Channel state information acquisition method and device, channel state information generation method and device and network equipment

Abstract

The invention provides a channel state information acquisition method, a channel state information generation method, a channel state information acquisition device, a channel state information generation device and network equipment. The method comprises the following steps: acquiring a first signal; under the condition that the duplex mode of the first signal is FDD, inputting the first input signal into a processing module, performing data processing by the processing module according to the first input signal, and outputting predicted downlink channel state information of the terminal; under the condition that the duplex mode of the first signal is TDD, inputting a second input signal into a first module in the signal processing module, performing data processing by the first module according to the second input signal, and outputting predicted downlink channel state information of the terminal; with the adoption of the method, a set of processing module obtained through model training is utilized, so that CSI acquisition of signals with FDD (Frequency Division Duplexing) as a duplex mode can be realized, and CSI acquisition of signals with TDD (Time Division Duplexing) as a duplex mode can also be realized.

Description

Technical Field

[0001] The present application relates to the field of wireless technology, and in particular to a channel state information acquisition method, generation method, device and network equipment. Background Art

[0002] Depending on whether the duplex mode of the wireless communication system is time division duplex (TDD) or frequency division duplex (FDD), the scheme for obtaining channel state information (CSI) is designed separately. In the TDD system, based on channel reciprocity, the base station obtains the downlink CSI based on the estimated uplink CSI. The downlink CSI obtained in this way has a certain error; in the FDD system, the terminal feeds back the downlink CSI to the base station through the precoding matrix index (PMI). The feedback overhead increases with the increase in the number of antennas. The rapid increase in feedback overhead will seriously reduce the capacity gain that can be brought by the ultra-large-scale antenna system.

[0003] In summary, the existing CSI acquisition methods do not consider the common channel reciprocity feature of TDD and FDD to design a unified CSI feedback method to further reduce the overhead of CSI feedback or improve the accuracy of CSI feedback. Summary of the invention

[0004] The purpose of the present application is to provide a channel state information acquisition method, generation method, device and network equipment, which uses a set of processing modules obtained through model training to achieve CSI acquisition of signals with FDD duplex mode and CSI acquisition of signals with TDD duplex mode.

[0005] One embodiment of the present application provides a method for acquiring channel state information, wherein the method includes:

[0006] Acquire a first signal sent by a terminal;

[0007] In the case where the duplex mode of the first signal is frequency division duplex FDD, the first input signal is input into the processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency point information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0008] When the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, and the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0009] Optionally, in the method, the first input signal further comprises:

[0010] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0011] Optionally, in the method, the second input signal further comprises:

[0012] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0013] Optionally, in the method, the processing module further comprises a second module; a signal output terminal of the second module is connected to a signal input terminal of the first module;

[0014] The step of inputting the first input signal into the processing module includes:

[0015] The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0016] Optionally, in the method, the processing module further comprises a second module; the signal output terminal of the first module is connected to the signal input terminal of the second module;

[0017] The step of inputting the first input signal into the processing module includes:

[0018] The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0019] Optionally, in the method, the at least two modules in the processing module respectively include a plurality of data processing models;

[0020] The method further comprises:

[0021] The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

[0022] One embodiment of the present application further provides a method for generating a processing module for channel state information prediction, wherein the method comprises:

[0023] Acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0024] According to the second channel state information at multiple historical times, the first neural network model is trained to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0025] Keeping the parameters of the first data processing model unchanged, train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0026] Optionally, in the method, the first channel state information further includes:

[0027] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0028] Optionally, in the method, the second channel state information further includes:

[0029] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0030] Optionally, the method, wherein the first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at a next historical time corresponding to the plurality of second uplink channel states in the first data set;

[0031] The method of training the first neural network model according to the second channel state information at multiple historical times to obtain the first model includes:

[0032] The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

[0033] Optionally, the method, wherein the first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at a next historical time corresponding to the plurality of first uplink channel states in the first data set;

[0034] Keeping the parameters of the first data processing model unchanged, training at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtaining the second data processing model of the second module, including:

[0035] Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

[0036] Optionally, in the method, the signal output terminal of the second module is connected to the signal input terminal of the first module;

[0037] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including:

[0038] The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

[0039] Optionally, the method, wherein the signal output terminal of the first module is connected to the signal input terminal of the second module;

[0040] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including:

[0041] The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

[0042] Optionally, the method, wherein the second channel state information at multiple historical times is used as training data, input into the first module, the first label data is used as a training label, the first neural network model is trained, and the first data processing model is obtained, includes one or more of the following:

[0043] The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module;

[0044] The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

[0045] Optionally, the method, wherein the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time are used as training data, input into the processing module, and the second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, includes one or more of the following:

[0046] The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module;

[0047] The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

[0048] One embodiment of the present application provides a network device, which includes a memory, a transceiver, and a processor:

[0049] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0050] Acquire a first signal sent by a terminal;

[0051] In the case where the duplex mode of the first signal is frequency division duplex FDD, the first input signal is input into the processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency point information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0052] When the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, and the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0053] Optionally, in the network device, the first input signal further includes:

[0054] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0055] Optionally, in the network device, the second input signal further includes:

[0056] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0057] Optionally, in the network device, the processing module further comprises a second module; a signal output terminal of the second module is connected to a signal input terminal of the first module;

[0058] The processor reads the computer program in the memory and inputs the first input signal into the processing module, including:

[0059] The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0060] Optionally, in the network device, the processing module further comprises a second module; the signal output end of the first module is connected to the signal input end of the second module;

[0061] The processor reads the computer program in the memory and inputs the first input signal into the processing module, including:

[0062] The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0063] Optionally, in the network device, the at least two modules in the processing module respectively include a plurality of data processing models;

[0064] The processor is further configured to read the computer program in the memory and perform the following operations:

[0065] The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

[0066] One embodiment of the present application provides a network device, which includes a memory, a transceiver, and a processor:

[0067] A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations:

[0068] Acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0069] According to the second channel state information at multiple historical times, the first neural network model is trained to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0070] Keeping the parameters of the first data processing model unchanged, train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0071] Optionally, in the network device, the first channel state information further includes:

[0072] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0073] Optionally, in the network device, the second channel state information further includes:

[0074] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0075] Optionally, the network device, wherein the first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of the second uplink channel states in the first data set;

[0076] The processor reads the computer program in the memory, trains the first neural network model according to the second channel state information at multiple historical times, and obtains the first data processing model, including:

[0077] The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

[0078] Optionally, the network device, wherein the first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of first uplink channel states in the first data set;

[0079] The processor reads the computer program in the memory, keeps the parameters of the first data processing model unchanged, trains at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtains the second data processing model of the second module, including:

[0080] Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

[0081] Optionally, in the network device, the signal output terminal of the second module is connected to the signal input terminal of the first module;

[0082] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor reads the computer program in the memory, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times as training data, and inputs them into the processing module, including:

[0083] The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

[0084] Optionally, in the network device, the signal output terminal of the first module is connected to the signal input terminal of the second module;

