A sub-6GHz channel-assisted hybrid precoding method for millimeter wave channel systems

Abstract

The present application relates to a sub-6GHz channel-assisted hybrid precoding method for a millimeter wave channel system, which utilizes sub-6GHz channel information to assist in realizing millimeter wave hybrid precoding. By designing a hybrid precoding prediction model, the model adopts a linear neural network layer structure or a deeply separable convolutional layer neural network structure. Under the premise of reducing the complexity of obtaining millimeter wave real-time channel information, it can effectively realize the estimation of millimeter wave hybrid precoding matrix in line-of-sight transmission and non-line-of-sight transmission scenarios.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system. Background Art

[0002] As a key technology in 6G, millimeter wave communication can provide high-bandwidth, low-latency and high-capacity communication performance. In order to improve the signal propagation capability of millimeter wave communication, precoding technology is used to enhance the received signal strength. Among them, there are three main types of precoding technology: digital precoding, analog precoding and hybrid precoding. Digital precoding has higher flexibility and accuracy, but higher complexity and implementation cost; analog precoding is simpler and more expensive but lacks adaptability and interference suppression capabilities; hybrid precoding can balance the disadvantages of digital precoding and analog precoding, and better improve communication performance.

[0003] Traditional hybrid precoding technologies mostly rely on the user's channel state information, which will bring about large pilot overhead and computational complexity. The computational complexity of the traditional signal processing process is reduced by introducing a neural network, but if it relies on partial or complete millimeter wave channel state information, it will still consume a large amount of time and frequency resources in the millimeter wave system. Due to the additional pilot overhead, the estimation process of millimeter wave channel information is more complicated. Summary of the invention

[0004] In order to overcome at least one deficiency in the prior art, the present application provides a sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system.

[0005] In a first aspect, a sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system is provided, comprising:

[0006] Constructing a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module;

[0007] Obtain sub-6GHz channel information, input the real part and imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end;

[0008] The hybrid precoding matrix is ​​determined according to the phase of the intermediate quantity of analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate quantity of analog beam precoding at the receiving end, the real part and the imaginary part of the digital beam precoding at the receiving end, combined with the power constraint and the constant modulus constraint; the hybrid precoding matrix includes the digital beam precoding matrix at the transmitting end, the digital beam precoding matrix at the receiving end, the analog beam precoding matrix at the transmitting end, and the analog beam precoding matrix at the receiving end.

[0009] In one embodiment, the feature extraction module includes two identical feature extraction branches, and the two feature extraction branches are used to extract features of the real part and the imaginary part of the sub-6GHz channel information, respectively, to obtain real part features and imaginary part features;

[0010] The real feature is divided into two parts, recorded as the first real feature and the second real feature, and the dimension of the first real feature is the same as the dimension of the digital beam precoding matrix at the transmitting end; the imaginary feature is divided into two parts, recorded as the first imaginary feature and the second imaginary feature; the dimension of the first imaginary feature is the same as the dimension of the digital beam precoding matrix at the receiving end;

[0011] The fusion module is used to fuse the second real feature and the second imaginary feature to obtain a fused feature;

[0012] The prediction module is used to predict the intermediate phase of the analog beam precoding at the transmitting end, the real and imaginary parts of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real and imaginary parts of the digital beam precoding at the receiving end based on the first real feature, the first imaginary feature, and the fused features.

[0013] In one embodiment, the feature extraction branch includes three feature extraction units connected in sequence, and each feature extraction unit includes a linear connection layer and a normalization layer.

[0014] In one embodiment, the fusion module includes a merging module, a linear connection layer, and a normalization layer connected in sequence.

[0015] In one embodiment, the hybrid precoding matrix is ​​determined using the following formula:

[0016]

[0017]

[0018]

[0019]

[0020] in, is the digital beam precoding matrix at the transmitting end, is the digital beam precoding matrix at the receiving end, Simulate the beam precoding matrix for the transmitter, Simulate the beam precoding matrix for the receiving end; is the F norm;

[0021] is the real part of the digital beam precoding at the transmitting end, is the imaginary part of the digital beam precoding at the transmitting end, is the number of subcarriers, is the real part of the digital beam precoding at the receiving end, is the imaginary part of the digital beam precoding at the receiving end, Simulate the intermediate phase of the beam precoding at the transmitter. Simulate the beam precoding intermediate phase at the receiving end, Is an imaginary unit.

[0022] In one embodiment, constructing a hybrid precoding prediction model includes:

[0023] The hybrid precoding prediction model is trained, and the loss function used in the training process is:

[0024]

[0025] Among them, Loss is the loss function and R is the channel spectrum efficiency.

