the first channel statistic and the second channel statistic is determined, wherein the second difference is indicative of the first difference.
A data processing device, characterized by comprising a memory and a processor, the memory is used to store a program, and the processor is used to call the program in the memory to execute the program described in any one of claims 1-12. described method.
An apparatus, characterized by comprising a processor, configured to call a program from a memory to execute the method according to any one of claims 1-12.
A chip, characterized by comprising a processor, configured to call a program from a memory, so that a device installed with the chip executes the method according to any one of claims 1-12.
A computer-readable storage medium, characterized in that a program is stored thereon, and the program causes a computer to execute the method according to any one of claims 1-12.
A computer program product, characterized by comprising a program, the program causes a computer to execute the method according to any one of claims 1-12.
A computer program, characterized in that the computer program causes a computer to execute the method according to any one of claims 1-12.