[0085] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor reads the computer program in the memory, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times as training data, and inputs them into the processing module, including:

[0086] The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

[0087] Optionally, the network device, wherein the processor reads the computer program in the memory, uses the second channel state information of multiple historical times as training data, inputs it into the first module, uses the first label data as a training label, trains the first neural network model, and obtains the first data processing model, including one or more of the following:

[0088] The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module;

[0089] The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

[0090] Optionally, in the network device, the processor reads the computer program in the memory,

[0091] The first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time are used as training data and input into the processing module. The second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following:

[0092] The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module;

[0093] The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

[0094] One embodiment of the present application further provides a channel state information acquisition device, wherein the device includes:

[0095] A signal acquisition unit, configured to acquire a first signal sent by a terminal;

[0096] A first input unit is used to input the first input signal into a processing module when the duplex mode of the first signal is frequency division duplex FDD, and the processing module performs data processing according to the first input signal and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0097] A second input unit is used to input the second input signal into the first module in the processing module when the duplex mode of the first signal is time division duplex TDD, and the first module performs data processing according to the second input signal and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0098] One embodiment of the present application further provides a device for generating a processing module for channel state information prediction, wherein the device comprises:

[0099] A data acquisition unit, configured to acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0100] A first training unit is used to train a first neural network model according to the second channel state information at multiple historical times to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0101] The second training unit is used to keep the parameters of the first data processing model unchanged, and train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times to obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0102] One embodiment of the present application further provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute any of the methods described above.

[0103] The beneficial effects of the above technical solution of the present application are as follows:

[0104] By adopting the method described in the embodiment of the present application and utilizing a set of processing modules obtained through model training, it is possible to acquire CSI for signals whose duplex mode is FDD and also to acquire CSI for signals whose duplex mode is TDD, thereby achieving unified TDD and FDD channel prediction, thereby reducing or avoiding the CSI feedback overhead of the FDD system and improving the accuracy of the downlink CSI acquired by the TDD system. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1A schematic diagram of a flow chart of a method for acquiring channel state information according to an embodiment of the present application;

[0106] Figure 2 This is a signal diagram of the first module in an embodiment of the present application;

[0107] Figure 3 This is a schematic diagram of one implementation structure of a processing module in an embodiment of the present application;

[0108] Figure 4 Another schematic diagram of the structure of the processing module in the embodiment of the present application;

[0109] Figure 5 A schematic flow chart of a method for generating a processing module for channel state information prediction according to an embodiment of the present application;

[0110] Figure 6 This is a schematic diagram of the structure of a network device according to one embodiment of the present application;

[0111] Figure 7 This is a schematic diagram of the structure of a network device according to another embodiment of the present application;

[0112] Figure 8 This is a schematic diagram of the structure of the channel state information acquisition device described in an embodiment of the present application;

[0113] Fig. 9 This is a schematic diagram of the structure of the generating device described in the embodiment of the present application. DETAILED DESCRIPTION

[0114] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0115] In the embodiments of the present application, the term "and / or" describes the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0116] In the embodiments of the present application, the term "plurality" refers to two or more than two, and other quantifiers are similar.

[0117] The technical solution provided in the embodiment of the present application can be applicable to a variety of systems, especially 5G systems. For example, the applicable system can be a global system of mobile communication (GSM) system, a code division multiple access (CDMA) system, a wideband code division multiple access (WCDMA) general packet radio service (GPRS) system, a long term evolution (LTE) system, a LTE frequency division duplex (FDD) system, a LTE time division duplex (TDD) system, an advanced long term evolution (LTE-A) system, a universal mobile telecommunication system (UMTS), a world-wide interoperability for microwave access (WiMAX) system, a 5G new air interface (NR) system, etc. These various systems include terminal equipment and network equipment. The system may also include a core network part, such as an evolved packet.

[0118] The following first introduces the contents involved in the solution provided in the embodiment of the present application.

[0119] Currently, the CSI acquisition scheme is designed separately and independently according to whether the duplex mode of the wireless communication system is time division duplex TDD or frequency division duplex FDD.

[0120] For TDD systems, the terminal sends an uplink sounding reference signal (SRS). After receiving the SRS, the base station estimates the uplink channel. Then, based on channel reciprocity, the uplink CSI estimated by the base station is considered to be the downlink CSI. However, due to various non-ideal factors, channel reciprocity is not strictly true, that is, there are still differences between the uplink channel and the downlink channel in the TDD system. Therefore, there is a certain error in obtaining the downlink CSI based on channel reciprocity.

[0121] For FDD systems, the base station sends a downlink reference signal, such as a channel state information reference signal (CSI-RS). After receiving the CSI-RS, the terminal estimates the downlink channel, and then feeds back the estimated CSI to the base station. In order to reduce feedback overhead, the commonly used CSI feedback method in FDD systems is to feed back the downlink CSI to the base station through the precoding matrix index (PMI), rather than feeding back the entire downlink channel estimated by the terminal to the base station. Specifically, the terminal selects the most appropriate codebook from a predefined codebook set based on the principle of maximum system capacity and other principles and feeds it back to the base station. The base station performs downlink beamforming based on the PMI fed back by the terminal. In FDD systems, even for codebook-based CSI feedback schemes, the feedback overhead will increase with the increase in the number of antennas. In particular, in future ultra-large-scale antenna systems, the rapid increase in feedback overhead will seriously reduce the capacity gain that can be brought by ultra-large-scale antenna systems.

[0122] Therefore, the existing CSI acquisition method does not consider the common channel reciprocity feature of TDD and FDD to design a unified CSI feedback method to further reduce the overhead of CSI feedback or improve the accuracy of CSI feedback.

[0123] To solve the above problems, an embodiment of the present application provides a method for generating a processing module for channel state information prediction and a method for acquiring channel state information. Through the method for generating the processing module, the processing module can be obtained through neural network model training, which can realize CSI acquisition of signals with a duplex mode of FDD and CSI acquisition of signals with a duplex mode of TDD. The processing module generated by this method can obtain the downlink channel state information of the terminal according to the type of signal sent by the terminal, and realize unified TDD and FDD channel prediction, thereby reducing or avoiding the CSI feedback overhead of the FDD system and improving the accuracy of the TDD system in acquiring downlink CSI.

[0124] One embodiment of the present application provides a method for obtaining channel state information, such as Figure 1 As shown, the method includes:

[0125] S110, acquiring a first signal sent by a terminal;

[0126] S120, when the duplex mode of the first signal is frequency division duplex FDD, input the first input signal into a processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0127] S130, when the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0128] By using the method described in the embodiment of the present application, when the duplex mode of the first signal sent by the terminal is TDD, the first module in the processing module is used to process the second input signal to obtain the predicted downlink channel state information of the terminal; when the duplex mode of the first signal sent by the terminal is FDD, the processing module is used to process the first input signal to obtain the predicted downlink channel state information of the terminal. Therefore, by using a set of processing modules obtained through model training, it is possible to obtain CSI for signals with a duplex mode of FDD and signals with a duplex mode of TDD.

