Millimeter wave channel prediction preprocessing method and preprocessing system for multipath and doppler scenarios

Abstract

The application relates to a wireless communication technology, and relates to a millimeter wave channel prediction preprocessing method and system for a multipath and Doppler scene, which comprises the following steps: establishing a single-cell millimeter wave MIMO channel model comprising a base station and a user; establishing a channel prediction model, which comprises a connected channel estimator, a channel data transformer, a millimeter wave channel prediction network, and a channel data inverse transformer; the channel estimator is used for acquiring known channel data according to a pilot signal transmitted by the user; the channel data transformer is used for preprocessing the channel data; the millimeter wave channel prediction network is used for predicting future channel data according to the known channel data; and the channel data inverse transformer is used for transforming channel data in a time-frequency angle domain into channel data in a time-space domain. The application can improve the channel prediction accuracy in the multipath and Doppler scene by preprocessing the channel data, and can effectively improve the communication quality.

Description

Technical Field

[0001] This invention relates to wireless communication technology, and more particularly to a millimeter-wave channel prediction preprocessing method and preprocessing system for multipath and Doppler scenarios. Background Technology

[0002] Channel prediction technology is one of the key technologies in modern wireless communication. In millimeter-wave MIMO systems, the channel state changes rapidly, requiring more frequent acquisition of channel state information. However, acquiring channel state information by transmitting pilot signals consumes a lot of communication system resources. Using channel prediction technology to directly predict and obtain future channel state information is one way to solve this problem.

[0003] However, when using channel prediction techniques to predict future channel conditions, predictions are generally based on channel data represented by a spatiotemporal channel matrix. In this spatiotemporal channel matrix representation, different multipath paths interfere with each other, and the Doppler effect introduces frequency offsets. This makes the channel data represented by the spatiotemporal channel matrix susceptible to the influence of multipath and Doppler effects, making accurate channel prediction difficult.

[0004] Most existing channel prediction methods are based on channel data represented by spatiotemporal channel matrices, which are not suitable for channel prediction in multipath and Doppler scenarios, especially millimeter-wave channel prediction in multipath and Doppler scenarios. Therefore, how to improve the prediction accuracy of existing channel prediction methods in multipath and Doppler scenarios through appropriate preprocessing techniques is an urgent problem to be solved. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides a millimeter-wave channel prediction preprocessing method and preprocessing system for multipath and Doppler scenarios. Based on existing channel prediction methods, a preprocessing method is designed to preprocess channel data to adapt it to multipath and Doppler scenarios. This can effectively improve the prediction accuracy of millimeter-wave MIMO systems, enhance precoding quality, achieve higher system capacity, and improve communication quality.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide a millimeter-wave channel prediction preprocessing method for multipath and Doppler scenarios, comprising the following steps:

[0008] S1. Establish a single-cell millimeter-wave MIMO channel model, including one base station and one user;

[0009] S2. Establish a channel prediction model, which includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected in series. The channel estimator is used to obtain known channel data based on the pilot signal sent by the user. The channel data transformer preprocesses the channel data. The millimeter-wave channel prediction network predicts future channel data based on the known channel data. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back into the channel data represented by the time-space domain channel matrix.

[0010] Secondly, this invention provides a millimeter-wave channel prediction preprocessing system for multipath and Doppler scenarios, implemented based on the aforementioned millimeter-wave channel prediction preprocessing method; it includes the following modules:

[0011] The millimeter-wave MIMO channel model construction module includes a base station and a user;

[0012] The channel prediction model building module is used to build a channel prediction model. It includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected to each other. The channel estimator is used to obtain known channel data based on the pilot signal. The channel data transformer preprocesses the channel data, transforming the channel data represented by the spatiotemporal domain channel matrix into the channel data represented by the time-frequency angular domain channel matrix. The millimeter-wave channel prediction network predicts future channel data based on the known channel data. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back into the channel data represented by the spatiotemporal domain channel matrix.