[0026] In a second aspect, a sub-6 GHz channel-assisted hybrid precoding device for a millimeter wave channel system is provided, comprising:

[0027] A model building module is used to build a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module;

[0028] The intermediate quantity prediction module is used to obtain sub-6GHz channel information, input the real part and imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the intermediate quantity phase of the analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the intermediate quantity phase of the analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end;

[0029] The hybrid precoding matrix generation module is used to determine the hybrid precoding matrix according to the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, the real part and the imaginary part of the digital beam precoding at the receiving end, combined with the power constraint and the constant modulus constraint; the hybrid precoding matrix includes the digital beam precoding matrix at the transmitting end, the digital beam precoding matrix at the receiving end, the analog beam precoding matrix at the transmitting end, and the analog beam precoding matrix at the receiving end.

[0030] In a third aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned sub-6GHz channel-assisted hybrid precoding method for millimeter wave channel systems.

[0031] In a fourth aspect, a computer program product is provided, comprising a computer program / instruction, which, when executed by a processor, implements the above-mentioned sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system.

[0032] Compared with the prior art, this application has the following beneficial effects:

[0033] 1. This application uses sub-6GHz channel information to assist in realizing millimeter wave hybrid precoding. On the premise of reducing the complexity of obtaining millimeter wave real-time channel information, it can effectively realize the estimation of millimeter wave hybrid precoding matrix in line-of-sight transmission and non-line-of-sight transmission scenarios.

[0034] 2. This application designs a hybrid precoding prediction model, which adopts a linear neural network layer structure or a deep separable convolutional layer neural network structure, and can effectively realize the millimeter wave hybrid beam precoding matrix estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present application may be better understood by referring to the following description given in conjunction with the accompanying drawings, which together with the following detailed description are included in this specification and form a part of this specification. In the drawings:

[0036] Figure 1 shows a schematic diagram of a millimeter wave and sub-6GHz radio transceiver;

[0037] Figure 2 A flow chart of a sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system according to an embodiment of the present application is shown;

[0038] Figure 3 A schematic diagram of a hybrid precoding prediction model according to an embodiment of the present application is shown;

[0039] Figure 4 A schematic diagram of a hybrid precoding prediction model according to another embodiment of the present application is shown;

[0040] Figure 5 A structural block diagram of a sub-6 GHz channel-assisted hybrid precoding device for a millimeter wave channel system according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0041] The exemplary embodiments of the present application will be described below in conjunction with the accompanying drawings. For the sake of clarity and conciseness, not all features of the actual embodiments are described in the specification. However, it should be understood that many implementation-specific decisions can be made in the process of developing any such actual embodiments in order to achieve the specific goals of the developer, and these decisions may vary from embodiment to embodiment.

[0042] It is also necessary to explain here that, in order to avoid obscuring the present application due to unnecessary details, only the device structure closely related to the scheme according to the present application is shown in the drawings, while other details that are not closely related to the present application are omitted.

[0043] It should be understood that the present application is not limited to the described implementation forms due to the following description with reference to the accompanying drawings. In this article, where feasible, the embodiments can be combined with each other, features between different embodiments can be replaced or borrowed, and one or more features can be omitted in one embodiment.

[0044] Figure 1 A schematic diagram of a millimeter wave and sub-6GHz radio transceiver is shown. The present application provides a sub-6GHz channel-assisted hybrid precoding method for a millimeter wave channel system for millimeter wave and sub-6GHz radio transceivers, and designs a self-supervised learning framework to predict the millimeter wave downlink hybrid precoding matrix based on the sub-6GHz uplink channel. The present application uses sub-6GHz channel state information as out-of-band information to solve the millimeter wave channel overhead problem. The method can achieve the acquisition of the millimeter wave hybrid beam precoding matrix with low computational complexity and parameter quantity.

[0045] Figure 2 A flow chart of a sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system according to an embodiment of the present application is shown. Figure 2 , the method comprises the following steps:

[0046] Step S1, constructing a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module.

[0047] Step S2, obtain sub-6GHz channel information, input the real part and imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end.

[0048] Step S3, according to the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, the real part and the imaginary part of the digital beam precoding at the receiving end, combined with the power constraint and the constant modulus constraint, determine the hybrid precoding matrix; the hybrid precoding matrix includes the digital beam precoding matrix at the transmitting end, the digital beam precoding matrix at the receiving end, the analog beam precoding matrix at the transmitting end, and the analog beam precoding matrix at the receiving end.