[0129] Optionally, the channel state information acquisition method is executed by a network side, such as by a base station.

[0130] The method described in the embodiment of the present application can use a set of processing modules obtained through model training to obtain the downlink channel state information of the terminal for signals with duplex modes of TDD and FDD. Compared with the prior art method of using the uplink CSI estimated by the base station to determine the downlink CSI for signals with duplex mode of TDD, the accuracy of the obtained downlink CSI can be improved; compared with the prior art method of using the terminal to estimate the CSI and feed it back to the base station for signals with duplex mode of FDD, the overhead of CSI feedback can be reduced or avoided.

[0131] It should be noted that, in the embodiment of the present application, the above-mentioned first module is one of the at least two modules in the processing module that perform data processing on the first input signal when the first signal is FDD.

[0132] In one embodiment of the present application, optionally, the first input signal further includes:

[0133] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0134] Optionally, the second input signal further includes:

[0135] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0136] In an embodiment of the present application, the first module is one of the modules in a processing module including multiple modules, and the first module is used to output the predicted downlink channel state information of the terminal at the next moment based on the uplink channel state information at the previous moment and at least part of the downlink channel state information corresponding to the uplink channel state information at the previous moment fed back by the terminal.

[0137] Optionally, the base station obtains a first signal, which is an uplink reference signal. When the time division duplex of the first signal is TDD, the base station estimates the uplink channel state information of the previous moment according to the uplink reference signal, that is, obtains second uplink channel state information.

[0138] Alternatively, if Figure 2 As shown, the second input signal includes the uplink channel state information of the previous moment (second uplink channel state information), which is input into the first module. The first module outputs the downlink channel state information of the next moment according to the uplink channel state information of the previous moment; that is, the predicted downlink channel state information is obtained.

[0139] Optionally, Figure 2 As shown, the second input signal may also include at least part of the downlink channel state information corresponding to the second uplink channel state information fed back by the terminal, that is, including the second downlink channel state information. The second input signal includes the second uplink channel state information and the second downlink channel state information, which are input into the first module, and the first module outputs the downlink channel state information at the next moment according to the uplink channel state information (the second uplink channel state information) at the previous moment and the downlink channel state information (the second downlink channel state information) corresponding to the uplink channel state information fed back by the terminal; that is, the predicted downlink channel state information is obtained.

[0140] Optionally, the machine learning model (or neural network model) adopted by the first module includes but is not limited to: one or more of: recurrent neural network (RNN), long short-term memory network (LSTM), multi-layer perception, neural network based on attention mechanism, etc.

[0141] In the embodiments of the present application, one implementation method, optionally, Figure 3 As shown, the processing module further includes a second module, and a signal output end of the second module is connected to a signal input end of the first module;

[0142] The step of inputting the first input signal into the processing module includes:

[0143] The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs the preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0144] In this embodiment, the processing module includes a first module and a second module cascaded before a signal input end of the first module.

[0145] Among them, Figure 3 As shown, the input signal of the second module includes: first uplink channel state information and uplink and downlink frequency information of the terminal;

[0146] Optionally, when the duplex mode of the first signal is FDD, the base station estimates the uplink channel state information according to the uplink reference signal (first signal), that is, obtains the first uplink channel state information.

[0147] The output signal of the second module includes: preprocessed uplink channel state information corresponding to the first uplink channel state information; the second module is used to preprocess the input first uplink channel state information so that the preprocessed uplink channel state can represent the channel characteristics of the corresponding downlink frequency information.

[0148] Optionally, the machine learning model adopted by the second module is a neural network model, including but not limited to one or more of a deep neural network (DNN), a recurrent neural network (RNN), a neural network based on an attention mechanism, and a transformer model.

[0149] Optionally, in this embodiment, one implementation method, such as Figure 3 As shown, the input signal of the first module includes: the pre-processed uplink channel state information output by the second module;

[0150] The output signal of the first module includes: downlink channel state information corresponding to the terminal and the corresponding downlink frequency point information. In this implementation mode, the first module obtains the downlink channel state information corresponding to the corresponding downlink frequency point information at the next moment according to the preprocessed uplink channel state information at the previous moment; that is, obtains the predicted downlink channel state information.

[0151] Another embodiment, such as Figure 3 As shown, the input signal of the first module includes: the preprocessed uplink channel state information output by the second module and the first downlink channel state information; wherein the first downlink channel state information is also at least part of the downlink channel state information corresponding to the first uplink channel state information fed back by the terminal.

[0152] In this implementation, the output signal of the first module includes: downlink channel state information corresponding to the corresponding downlink frequency point information of the terminal. In this implementation, the first module obtains the downlink channel state information corresponding to the corresponding downlink frequency point information at the next moment based on the preprocessed uplink channel state information at the previous moment and at least part of the downlink channel state information corresponding to the uplink channel state information fed back by the terminal; that is, obtains the predicted downlink channel state information.

[0153] In another implementation manner in the embodiment of the present application, optionally, the processing module includes a first module and a second module cascaded after a signal output end of the first module.

[0154] Among them, Figure 4 As shown, the signal output end of the first module is connected to the signal input end of the second module.

[0155] In this implementation, the first input signal is input into the processing module, including:

[0156] The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0157] In this implementation manner, the signal input end of the first module includes: the first uplink channel state information or the first uplink channel state information and the first downlink channel state information.

[0158] Optionally, the signal input end of the first module includes first uplink channel state information. Specifically, the base station estimates the uplink channel state information according to the uplink reference signal (first signal), that is, obtains the first uplink channel state information.

[0159] The first module performs data processing on the first uplink channel state information and outputs initial downlink channel state information.

[0160] In another embodiment, the signal input end of the first module includes first uplink channel state information and at least part of the downlink channel state information corresponding to the first uplink channel state information fed back by the terminal, that is, includes the first uplink channel state information and the first downlink channel state information.

[0161] The first module performs data processing on the first uplink channel state information and the first downlink channel state information, and outputs initial downlink channel state information.

[0162] Specifically, the signal output end of the first module includes: initial downlink channel state information.

[0163] In the embodiment of the present application, optionally, the signal input end of the second module includes: the initial downlink channel state information output by the first module and the uplink and downlink frequency information of the terminal;

[0164] The signal output end of the second module includes: predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency point information.

[0165] With this implementation, the first module outputs initial downlink channel state information, and outputs predicted initial downlink channel state information without considering the impact of different uplink and downlink frequencies; then, the second module outputs predicted downlink channel state information of the terminal based on the initial downlink channel state information output by the first module and the uplink and downlink frequency information of the terminal, so that the obtained predicted downlink channel state information represents the downlink frequency information of the terminal, that is, the downlink channel state information corresponding to the terminal and the corresponding downlink frequency information is obtained.

[0166] Optionally, in this embodiment, the machine learning model (or neural network model) adopted by the second module includes but is not limited to only a deep neural network (DNN) model.