[0013] Compared with the prior art, the beneficial effects of the present invention include at least the following:

[0014] The millimeter-wave channel prediction preprocessing method and preprocessing system provided by this invention can reduce the impact of multipath and Doppler on channel prediction under complex and variable millimeter-wave channels, improve channel prediction accuracy, and enhance precoding quality. It can effectively alleviate the problem of millimeter-wave channel prediction being susceptible to interference under multipath and Doppler scenarios, achieve higher system capacity, and improve wireless communication quality. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the millimeter-wave channel prediction preprocessing method for multipath and Doppler scenarios provided in an embodiment of the present invention;

[0016] Figure 2 This is a schematic diagram of the channel prediction model in an embodiment of the present invention;

[0017] Figure 3 This is a schematic diagram of the channel data converter in an embodiment of the present invention;

[0018] Figure 4 This is a schematic diagram of the channel data inverse transformer in an embodiment of the present invention;

[0019] Figure 5 This is a schematic diagram comparing the mean square error of predicted future channel data between millimeter-wave channel prediction networks based on RNN, LSTM, and Transformer models using channel prediction preprocessing methods and without channel prediction preprocessing methods, as shown in this embodiment of the invention.

[0020] Figure 6 This is a schematic diagram comparing the mean square error of predicted future channel data using a channel prediction preprocessing method and without using a channel prediction preprocessing method under different antenna configurations in a millimeter-wave channel prediction network based on the Transformer model according to an embodiment of the present invention.

[0021] Figure 7 This is a schematic diagram comparing the mean square error of future channel data when using channel data represented by the Transformer model in the millimeter-wave channel prediction network in this embodiment of the invention, using channel data represented by the spatiotemporal domain channel matrix, the time-angle domain channel matrix, the time-frequency spatial domain channel matrix, and the time-frequency angle domain channel matrix respectively to predict the channel data.

[0022] Figure 8 This is a block diagram of a millimeter-wave channel prediction and preprocessing system for multipath and Doppler scenarios provided in an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Example 1

[0025] Please see Figure 1 This embodiment provides a millimeter-wave channel prediction preprocessing method for multipath and Doppler scenarios, including the following steps:

[0026] S1. Establish a single-cell millimeter-wave MIMO channel model, which includes a base station and a user.

[0027] In the single-cell millimeter-wave MIMO channel model, both the base station and the user's antennas are configured as uniform linear arrays. In this embodiment, the base station's uniform linear array is equipped with... The user's uniform linear array is equipped with a single transmitting antenna. The base station provides services to users through its transmitting antenna. Assume there are... There are multiple path propagation paths, and the base station and the user are on the first path. The channel state information of a frame can be represented as:

[0028] ;

[0029] in, Indicates the first Complex decay factor of each propagation path Indicates the first Doppler frequency shift of the propagation path Indicates frame duration, superscript This indicates the conjugate transpose. , Indicates the array steering vector. Indicates the first The angle of arrival of the path, Indicates the first The departure angle of the path.

[0030] The channel state information is considered to be constant during the coherence time of the channel, but changes between two different coherence times.

[0031] The array guide vector and They are represented as follows:

[0032] ;

[0033] ;

[0034] in, The wavelength of the carrier wave. Indicates antenna spacing, superscript This indicates transpose.

[0035] S2. Establish a channel prediction model, which includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected together. The channel estimator is used to obtain known channel data based on the pilot signal sent by the user. The channel data transformer preprocesses the channel data. The millimeter-wave channel prediction network predicts future channel data based on the known channel data, that is, it predicts the future channel conditions. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back to the channel data represented by the time-space domain channel matrix.

[0036] like Figure 2As shown, in the channel prediction model of this embodiment, known channel data is first obtained based on the pilot signal. Then, a channel data transformer is used to preprocess the channel data, transforming the channel data represented by the spatiotemporal domain channel matrix into channel data represented by the time-frequency angular domain channel matrix. The transformed channel data is input into the millimeter-wave channel prediction network; the millimeter-wave channel prediction network outputs... The system generates future channel data (i.e., channel state information to be predicted). Finally, a channel data inverse transformer transforms the future channel data, represented by the time-frequency angular domain channel matrix, back to the future channel data represented by the time-space domain channel matrix. This helps the wireless communication system intelligently select communication strategies and improve communication quality. Specifically, the process includes the following steps:

[0037] S21. Establish a channel estimator to obtain known channel data based on the pilot signals sent by the user.

[0038] Specifically, the base station and the user communicate using time-division duplex (TDM). The user sends a pilot signal to the base station at regular intervals to obtain channel state information, i.e., to acquire known channel data. The process by which the channel estimator acquires known channel data can be represented as follows:

[0039] ;

[0040] in, The length of the pilot channel. For the base station in the The signal received by the frame; For the first Channel state information of the frame, i.e., the first frame The channel data, represented by the spatiotemporal channel matrix of the frame, is estimated by the channel estimator; For users in the first The pilot signal transmitted in the frame, It is additive white Gaussian noise.