[0049] This embodiment uses sub-6GHz channel information to assist in realizing millimeter wave hybrid precoding. Under the premise of reducing the complexity of acquiring millimeter wave real-time channel information, it can effectively realize the estimation of millimeter wave hybrid precoding matrix in line-of-sight transmission and non-line-of-sight transmission scenarios.

[0050] In one embodiment, the hybrid precoding prediction model may adopt a linear neural network layer structure. Figure 3 A schematic diagram of a hybrid precoding prediction model according to an embodiment of the present application is shown.

[0051] See also Figure 3 ,The feature extraction module includes two identical feature extraction branches, which are respectively used to extract the real and imaginary parts of the sub-6GHz channel information to obtain the real and imaginary features.

[0052] The real feature is divided into two parts, recorded as the first real feature and the second real feature, and the dimension of the first real feature is the same as the dimension of the digital beam precoding matrix at the transmitting end; the imaginary feature is divided into two parts, recorded as the first imaginary feature and the second imaginary feature; the dimension of the first imaginary feature is the same as the dimension of the digital beam precoding matrix at the receiving end.

[0053] The fusion module is used to fuse the second real feature and the second imaginary feature to obtain a fused feature.

[0054] The prediction module is used to predict the intermediate phase of the analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end according to the first real feature, the first imaginary feature, and the fused feature. Here, the prediction module adopts an existing network structure, for example, it can be a prediction module in a convolutional neural network (CNN).

[0055] After using the prediction module to predict the intermediate phase of the analog beam precoding at the transmitting end, the real and imaginary parts of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real and imaginary parts of the digital beam precoding at the receiving end, the hybrid precoding matrix is ​​determined in combination with the power constraint and the constant modulus constraint; the prediction module and the specific implementation function of this step can be implemented using a matrix reconstruction module.

[0056] Specifically, the feature extraction branch includes three feature extraction units connected in sequence, and each feature extraction unit includes a linear connection layer and a normalization layer.

[0057] Specifically, the fusion module includes a merging module, a linear connection layer and a normalization layer which are connected in sequence.

[0058] In this embodiment, the feature extraction module is composed of multiple linear connection layers and normalization layers. The real and imaginary parts of the channel are combined into a dual-path parallel structure to obtain the decoupling of the real and imaginary features. Secondly, the interval fusion method is used to represent the real and imaginary features as segmented intervals of different feature lengths; then, the corresponding intervals are fused to obtain complex features with complete channel information.

[0059] In other embodiments, the hybrid precoding prediction model may adopt a deep separable convolutional layer neural network structure. Figure 4 A schematic diagram of a hybrid precoding prediction model according to another embodiment of the present application is shown.

[0060] See also Figure 4, the feature extraction module adopts a pyramid structure, and replaces the original convolution layer in the pyramid structure with a convolution layer with depth separable, and extracts features from the real and imaginary parts of the input sub-6GHz channel information to obtain real features, imaginary features and amplitudes. The pyramid structure includes a depth-separable convolution layer and a pooling operation, which can analyze channel features of different dimensions and effectively control the computational complexity of the convolution operation. Through the pooling operation, the number of model channels increases layer by layer, and the size of each layer of feature maps decreases layer by layer. With the increase in the number of channels of the traditional two-dimensional convolution model, the number of model parameters and the computational complexity also increase. Therefore, this embodiment designs a depth-separable convolution structure. Only one convolution kernel is used to convolve each input model channel, thereby reducing the number of model parameters. The depth-separable convolution fully learns the features, and the dimensional transformation of the model channel can be achieved through point convolution. Therefore, by enhancing the information interaction across channels, the representation capabilities of the network width, nonlinear characteristics and patterns are increased with minimal parameters.

[0061] In the feature fusion module, the interval fusion structure is adopted to realize the fusion of real features and imaginary features; the fused features and amplitudes are input into the prediction module to predict the intermediate phase of the analog beam precoding at the transmitting end, the real and imaginary parts of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real and imaginary parts of the digital beam precoding at the receiving end.

[0062] In this embodiment, the prediction module is obtained by improving the prediction module in the existing network structure, specifically replacing the convolutional layer therein with a depth-separable convolutional layer. The existing network structure may be a convolutional neural network CNN.