[0167] In an embodiment of the present application, optionally, the at least two modules in the processing module respectively include a plurality of data processing models;

[0168] The method further comprises:

[0169] The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

[0170] Optionally, the type of predicted downlink channel state information required to be obtained includes but is not limited to only one or more of channel quality indication (Channel Quality Indicator, CQI), precoding matrix indication (Precoding Matrix Indicator, PMI), rank indication (Rank Indicator, RI), layer indication (Layer Indicator, LI), layer 1 reference signal received power (Layer-1 Reference Signal Received Power, L1-RSRP), etc.

[0171] Optionally, the type of the predicted downlink channel state information may also include multiple combinations of the above-mentioned types of downlink channel state information.

[0172] In the method described in the embodiment of the present application, corresponding to different first input signals, different second input signals, and different types of predicted downlink channel state information, the first module and the second module may respectively include corresponding data processing models. Specifically, according to the type of predicted downlink channel state information to be obtained and the difference between the first input signal and the second input signal, the corresponding first module and the second module respectively select corresponding data processing models.

[0173] By adopting the method described in the embodiment of the present application, optionally, when the duplex mode of the first signal obtained by the base station is TDD, the second module cascaded with the first module in the processing module is kept in a deactivated state, and the predicted downlink channel state information of the terminal is obtained by inputting the second input signal to the first module; when the duplex mode of the first signal obtained by the base station is FDD, the second module cascaded with the first module in the processing module is activated, and the predicted downlink channel state information of the terminal is obtained by inputting the first input signal to the processing module.

[0174] Therefore, after acquiring the first signal sent by the terminal, the base station needs to determine whether to activate the second module cascaded with the first module according to the type of the first signal.

[0175] Optionally, it can be determined whether the duplex mode of the first signal is FDD or TDD based on whether the first signal includes uplink and downlink frequency information. If the first signal includes uplink and downlink frequency information, it means that the duplex mode of the first signal is FDD; if the first signal does not include uplink and downlink frequency information, it means that the duplex mode of the first signal is TDD.

[0176] By adopting the channel state information acquisition method described in the embodiment of the present application, the downlink channel state information of the terminal can be obtained according to the type of duplex mode in which the terminal sends the signal, and unified TDD and FDD channel prediction can be achieved, thereby reducing or avoiding the CSI feedback overhead of the FDD system and improving the accuracy of the TDD system in acquiring downlink CSI.

[0177] On the other hand, the embodiment of the present application also provides a method for generating a processing module for channel state information prediction, such as Figure 5 As shown, the method includes:

[0178] S510, obtaining a first data set; the first data set includes first channel state information of the terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0179] S520, training a first neural network model according to the second channel state information at multiple historical times to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0180] S530, keeping the parameters of the first data processing model unchanged, training at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtaining the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0181] The generation method described in this embodiment is adopted, based on a deep learning prediction method, and utilizing the channel reciprocity characteristics shared by TDD and FDD systems. First, TDD channel feedback prediction is mainly used to train a machine learning model (or neural network model) as the first module; the parameters of the first module are kept unchanged, and a neural network model of multiple modules including the first module is trained to obtain a data processing model of multiple modules in the processing module including the first module, which is suitable for FDD channel prediction to achieve unified TDD and FDD channel prediction, thereby reducing or avoiding the CSI feedback overhead of the FDD system and improving the downlink CSI accuracy obtained by the TDD system.

[0182] It should be noted that the network device that executes the method for generating the processing module for channel state information prediction described in the embodiment of the present application can be the same device as the network device that executes the channel state information acquisition method, such as both are base stations, and of course they can also be different network devices.

[0183] Optionally, in step S510, the first data set obtained includes:

[0184] First channel state information of the terminal at multiple historical times;

[0185] Second channel state information of the terminal at multiple historical times;

[0186] The uplink and downlink frequency information corresponding to the terminal at the historical time.

[0187] Optionally, the first channel state information includes:

[0188] first uplink channel state information; wherein the first uplink channel state information is determined according to the first signal whose duplex mode is FDD; or,

[0189] The first channel state information includes the first uplink channel state information and first downlink channel state information; wherein the first downlink channel state information includes at least part of the downlink channel state information corresponding to the first uplink channel state information fed back by the terminal.

[0190] Optionally, the second channel state information includes:

[0191] second uplink channel state information; wherein the second uplink channel state information is determined according to the first signal whose duplex mode is TDD; or,

[0192] The second channel state information includes the second uplink channel state information and the second downlink channel state information;

[0193] The second downlink channel state information includes at least a portion of downlink channel state information corresponding to the second uplink channel state information fed back by the terminal.

[0194] In the embodiment of the present application, optionally, the first uplink channel state information is obtained by the base station according to a signal in which the duplex mode is FDD. Specifically, the uplink channel state information is estimated and stored respectively according to the signal in which the duplex mode is FDD obtained within a preset time period, so as to obtain the first uplink channel state information;

[0195] The second uplink channel state information is obtained by the base station according to the TDD duplex mode signal. Specifically, the uplink channel state information is estimated and stored according to the TDD duplex mode signals obtained within a preset time period to obtain the second uplink channel state information.

[0196] Optionally, the first uplink channel state information obtained according to the multiple signals and the second uplink channel state information obtained according to the multiple signals constitute an uplink channel state information input data set.

[0197] In the embodiment of the present application, when the duplex mode of the signal obtained by the base station is FDD, optionally, the base station also obtains corresponding uplink and downlink frequency information, wherein the uplink and downlink frequency information is associated with the first uplink channel state information in the first data set. Specifically, for the base station, the uplink and downlink frequency information is known, and it is only necessary to store the uplink and downlink frequency information of the signal whose duplex mode is FDD received within a preset time period, and associate the stored uplink and downlink frequency information with the corresponding first uplink channel state information.

[0198] Optionally, the uplink and downlink frequency information corresponding to the multiple signals obtained within the preset time period constitutes an uplink and downlink frequency information input data set.

[0199] In the embodiment of the present application, optionally, the first data set also includes one or more of the following:

[0200] The first downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information;

[0201] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0202] Specifically, the base station needs to store part or all of the downlink channel state information fed back by the terminal for model input within a preset time period, and associate the stored downlink channel state information with the first uplink channel state information in the first data set, or associate it with the second uplink channel state information to obtain the corresponding first downlink channel state information and / or second downlink channel state information.

[0203] Optionally, the first downlink channel state information and the second downlink channel state information obtained above constitute a downlink channel state information input data set.

[0204] In the embodiment of the present application, optionally, the first data set further includes a plurality of first label data and a plurality of second label data; wherein the first label data includes:

[0205] Downlink channel state information at a next historical time corresponding to the second uplink channel state in the second channel state information at multiple historical times;

[0206] The second tag data includes:

[0207] Downlink channel state information at a next historical time corresponding to a first uplink channel state in the first channel state information at multiple historical times.

[0208] Specifically, the first label data and the second label data are used as label signals for machine model training, that is, output signals.