[0041] S22. Establish a channel data transformer to preprocess the known channel data obtained by the channel estimator, transforming the channel data represented by the spatiotemporal domain channel matrix into the channel data represented by the time-frequency angular domain channel matrix.

[0042] like Figure 3 As shown, the transformation of channel data from the spatiotemporal domain to the time-frequency angular domain is achieved in two steps:

[0043] S221. Transform the channel data represented by the spatiotemporal channel matrix into the channel data represented by the time-angle channel matrix using the spatial discrete Fourier transform:

[0044] ;

[0045] in, The channel data represented by the spatiotemporal domain channel matrix reflects the channel conditions between each base station antenna and each user antenna; This represents the channel data represented by the time-angle domain channel matrix. Representing the discrete Fourier transform matrix:

[0046] ;

[0047] in, The number of points in the spatial domain discrete Fourier transform; matrix In this context, the superscript of each element represents the product of the row and column numbers of the corresponding element, indicating exponentiation, such as... express power of 0, It also means power of 0, express of Power of 1.

[0048] S222. Transform the channel data represented by the time-angle domain channel matrix to the channel data represented by the time-frequency-angle domain channel matrix through the discrete short-time Fourier transform.

[0049] Preferably, a separate discrete short-time Fourier transform is performed on each element of the time-angle domain channel matrix; let... If is an element in the time-angle domain channel matrix, then the discrete short-time Fourier transform can be expressed as:

[0050] ;

[0051] in, This represents an element in the channel data represented by the time-frequency angular domain channel matrix. Indicates a time index. Indicates the frequency index; The window length represents the discrete short-time Fourier transform. Indicates the number of jumps. Indicates the window length is Window function, This represents an element in the time-angle domain channel matrix.

[0052] Each channel data element in the time-angle domain channel matrix is ​​then... All are transformed into channel data elements represented by the time-frequency angular domain channel matrix. Then, the channel data is represented by the time-angle domain channel matrix. The channel data was transformed into a time-frequency angular domain channel matrix representation. .

[0053] In this embodiment, the effects of the spatial Fourier transform and the discrete short-time Fourier transform on the channel matrix are orthogonal to each other. The effectiveness of these two transforms will be discussed below.

[0054] Through spatial Fourier transform, channel data can be transformed into a time-angular domain channel matrix representation. In the time-angular domain channel matrix, when the number of base station antennas is sufficient, any two multipaths are orthogonal to each other and can be separated. In other words, in the time-angular domain channel matrix, any two multipaths do not affect each other, making channel characteristics more apparent and channel processing easier. Furthermore, when the number of antennas is limited, most multipaths can remain nearly orthogonal in the time-angular domain channel matrix, with minimal influence between different multipaths. If the angles of two multipaths are close, their mutual influence will be greater, and they will be processed jointly. However, this is generally reasonable because multipaths with similar angles usually have similar propagation paths. Therefore, the time-angular domain channel matrix can effectively handle complex multipath scenarios.

[0055] The time-domain channel data can be transformed into a time-frequency domain channel matrix representation using the discrete short-time Fourier transform.

[0056] ;

[0057] in, This represents an element in the channel data represented by the time-frequency domain channel matrix; This represents the discrete Fourier transform of channel information that does not include Doppler frequencies, i.e., the channel information... Partially corresponding Discrete Fourier Transform; The window length represents the discrete short-time Fourier transform. Indicates the Doppler frequency. Indicates the frame duration. It is a function.

[0058] In the time-frequency domain channel matrix, the Doppler effect's influence is only related to the Doppler frequency. The Doppler frequency is related to the relative motion velocity. Since the estimation interval for acquiring known channel data by the channel estimator is typically on the order of milliseconds, the relative motion velocity remains essentially unchanged within such a short interval. Therefore, the Doppler frequency can be considered constant across consecutive frames. The effect will not change significantly. Therefore, the impact of the Doppler effect on the time-frequency domain channel matrix remains almost unchanged across consecutive frames, but it will cause frequency offset in the time-space domain channel matrix. Thus, by transforming the channel data represented by the time-space domain channel matrix to the channel data represented by the time-frequency domain channel matrix, the influence of the Doppler effect can be effectively handled, making channel prediction more accurate.

[0059] S23. Establish a millimeter-wave channel prediction network to receive known channel data output from the channel data converter and explore the relationship between known channel data and future channel data in order to predict future channel data.

[0060] Furthermore, the millimeter-wave channel prediction network outputs the predicted future channel data to the channel data inverse transformer.