[0063] In one embodiment, the hybrid precoding matrix is ​​determined using the following formula:

[0064]

[0065]

[0066]

[0067]

[0068] in, is the digital beam precoding matrix at the transmitting end, is the digital beam precoding matrix at the receiving end, Simulate the beam precoding matrix for the transmitter, Simulate the beam precoding matrix for the receiving end; is the F norm;

[0069] is the real part of the digital beam precoding at the transmitting end, is the imaginary part of the digital beam precoding at the transmitting end, is the number of subcarriers, is the real part of the digital beam precoding at the receiving end, is the imaginary part of the digital beam precoding at the receiving end, Simulate the intermediate phase of the beam precoding at the transmitter. Simulate the beam precoding intermediate phase at the receiving end, Is an imaginary unit.

[0070] In this embodiment, the power constraint and the constant modulus constraint in millimeter wave hybrid precoding are combined so that the generated hybrid precoding matrix meets the constraint conditions.

[0071] In one embodiment, constructing a hybrid precoding prediction model includes:

[0072] The hybrid precoding prediction model is trained, and the loss function used in the training process is:

[0073]

[0074] Wherein, Loss is the loss function, and R is the channel spectrum efficiency. Here, the channel spectrum efficiency is constructed according to the hybrid precoding matrix.

[0075] In this embodiment, the network structure evaluation criterion is designed based on the channel spectrum efficiency, and the loss function is constructed based on the channel spectrum efficiency, which can avoid the impact of gradient explosion on the model prediction results as much as possible.

[0076] Furthermore, in order to improve the generalization ability of the model, curriculum learning is introduced during the training process. The existing deep learning-based methods establish a joint optimization model of multi-channel matrices and establish a mapping relationship between data and matrices. However, when the number of samples becomes larger and more complex, some models cannot be effectively applied, which increases the difficulty of model prediction. During the training process, data subsets are gradually introduced from easy to difficult according to the training plan. Iterative training allows the model to learn over time and adapt to increasingly challenging scenarios, starting from acquiring basic knowledge from simple samples and gradually transitioning to more challenging samples to gain a deeper understanding. The optimal training strategy is obtained by dynamically adjusting the ratio of simple and challenging samples.

[0077] Based on the same inventive concept as the sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system, this embodiment also provides a corresponding sub-6 GHz channel-assisted hybrid precoding device for a millimeter wave channel system. Figure 5The structure block diagram of a sub-6 GHz channel-assisted hybrid precoding device for a millimeter wave channel system according to an embodiment of the present application is shown, including:

[0078] The model building module 51 is used to build a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module;

[0079] The intermediate quantity prediction module 52 is used to obtain sub-6GHz channel information, input the real part and imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the intermediate quantity phase of the analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the intermediate quantity phase of the analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end;

[0080] The hybrid precoding matrix generation module 53 is used to determine the hybrid precoding matrix according to the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, the real part and the imaginary part of the digital beam precoding at the receiving end, combined with the power constraint and the constant modulus constraint; the hybrid precoding matrix includes the digital beam precoding matrix at the transmitting end, the digital beam precoding matrix at the receiving end, the analog beam precoding matrix at the transmitting end, and the analog beam precoding matrix at the receiving end.

[0081] The hybrid precoding device for sub-6GHz channel assistance for millimeter wave channel system of this embodiment has the same inventive concept as the hybrid precoding method for sub-6GHz channel assistance for millimeter wave channel system mentioned above. Therefore, the specific implementation method of the device can be seen in the implementation example part of the hybrid precoding method for sub-6GHz channel assistance for millimeter wave channel system mentioned above, and its technical effect corresponds to the technical effect of the above method, which will not be repeated here.

[0082] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned sub-6GHz channel-assisted hybrid precoding method for a millimeter wave channel system.

[0083] An embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system.

[0084] The above are only various implementations of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A sub-6GHz channel-assisted hybrid precoding method for a millimeter wave channel system, characterized in that: include: Constructing a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module; Obtain sub-6GHz channel information, input the real part and the imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the intermediate phase of the analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real part and the imaginary part of the digital beam precoding at the receiving end; Determine a hybrid precoding matrix according to the phase of the transmitting end analog beam precoding intermediate quantity, the real part and the imaginary part of the transmitting end digital beam precoding, the phase of the receiving end analog beam precoding intermediate quantity, the real part and the imaginary part of the receiving end digital beam precoding, combined with power constraints and constant modulus constraints; the hybrid precoding matrix includes the transmitting end digital beam precoding matrix, the receiving end digital beam precoding matrix, the transmitting end analog beam precoding matrix, and the receiving end analog beam precoding matrix; The feature extraction module includes two identical feature extraction branches, and the two feature extraction branches are used to extract features of the real part and the imaginary part of the sub-6GHz channel information respectively to obtain real part features and imaginary part features; The real part feature is divided into two parts, recorded as a first real part feature and a second real part feature, and the dimension of the first real part feature is the same as the dimension of the digital beam precoding matrix at the transmitting end; the imaginary part feature is divided into two parts, recorded as a first imaginary part feature and a second imaginary part feature; the dimension of the first imaginary part feature is the same as the dimension of the digital beam precoding matrix at the receiving end; The fusion module is used to fuse the second real feature and the second imaginary feature to obtain a fused feature; The prediction module is used to predict the intermediate phase of the analog beam precoding at the transmitting end, the real and imaginary parts of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real and imaginary parts of the digital beam precoding at the receiving end based on the first real feature, the first imaginary feature, and the fused feature.