[0209] It should be noted that, optionally, the first label data and the second label data may belong to the same data set as the first channel state information, the second channel state information, and the uplink and downlink frequency information. In another embodiment, optionally, the first label data and the second label data may also belong to different data sets as the first channel state information, the second channel state information, and the uplink and downlink frequency information.

[0210] In the present application, in order to more accurately predict the required channel state information CSI type, optionally, different network models are trained for each type of CSI, and the corresponding network model is selected for prediction according to the CSI type actually required to be predicted. When a combination of different CSI types needs to be fed back, it is only necessary to use the corresponding network models for prediction respectively, and then combine the results.

[0211] In addition, in the present application, optionally, the predicted CSI type may include a downlink channel matrix H, and corresponding to the downlink channel matrix H to be predicted, the corresponding first label data includes the downlink channel matrix H of TDD or FDD.

[0212] By adopting the method described in the embodiment of the present application, in the data collection and acquisition stage, based on the need to obtain multiple types of downlink channel state information CSI within a preset time period as the first label data or the second label data, training is performed separately to obtain network models corresponding to the multiple CSI types.

[0213] Optionally, the first label data and the second label data respectively include:

[0214] CQI label data set: The base station needs to store the CQI fed back by the terminal and associate the fed back CQI with the uplink CSI at the corresponding time in the uplink CSI input data set;

[0215] PMI label data set: The base station needs to store the PMI fed back by the terminal and associate the fed back PMI with the corresponding uplink CSI in the uplink CSI input data set;

[0216] RI label data set: The base station needs to store the RI fed back by the terminal and associate the fed back RI with the corresponding uplink CSI in the uplink CSI input data set;

[0217] LI label data set: The base station needs to store the LI fed back by the terminal and associate the fed back LI with the corresponding uplink CSI in the uplink CSI input data set;

[0218] L1-RSRP label data set: The base station needs to store the L1-RSRP fed back by the terminal and associate the fed back L1-RSRP with the corresponding uplink CSI in the uplink CSI input data set;

[0219] Downlink channel matrix H label data set: Since the actual downlink channel matrix H is rarely fed back in the actual communication system, when obtaining the downlink channel matrix H label data set, the base station needs to notify the corresponding terminal to upload the downlink channel matrix H estimated by the terminal when the terminal is relatively idle, and then the terminal sends the estimated downlink channel matrix H to the base station. The base station stores the received downlink channel matrix H and associates the stored downlink with the corresponding uplink CSI in the uplink CSI input data set.

[0220] The method for generating a processing module for channel state information prediction described in an embodiment of the present application performs model training based on the first data set obtained as mentioned above, and obtains a processing module for realizing unified TDD and FDD channel prediction, which includes two training stages: a training stage of the first model of the first module; and a training stage of the second model of the second module.

[0221] Among them, in the training stage of the first model of the first module, based on the collected second channel state information and the first label data, a corresponding neural network model is selected for training until the trained neural network model converges to obtain the first data processing model of the first module.

[0222] Optionally, depending on whether the second channel state information in the input data set of the first module includes part of the terminal feedback or downlink CSI, that is, depending on whether the second channel state information includes second downlink channel state information, and the type of downlink channel state information that needs to be predicted, different neural network model parameters need to be trained to obtain respectively corresponding different first data processing models. In this way, when the first data processing model is used to obtain the predicted downlink channel state information, the corresponding first data processing model can be selected for data processing according to different input signals and / or different types of downlink CSI that need to be predicted, so as to more accurately predict the downlink CSI.

[0223] According to the above, in an embodiment of the present application, the first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of the second uplink channel states in the first data set;

[0224] The method of training the first neural network model according to the second channel state information at multiple historical times to obtain the first model includes:

[0225] The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

[0226] Optionally, the second channel state information at multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model, including one or more of the following:

[0227] The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module;

[0228] The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

[0229] By adopting the above process, for the training of the first model of the first module, the corresponding first model can be selected for data processing according to different input signals and / or different types of downlink CSI to be predicted, so as to more accurately predict the downlink CSI.

[0230] In the training phase of the second model of the second module, the network parameters in the first module are kept unchanged, the neural network structure of the second module is selected, and the second channel state information and the second label data collected in the first data set are used. Figure 3 or Figure 4 structure, train the neural network model corresponding to the second module until the trained neural network model converges, and obtain the second data processing model of the second module.

[0231] Optionally, when the second module is located before the first module, the second module needs to train parameters of different machine learning models to obtain different second models according to whether the input signal includes at least part of the downlink CSI fed back by the terminal and the type of CSI to be predicted. In this way, when the second data processing model is used to obtain the predicted downlink channel state information, the corresponding second data processing model can be selected for data processing according to the different input signals and / or the different types of downlink CSI to be predicted, so as to more accurately predict the downlink CSI.

[0232] Therefore, according to the above, the first data set also includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of the first uplink channel states in the first data set;

[0233] Keeping the parameters of the first data processing model unchanged, training at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtaining the second data processing model of the second module, including:

[0234] Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

[0235] Optionally, one embodiment, such as Figure 3 As shown, the signal output end of the second module is connected to the signal input end of the first module;

[0236] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including:

[0237] The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

[0238] Another embodiment, optionally, as Figure 4 As shown, the signal output end of the first module is connected to the signal input end of the second module;

[0239] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including:

[0240] The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

[0241] In this implementation, optionally, the first channel state information at multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time are used as training data and input into the processing module, and the second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following:

[0242] The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module;

[0243] The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

[0244] By adopting the above process, for the training of the second model of the second module, the corresponding first model can be selected for data processing according to different input signals and / or different types of downlink CSI to be predicted, so as to more accurately predict the downlink CSI.

[0245] The following assumes that the duplex mode of the first input signal is FDD, combined with Figure 3 As shown, taking the example that the second module is located before the first module, the method for generating the processing module for channel state information prediction described in the embodiment of the present application is illustrated.

[0246] In this embodiment, combined with Figure 3 As shown, the input signal of the processing module including the first module and the second module includes first uplink channel state information and uplink and downlink frequency information, and the first uplink channel state information is the uplink channel state information estimated by the base station according to the signal with the duplex mode of FDD; wherein the first module is used to pre-process the first uplink channel state information according to the first uplink channel state information and the uplink and downlink frequency information, so that the pre-processed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information.

[0247] Combination Figure 3 As shown, the method of this embodiment is adopted, and the specific implementation process includes the following steps:

[0248] 1) Obtaining an input data set for a machine learning model; in this embodiment, the obtained input data set includes:

[0249] First uplink channel state information of the terminal at multiple historical times;

[0250] Second uplink channel state information of the terminal at multiple historical times;

[0251] The uplink and downlink frequency information corresponding to the terminal at the historical time.

[0252] Optionally, the first uplink channel state information and the second uplink channel state information constitute an uplink CSI input data set; and the uplink and downlink frequency point information constitute an uplink and downlink frequency point input data set.