[0061] Specifically, the millimeter-wave channel prediction network is based on Known channel data represented by a time-frequency angular domain channel matrix is ​​obtained. The future channel data is represented by a time-frequency angular domain channel matrix; where the channel data is represented by the time-frequency angular domain channel matrix. It includes information in three dimensions: time, frequency, and angle. When performing channel prediction, the channel data is divided according to the time dimension. Predicting using channel data from past times (i.e., known channel data) Channel prediction for future time slots (i.e., future channel data) can be represented as follows:

[0062] ;

[0063] in, This represents a millimeter-wave channel prediction network. It is the set of parameters for millimeter-wave channel prediction networks; The time exponent predicted by the millimeter-wave channel prediction network is: The future channel data represented by the time-frequency angular domain channel matrix; It includes a time index. Channel data information in terms of frequency and angle dimensions. ; Indicates the time index as Known channel data represented by the time-frequency angular domain channel matrix. ; This indicates the frequency index.

[0064] In this embodiment, the millimeter-wave channel prediction network can use common RNN channel prediction networks, LSTM channel prediction networks, or Transformer channel prediction networks; and before performing the channel prediction task, the millimeter-wave channel prediction network is pre-trained based on the channel data represented by the time-frequency angular domain channel matrix.

[0065] S24. Establish a channel data inverse transformer to inversely transform the channel data represented by the time-frequency angular domain channel matrix output by the millimeter-wave channel prediction network back to the channel data represented by the time-space domain channel matrix.

[0066] Specifically, a two-step transformation is used to transform the future channel data represented by the time-frequency angular domain channel matrix back to the channel data represented by the time-space domain channel matrix, such as... Figure 4 As shown.

[0067] S241. Transform the channel data represented by the time-frequency angular domain channel matrix to the channel data represented by the time-angular domain channel matrix through the discrete short-time Fourier inverse transform:

[0068] ;

[0069] in, This represents an element of the channel data represented by the time-angle domain channel matrix. This represents an element in the channel data represented by the time-frequency angular domain channel matrix; Indicates the window length is The synthesis window function, Indicates the window length is Window function, Represents the composite window function exist The function value at time t. Representing window functions exist Function value at time t; Composite window function With window function Satisfying the relation: , It is a constant.

[0070] Each channel data element in the time-frequency angular domain channel matrix is ​​then analyzed. All become channel data elements represented by the time-angle domain channel matrix. Subsequently, the channel data is represented by the time-frequency angular domain channel matrix. The channel data was transformed into a channel matrix representation in the time-angle domain. .

[0071] S242. Transform the channel data represented by the time-angle domain channel matrix to the channel data represented by the time-space domain channel matrix using the inverse Fourier transform in the spatial domain:

[0072] .

[0073] In this embodiment, mathematical modeling and simulation of the system were performed. The simulation used CDL-B channel data generated by MATLAB. The base station used a uniform linear array with 4 antennas, and the user used a uniform linear array with 2 antennas. The carrier frequency was 28 GHz, the user's moving speed was 60 km / h, and the signal-to-noise ratio was 20 dB. The simulation used 24 known spatiotemporal channel matrix representations of channel data to predict 6 future spatiotemporal channel matrix representations of channel data. The window function used for the Discrete Short-Time Fourier Transform (DSFT) was a rectangular window with a window length of 4 and a hop count of 2. Figures 5-7 As shown. Note that before using a millimeter-wave channel prediction network for channel prediction, the millimeter-wave channel prediction network needs to be trained. Specifically, if channel data represented by a spatiotemporal domain channel matrix is ​​used for channel prediction, the millimeter-wave channel prediction network needs to be trained in advance using the same channel data. Similarly, if channel data represented by a time-frequency angular domain channel matrix is ​​used for channel prediction, the millimeter-wave channel prediction network needs to be trained in advance using the same channel data.

[0074] from Figure 5 The comparison shows that after using the channel prediction preprocessing method of the present invention to transform the channel data into a time-frequency angular domain channel matrix representation, regardless of whether a millimeter-wave channel prediction network based on an RNN model, an LSTM model, or a Transformer model is used, the mean square error at various prediction lengths is smaller than the mean square error when using channel data represented by a time-space domain channel matrix for channel prediction, thus enabling better prediction of future channel conditions.