2. The method according to claim 1, characterized in that The feature extraction branch includes three feature extraction units connected in sequence, and each feature extraction unit includes a linear connection layer and a normalization layer.

3. The method according to claim 1, characterized in that The fusion module includes a merging module, a linear connection layer and a normalization layer which are connected in sequence.

4. The method according to claim 1, characterized in that The hybrid precoding matrix is ​​determined using the following formula: in, is the digital beam precoding matrix at the transmitting end, is the digital beam precoding matrix at the receiving end, Simulate the beam precoding matrix for the transmitter, Simulate the beam precoding matrix for the receiving end; is the F norm; is the real part of the digital beam precoding at the transmitting end, is the imaginary part of the digital beam precoding at the transmitting end, is the number of subcarriers, is the real part of the digital beam precoding at the receiving end, is the imaginary part of the digital beam precoding at the receiving end, Simulate the intermediate phase of the beam precoding at the transmitter. Simulate the beam precoding intermediate phase at the receiving end, Is an imaginary unit.

5. The method according to claim 1, characterized in that The constructing of the hybrid precoding prediction model comprises: The hybrid precoding prediction model is trained, and the loss function used in the training process is: Among them, Loss is the loss function and R is the channel spectrum efficiency.

6. A sub-6GHz channel-assisted hybrid precoding device for a millimeter wave channel system, characterized in that: include: A model building module, used to build a hybrid precoding prediction model; the hybrid precoding prediction model includes a feature extraction module, a feature fusion module and a prediction module; An intermediate quantity prediction module is used to obtain sub-6GHz channel information, input the real part and imaginary part of the sub-6GHz channel information into the hybrid precoding prediction model, and predict the intermediate quantity phase of the analog beam precoding at the transmitting end, the real part and imaginary part of the digital beam precoding at the transmitting end, the intermediate quantity phase of the analog beam precoding at the receiving end, and the real part and imaginary part of the digital beam precoding at the receiving end; A hybrid precoding matrix generation module is used to determine a hybrid precoding matrix according to the phase of the intermediate amount of analog beam precoding at the transmitting end, the real part and the imaginary part of the digital beam precoding at the transmitting end, the phase of the intermediate amount of analog beam precoding at the receiving end, the real part and the imaginary part of the digital beam precoding at the receiving end, combined with power constraints and constant modulus constraints; the hybrid precoding matrix includes the digital beam precoding matrix at the transmitting end, the digital beam precoding matrix at the receiving end, the analog beam precoding matrix at the transmitting end, and the analog beam precoding matrix at the receiving end; The feature extraction module includes two identical feature extraction branches, and the two feature extraction branches are used to extract features of the real part and the imaginary part of the sub-6GHz channel information respectively to obtain real part features and imaginary part features; The real part feature is divided into two parts, recorded as a first real part feature and a second real part feature, and the dimension of the first real part feature is the same as the dimension of the digital beam precoding matrix at the transmitting end; the imaginary part feature is divided into two parts, recorded as a first imaginary part feature and a second imaginary part feature; the dimension of the first imaginary part feature is the same as the dimension of the digital beam precoding matrix at the receiving end; The fusion module is used to fuse the second real feature and the second imaginary feature to obtain a fused feature; The prediction module is used to predict the intermediate phase of the analog beam precoding at the transmitting end, the real and imaginary parts of the digital beam precoding at the transmitting end, the intermediate phase of the analog beam precoding at the receiving end, and the real and imaginary parts of the digital beam precoding at the receiving end based on the first real feature, the first imaginary feature, and the fused feature.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the sub-6GHz channel-assisted hybrid precoding method for a millimeter wave channel system as described in any one of claims 1-5.

8. A computer program product, characterized in that It includes a computer program / instruction, which, when executed by a processor, implements the sub-6 GHz channel-assisted hybrid precoding method for a millimeter wave channel system as described in any one of claims 1 to 5.

Images