[0253] 2) Obtain a labeled data set for the machine learning model; Optionally, the base station can notify the terminal to feedback all types of CSI, including CQI, PMI, RI, LI, L1-RSRP, etc., during a relatively idle time. The base station receives and stores the received CSI feedback, and associates the stored CSI with the corresponding uplink CSI to obtain the first label data.

[0254] 3) The machine learning model of the first module (the first neural network model) is selected as the LSTM structure, and the machine learning model of the second module (the second neural network model) is selected as the RNN structure.

[0255] 4) Based on the acquired uplink TDD CSI input data set and various CSI label data sets, training a first neural network model until the model converges to obtain a first data processing model of the first module.

[0256] 5) Keeping the parameters of the first data processing model of the first module unchanged, training the second neural network model based on the uplink FDD CSI input data set, the uplink and downlink frequency information, and various CSI label data sets until the model converges, and obtaining the second data processing model of the second module;

[0257] 6) Model deduction stage:

[0258] When the input signal includes uplink TDD CSI (second uplink channel state information), the second module is deactivated, and only the first module is used to predict the downlink CSI based on the uplink TDD CSI;

[0259] When the input signal includes uplink FDD CSI (first uplink channel state information) and uplink and downlink frequency information, the second module is activated and combined with the first module to predict the downlink CSI based on the uplink FDD CSI and the uplink and downlink frequency information.

[0260] By adopting the generation method described in the embodiment of the present application, based on a deep learning prediction method and utilizing the channel reciprocity characteristics common to TDD and FDD systems, a data processing model for unified TDD and FDD channel prediction is generated, which can reduce or avoid the CSI feedback overhead of the FDD system, and improve the accuracy of the TDD system in obtaining downlink CSI, thereby reducing hardware costs and effectively reducing the burden of model training and computational complexity.

[0261] One embodiment of the present application also provides a network device, such as Figure 6 As shown, it includes a memory 610, a transceiver 620, and a processor 630:

[0262] The memory 610 is used to store computer programs; the transceiver 620 is used to send and receive data under the control of the processor 630; the processor 630 is used to read the computer program in the memory 610 and perform the following operations:

[0263] Acquire a first signal sent by a terminal;

[0264] In the case where the duplex mode of the first signal is frequency division duplex FDD, the first input signal is input into the processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency point information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0265] When the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, and the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0266] Optionally, in the network device, the first input signal further includes:

[0267] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0268] Optionally, in the network device, the second input signal further includes:

[0269] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0270] Optionally, in the network device, the processing module further comprises a second module; a signal output terminal of the second module is connected to a signal input terminal of the first module;

[0271] The processor 630 reads the computer program in the memory 610 and inputs the first input signal into the processing module, including:

[0272] The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0273] Optionally, in the network device, the processing module further comprises a second module; the signal output end of the first module is connected to the signal input end of the second module;

[0274] The processor 630 reads the computer program in the memory 610 and inputs the first input signal into the processing module, including:

[0275] The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0276] Optionally, in the network device, the at least two modules in the processing module respectively include a plurality of data processing models;

[0277] The processor 630 is further configured to read the computer program in the memory 610 and perform the following operations:

[0278] The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

[0279] Among them, Figure 6In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 630 and various circuits of memory represented by memory 610 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 620 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 630 is responsible for managing the bus architecture and general processing, and the memory 610 may store data used by the processor 630 when performing operations.

[0280] The processor 630 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0281] One embodiment of the present application also provides a network device, such as Figure 7 As shown, it includes a memory 710, a transceiver 720, and a processor 730:

[0282] The memory 710 is used to store computer programs; the transceiver 720 is used to send and receive data under the control of the processor 730; the processor 730 is used to read the computer program in the memory 710 and perform the following operations:

[0283] Acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0284] According to the second channel state information at multiple historical times, the first neural network model is trained to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0285] Keeping the parameters of the first data processing model unchanged, train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0286] Optionally, in the network device, the first channel state information further includes:

[0287] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0288] Optionally, in the network device, the second channel state information further includes:

[0289] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0290] Optionally, the network device, wherein the first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of the second uplink channel states in the first data set;

[0291] The processor 730 reads the computer program in the memory 710, trains the first neural network model according to the second channel state information at multiple historical times, and obtains the first data processing model, including:

[0292] The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

[0293] Optionally, the network device, wherein the first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of first uplink channel states in the first data set;

[0294] The processor 730 reads the computer program in the memory 710, keeps the parameters of the first data processing model unchanged, trains at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtains the second data processing model of the second module, including:

[0295] Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

[0296] Optionally, in the network device, the signal output terminal of the second module is connected to the signal input terminal of the first module;

[0297] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor 730 reads the computer program in the memory 710, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time as training data, and inputs them into the processing module, including:

[0298] The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

[0299] Optionally, in the network device, the signal output terminal of the first module is connected to the signal input terminal of the second module;

[0300] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor 730 reads the computer program in the memory 710, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time as training data, and inputs them into the processing module, including:

[0301] The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

[0302] Optionally, the network device, wherein the processor 730 reads the computer program in the memory 710, uses the second channel state information of multiple historical times as training data, inputs it into the first module, uses the first label data as a training label, trains the first neural network model, and obtains the first data processing model, including one or more of the following:

[0303] The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module;

[0304] The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

[0305] Optionally, in the network device, the processor 730 reads the computer program in the memory 710,

[0306] The first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time are used as training data and input into the processing module. The second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following:

[0307] The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module;

[0308] The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

[0309] Among them, Figure 7 In the embodiment, the bus architecture may include any number of interconnected buses and bridges, specifically one or more processors represented by processor 730 and various circuits of memory represented by memory 710 are linked together. The bus architecture may also link together various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. The bus interface provides an interface. The transceiver 720 may be a plurality of components, namely, a transmitter and a receiver, providing a unit for communicating with various other devices on a transmission medium, which transmission medium may include a wireless channel, a wired channel, an optical cable, and other transmission media. The processor 730 is responsible for managing the bus architecture and general processing, and the memory 710 may store data used by the processor 730 when performing operations.

[0310] The processor 730 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or a complex programmable logic device (CPLD). The processor may also adopt a multi-core architecture.