[0075] from Figure 6 The comparison shows that after transforming the channel data into a time-frequency angular domain channel matrix representation using the channel prediction preprocessing method of the present invention, the mean square error under various antenna configurations is smaller than the mean square error when using channel data represented by a time-space domain channel matrix for channel prediction, demonstrating the adaptability of the channel prediction preprocessing method of the present invention to different antenna configurations.

[0076] from Figure 7 The comparison shows that, compared to channel prediction using channel data represented by a time-angular domain channel matrix or a time-frequency spatial domain channel matrix, both methods can improve the accuracy of predicting future channel data. Furthermore, the preprocessing method of this invention, which transforms channel data into a time-frequency angular domain channel matrix for channel prediction, achieves even better channel prediction results. Therefore, the channel prediction preprocessing method of this invention can effectively improve the prediction performance of millimeter-wave channel prediction neural networks.

[0077] Example 2

[0078] Based on the same inventive concept as Embodiment 1 above, this embodiment provides a millimeter-wave channel prediction preprocessing system for multipath and Doppler scenarios. This system can be used to perform the steps of the above-described millimeter-wave channel prediction preprocessing method for multipath and Doppler scenarios.

[0079] Please see Figure 8 The millimeter-wave channel prediction preprocessing system 100 in this embodiment specifically includes the following modules:

[0080] Millimeter-wave MIMO channel model construction module 101 is used to establish a single-cell millimeter-wave MIMO channel model, including a base station and a user;

[0081] The channel prediction model establishment module 102 is used to establish a channel prediction model. It includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected to each other. The channel estimator is used to obtain known channel data based on the pilot signal. The channel data transformer preprocesses the channel data, transforming the channel data represented by the spatiotemporal domain channel matrix into the channel data represented by the time-frequency angular domain channel matrix. The millimeter-wave channel prediction network predicts future channel data based on the known channel data. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back into the channel data represented by the spatiotemporal domain channel matrix.

[0082] It should be noted that the millimeter-wave channel prediction preprocessing system and the millimeter-wave channel prediction preprocessing method of the present invention correspond one-to-one. The technical features and beneficial effects of the millimeter-wave channel prediction preprocessing method described in Example 1 are all applicable to the millimeter-wave channel prediction preprocessing system of Example 2. For details, please refer to the description in Example 1, which will not be repeated here.

[0083] Furthermore, in the above-described embodiment 2 of the millimeter-wave channel prediction preprocessing system for multipath and Doppler scenarios, the logical division of each module is merely an example. In practical applications, the corresponding functions can be assigned to different modules as needed, for example, for the sake of the corresponding hardware configuration requirements or the convenience of software implementation. That is, the internal structure of the millimeter-wave channel prediction preprocessing system can be divided into different modules to complete all or part of the functions described above.

[0084] In summary, the millimeter-wave channel prediction preprocessing method and preprocessing system for multipath and Doppler scenarios provided by this invention can improve the accuracy of channel prediction under complex and variable millimeter-wave channels, effectively alleviate the problem of channel data represented by spatiotemporal channel matrix being susceptible to interference in multipath and Doppler scenarios, achieve higher system capacity, and improve the quality of wireless communication.

[0085] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A millimeter-wave channel prediction preprocessing method for multipath and Doppler scenarios, characterized in that, Includes the following steps: S1. Establish a single-cell millimeter-wave MIMO channel model, including one base station and one user; S2. Establish a channel prediction model, which includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected together. The channel estimator is used to obtain known channel data based on the pilot signal sent by the user. The channel data transformer preprocesses the channel data. The millimeter-wave channel prediction network predicts future channel data based on the known channel data. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back to the channel data represented by the time-space domain channel matrix.

2. The millimeter-wave channel prediction preprocessing method according to claim 1, characterized in that, Step S2 includes: S21. Establish a channel estimator to obtain known channel data based on the pilot signals sent by the user; S22. Establish a channel data transformer to preprocess the known channel data obtained by the channel estimator and transform the channel data represented by the spatiotemporal domain channel matrix into the channel data represented by the time-frequency angular domain channel matrix. S23. Establish a millimeter-wave channel prediction network to receive known channel data output from the channel data converter and explore the relationship between known channel data and future channel data in order to predict future channel data. S24. Establish a channel data inverse transformer to inversely transform the channel data represented by the time-frequency angular domain channel matrix output by the millimeter-wave channel prediction network back to the channel data represented by the time-space domain channel matrix.