[0311] One embodiment of the present application also provides a channel state information acquisition device, such as Figure 8 As shown, the device comprises:

[0312] A signal acquisition unit 810, configured to acquire a first signal sent by a terminal;

[0313] The first input unit 820 is used to input the first input signal into the processing module when the duplex mode of the first signal is frequency division duplex FDD, and the processing module performs data processing according to the first input signal and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes the uplink and downlink frequency point information of the terminal and the first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other;

[0314] The second input unit 830 is used to input the second input signal into the first module in the processing module when the duplex mode of the first signal is time division duplex TDD, and the first module performs data processing according to the second input signal and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

[0315] Optionally, in the device, the first input signal further includes:

[0316] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0317] Optionally, in the device, the second input signal further includes:

[0318] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0319] Optionally, in the device, the processing module further comprises a second module; a signal output terminal of the second module is connected to a signal input terminal of the first module;

[0320] The first input unit 820 inputs the first input signal into the processing module, including:

[0321] The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0322] Optionally, in the device, the processing module further comprises a second module; the signal output end of the first module is connected to the signal input end of the second module;

[0323] The first input unit 820 inputs the first input signal into the processing module, including:

[0324] The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

[0325] Optionally, in the device, the at least two modules in the processing module respectively include a plurality of data processing models;

[0326] The device also includes:

[0327] The model determination unit 840 is used to determine the data processing model that needs to be called by each module according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

[0328] It should be noted here that the above-mentioned device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0329] One embodiment of the present application also provides a device for generating a processing module for channel state information prediction, such as Fig. 9 As shown, the device comprises:

[0330] The data acquisition unit 910 is used to acquire a first data set; the first data set includes first channel state information of the terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD;

[0331] A first training unit 920 is used to train a first neural network model according to the second channel state information at multiple historical times to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment;

[0332] The second training unit 930 is used to keep the parameters of the first data processing model unchanged, and train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times to obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

[0333] Optionally, in the device, the first channel state information further includes:

[0334] The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

[0335] Optionally, in the device, the second channel state information further includes:

[0336] The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

[0337] Optionally, in the device, the first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of second uplink channel states in the first data set;

[0338] The first training unit 920 trains the first neural network model according to the second channel state information at multiple historical times to obtain the first model, including:

[0339] The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

[0340] Optionally, in the device, the first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of first uplink channel states in the first data set;

[0341] The second training unit 930 keeps the parameters of the first data processing model unchanged, and trains at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times to obtain the second data processing model of the second module, including:

[0342] Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

[0343] Optionally, in the device, the signal output terminal of the second module is connected to the signal input terminal of the first module;

[0344] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the second training unit 930 uses the first channel state information of multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time as training data, and inputs them into the processing module, including:

[0345] The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

[0346] Optionally, in the device, the signal output terminal of the first module is connected to the signal input terminal of the second module;

[0347] Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the second training unit 930 uses the first channel state information of multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time as training data, and inputs them into the processing module, including:

[0348] The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

[0349] Optionally, the apparatus, wherein the first training unit 920 uses the second channel state information of multiple historical times as training data, inputs it into the first module, uses the first label data as a training label, trains the first neural network model, and obtains the first data processing model, including one or more of the following:

[0350] The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module;

[0351] The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

[0352] Optionally, in the device, the second training unit 930 uses the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, inputs them into the processing module, and uses the second label data as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following:

[0353] The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module;

[0354] The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

[0355] It should be noted here that the above-mentioned device provided in the embodiment of the present application can implement all the method steps implemented in the above-mentioned method embodiment, and can achieve the same technical effect. The parts and beneficial effects of this embodiment that are the same as those in the method embodiment will not be described in detail here.

[0356] It should be noted that the division of units in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional unit in each embodiment of the present application may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0357] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a processor-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0358] One embodiment of the present application further provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, and the computer program is used to enable the processor to execute any of the methods described above.

[0359] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.

[0360] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0361] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer executable instructions. These computer executable instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0362] These processor executable instructions may also be stored in a processor readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the processor readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0363] These processor-executable instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0364] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for acquiring channel state information, characterized in that: The method comprises: Acquire a first signal sent by a terminal; In the case where the duplex mode of the first signal is frequency division duplex FDD, the first input signal is input into the processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency point information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other; When the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, and the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

2. The method according to claim 1, characterized in that The first input signal also includes: The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

3. The method according to claim 1, characterized in that The second input signal also includes: The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

4. The method according to claim 2, characterized in that: The processing module further includes a second module; a signal output end of the second module is connected to a signal input end of the first module; The step of inputting the first input signal into the processing module includes: The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

5. The method according to claim 2, characterized in that: The processing module further includes a second module; the signal output end of the first module is connected to the signal input end of the second module; The step of inputting the first input signal into the processing module includes: The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

6. The method according to claim 1, characterized in that The at least two modules in the processing module respectively include a plurality of data processing models; The method further comprises: The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

7. A method for generating a processing module for channel state information prediction, characterized in that: The method comprises: Acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD; According to the second channel state information at multiple historical times, the first neural network model is trained to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment; Keeping the parameters of the first data processing model unchanged, train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

8. The method according to claim 7, characterized in that The first channel state information further includes: The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

9. The method according to claim 7, characterized in that: The second channel state information further includes: The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

10. The method according to any one of claims 7 to 9, characterized in that: The first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of second uplink channel states in the first data set; The method of training the first neural network model according to the second channel state information at multiple historical times to obtain the first model includes: The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

11. The method according to any one of claims 7 to 9, characterized in that: The first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of first uplink channel states in the first data set; Keeping the parameters of the first data processing model unchanged, training at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtaining the second data processing model of the second module, including: Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

12. The method according to claim 11, characterized in that The signal output terminal of the second module is connected to the signal input terminal of the first module; Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including: The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

13. The method according to claim 11, characterized in that The signal output terminal of the first module is connected to the signal input terminal of the second module; Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times are input as training data into the processing module, including: The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

14. The method according to claim 10, characterized in that The second channel state information at multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model, including one or more of the following: The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module; The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

15. The method according to claim 11, characterized in that The first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time are used as training data and input into the processing module. The second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following: The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module; The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

16. A network device, characterized in that: Including memory, transceiver, processor: A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Acquire a first signal sent by a terminal; In the case where the duplex mode of the first signal is frequency division duplex FDD, the first input signal is input into the processing module, the processing module performs data processing according to the first input signal, and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency point information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other; When the duplex mode of the first signal is time division duplex TDD, the second input signal is input into the first module in the processing module, and the first module performs data processing according to the second input signal, and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

17. The network device according to claim 16, characterized in that: The first input signal also includes: The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

18. The network device according to claim 16, characterized in that: The second input signal also includes: The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

19. The network device according to claim 17, characterized in that: The processing module further includes a second module; a signal output end of the second module is connected to a signal input end of the first module; The processor reads the computer program in the memory and inputs the first input signal into the processing module, including: The uplink and downlink frequency information and the first uplink channel state information are input into the second module; wherein the second module preprocesses the first uplink channel state information according to the uplink and downlink frequency information, and outputs preprocessed uplink channel state information corresponding to the first uplink channel state information, and the preprocessed uplink channel state information can characterize the channel characteristics of the corresponding downlink frequency information; the preprocessed uplink channel state information, or the preprocessed uplink channel state information and the first downlink channel state information are input into the first module, and the first module performs data processing and outputs the predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

20. The network device according to claim 17, characterized in that: The processing module further includes a second module; the signal output end of the first module is connected to the signal input end of the second module; The processor reads the computer program in the memory and inputs the first input signal into the processing module, including: The first uplink channel state information or the first uplink channel state information and the first downlink channel state information are input into the first module; wherein the first module performs data processing and outputs initial downlink channel state information; the initial downlink channel state information and the uplink and downlink frequency information are input into the second module, and the second module performs data processing according to the initial downlink channel state information and the uplink and downlink frequency information, and outputs predicted downlink channel state information corresponding to the terminal and the corresponding downlink frequency information.

21. The network device according to claim 17, characterized in that: The at least two modules in the processing module respectively include a plurality of data processing models; The processor is further configured to read the computer program in the memory and perform the following operations: The data processing model that needs to be called by each module is determined according to one or more of the first input signal, the second input signal and the type of predicted downlink channel state information that needs to be obtained.

22. A network device, characterized in that: Including memory, transceiver, processor: A memory for storing a computer program; a transceiver for transmitting and receiving data under the control of the processor; and a processor for reading the computer program in the memory and performing the following operations: Acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD; According to the second channel state information at multiple historical times, the first neural network model is trained to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment; Keeping the parameters of the first data processing model unchanged, train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times, and obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

23. The network device according to claim 22, characterized in that: The first channel state information further includes: The first downlink channel state information includes at least a portion of downlink channel state information fed back by the terminal and corresponding to the first uplink channel state information.

24. The network device according to claim 22, characterized in that: The second channel state information further includes: The second downlink channel state information includes at least a portion of the downlink channel state information fed back by the terminal and corresponding to the second uplink channel state information.

25. The network device according to any one of claims 22 to 24, characterized in that: The first data set further includes a plurality of first tag data; the first tag data includes downlink channel state information at the next historical time corresponding to the plurality of second uplink channel states in the first data set; The processor reads the computer program in the memory, trains the first neural network model according to the second channel state information at multiple historical times, and obtains the first data processing model, including: The second channel state information of multiple historical times is used as training data and input into the first module, and the first label data is used as a training label to train the first neural network model to obtain the first data processing model.

26. The network device according to any one of claims 22 to 24, characterized in that: The first data set further includes a plurality of second tag data; the second tag data includes downlink channel state information at the next historical time corresponding to the plurality of first uplink channel states in the first data set; The processor reads the computer program in the memory, keeps the parameters of the first data processing model unchanged, trains at least two neural network models cascaded in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information at multiple historical times, and obtains the second data processing model of the second module, including: Keep the parameters of the first data processing model unchanged, use the first channel state information of multiple historical times and the uplink and downlink frequency information corresponding to the terminal at the historical time as training data, input them into the processing module, use the second label data as training labels, train at least two cascaded neural network models, and obtain the second data processing model of the second module.

27. The network device according to claim 26, characterized in that: The signal output terminal of the second module is connected to the signal input terminal of the first module; Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor reads the computer program in the memory, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times as training data, and inputs them into the processing module, including: The uplink and downlink frequency point information corresponding to the terminal at the historical time and the first uplink channel state information are input into the second module as training data, and the first downlink channel state information is input into the first module as training data.

28. The network device according to claim 26, characterized in that: The signal output terminal of the first module is connected to the signal input terminal of the second module; Wherein, when the first channel state information includes first uplink channel state information and first downlink channel state information, the processor reads the computer program in the memory, and uses the first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical times as training data, and inputs them into the processing module, including: The first uplink channel state information and the first downlink channel state information are input into the first module as training data, and the uplink and downlink frequency point information corresponding to the terminal at the historical time is input into the second module as training data.

29. The network device according to claim 25, characterized in that: The processor reads the computer program in the memory, uses the second channel state information of multiple historical times as training data, inputs it into the first module, uses the first label data as a training label, trains the first neural network model, and obtains the first data processing model, including one or more of the following: The second uplink channel state information in the second channel state information at multiple historical times is used as training data, input into the first module, and different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively, and the first neural network model is trained to obtain multiple first data processing models of the first module; The second uplink channel state information and the corresponding second downlink channel state information in the second channel state information at multiple historical times are used as training data and input into the first module. Different types of first label data corresponding to the second channel state information in the training data are used as training labels respectively to train the first neural network model to obtain multiple first data processing models of the first module.

30. The network device according to claim 26, characterized in that The processor reads the computer program in the memory, The first channel state information at multiple historical times and the uplink and downlink frequency point information corresponding to the terminal at the historical time are used as training data and input into the processing module. The second label data is used as a training label to train at least two cascaded neural network models to obtain the second data processing model of the second module, including one or more of the following: The first uplink channel state information in the first channel state information at multiple historical times is used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are respectively used as training data, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module; The first uplink channel state information and the corresponding first downlink channel state information in the first channel state information at multiple historical times are used as training data and input into the processing module, wherein different types of second label data corresponding to the first channel state information in the training data are used as training data respectively, and at least two cascaded neural network models are trained respectively to obtain multiple second data processing models of the second module.

31. A channel state information acquisition device, characterized in that: The device comprises: A signal acquisition unit, configured to acquire a first signal sent by a terminal; A first input unit is used to input the first input signal into a processing module when the duplex mode of the first signal is frequency division duplex FDD, and the processing module performs data processing according to the first input signal and outputs the predicted downlink channel state information of the terminal; wherein the first input signal includes uplink and downlink frequency information of the terminal and first uplink channel state information determined according to the first signal whose duplex mode is FDD; the processing module includes at least two modules cascaded to each other; A second input unit is used to input the second input signal into the first module in the processing module when the duplex mode of the first signal is time division duplex TDD, and the first module performs data processing according to the second input signal and outputs the predicted downlink channel state information of the terminal; wherein the second input signal includes the second uplink channel state information determined according to the first signal whose duplex mode is TDD.

32. A device for generating a processing module for channel state information prediction, characterized in that: The device comprises: A data acquisition unit, configured to acquire a first data set; the first data set includes first channel state information of a terminal at multiple historical times, second channel state information, and uplink and downlink frequency point information corresponding to the terminal at the historical times; wherein the first channel state information includes first uplink channel state information, and the first uplink channel state information is determined according to a first signal whose duplex mode is frequency division duplex FDD; the second channel state information includes second uplink channel state information, and the second uplink channel state information is determined according to a first signal whose duplex mode is time division duplex TDD; A first training unit is used to train a first neural network model according to the second channel state information at multiple historical times to obtain a first data processing model of the first module; the first data processing model is used to output the predicted downlink channel state information of the terminal at the next moment according to the input second input signal; the second input signal includes the second channel state information of the terminal at the previous moment; The second training unit is used to keep the parameters of the first data processing model unchanged, and train at least two cascaded neural network models in the processing module according to the first channel state information and the corresponding uplink and downlink frequency information of multiple historical times to obtain the second data processing model of the second module in the processing module; wherein, the trained at least two cascaded neural network models include the first neural network model; the processing module includes the first module; the processing module is used to output the predicted downlink channel state information of the terminal at the next moment according to the input first input signal; the first input signal includes the first channel state information of the terminal at the previous moment and the corresponding uplink and downlink frequency information.

33. A processor-readable storage medium, characterized in that: The processor-readable storage medium stores a computer program, and the computer program is used to cause the processor to execute the method according to any one of claims 1 to 6, or to cause the processor to execute the method according to any one of claims 7 to 15.

Images