3. The millimeter-wave channel prediction preprocessing method according to claim 2, characterized in that, In the single-cell millimeter-wave MIMO channel model of step S1, there is a given... There are multiple path propagation paths, and the base station and the user are on the first path. The channel state information of a frame is represented as follows: ； in, Indicates the first Complex decay factor of each propagation path Indicates the first Doppler frequency shift of the propagation path Indicates frame duration, superscript This indicates the conjugate transpose. , Indicates the array steering vector. Indicates the first The angle of arrival of the path, Indicates the first The departure angle of the path.

4. The millimeter-wave channel preprocessing prediction method according to claim 3, characterized in that, Array guide vector and They are represented as follows: ； ； in, The wavelength of the carrier wave. Indicates antenna spacing. This indicates the number of transmit antennas at the base station. Indicates the number of transmit antennas for a user, superscript This indicates transpose.

5. The millimeter-wave channel prediction preprocessing method according to claim 2, characterized in that, Step S22 includes: S221. Transform the channel data represented by the spatiotemporal channel matrix into the channel data represented by the time-angle channel matrix using the spatial discrete Fourier transform: ； in, This represents the channel data represented by the spatiotemporal domain channel matrix. This represents the channel data represented by the time-angle domain channel matrix. Representing the discrete Fourier transform matrix: ； in, The number of points in the spatial domain discrete Fourier transform; matrix The superscript of each element is the product of the row and column numbers of the corresponding element, representing the exponentiation operation; S222. Transform the channel data represented by the time-angle domain channel matrix to the channel data represented by the time-frequency-angle domain channel matrix through the discrete short-time Fourier transform.

6. The millimeter-wave channel prediction preprocessing method according to claim 5, characterized in that, In step S222, a discrete short-time Fourier transform is performed on each element of the time-angle domain channel matrix; the discrete short-time Fourier transform is expressed as: ； in, This represents an element in the channel data represented by the time-frequency angular domain channel matrix. Indicates time index, Indicates the frequency index; The window length represents the discrete short-time Fourier transform. Indicates the number of jumps. Indicates the window length is Window function, This represents an element in the time-angle domain channel matrix.

7. The millimeter-wave channel prediction preprocessing method according to claim 2, characterized in that, In step S23, the millimeter-wave channel prediction network predicts according to... Known channel data represented by a time-frequency angular domain channel matrix is ​​obtained. The process of predicting future channel data, represented by a time-frequency angular domain channel matrix, is as follows: ； in, This represents a millimeter-wave channel prediction network. It is the set of parameters for millimeter-wave channel prediction networks; The time exponent predicted by the millimeter-wave channel prediction network is: The future channel data represented by the time-frequency angular domain channel matrix. ; Indicates time index as Known channel data represented by the time-frequency angular domain channel matrix. ; Indicates the frequency index; Indicates the time index.

8. The millimeter-wave channel prediction preprocessing method according to claim 2, characterized in that, Step S24 includes: S241. Transform the channel data represented by the time-frequency angular domain channel matrix to the channel data represented by the time-angular domain channel matrix through the discrete short-time Fourier inverse transform: ； in, This represents an element of the channel data represented by the time-angle domain channel matrix. This represents an element in the channel data represented by the time-frequency angular domain channel matrix; Indicates the window length is The synthesis window function, Indicates the window length is Window function, Represents the composite window function exist The function value at time t. Representing window functions exist The function value at time t; Indicates time index, Indicates the number of jumps; S242. Transform the channel data represented by the time-angle domain channel matrix to the channel data represented by the time-space domain channel matrix through the inverse Fourier transform in the spatial domain.

9. The millimeter-wave channel prediction preprocessing method according to claim 8, characterized in that, Synthesizing window function in step S241 With window function Satisfying the relation: , It is a constant.

10. A millimeter-wave channel prediction preprocessing system for multipath and Doppler scenarios, implemented based on the millimeter-wave channel prediction preprocessing method according to any one of claims 1-9, characterized in that, The system includes the following modules: The millimeter-wave MIMO channel model construction module includes a base station and a user; The channel prediction model building module is used to build a channel prediction model. It includes a channel estimator, a channel data transformer, a millimeter-wave channel prediction network, and a channel data inverse transformer connected to each other. The channel estimator is used to obtain known channel data based on the pilot signal. The channel data transformer preprocesses the channel data, transforming the channel data represented by the spatiotemporal domain channel matrix into the channel data represented by the time-frequency angular domain channel matrix. The millimeter-wave channel prediction network predicts future channel data based on the known channel data. The channel data inverse transformer transforms the channel data represented by the time-frequency angular domain channel matrix back into the channel data represented by the spatiotemporal domain channel matrix.

Images