Information transmission method and device

Abstract

The invention provides an information transmission method and device used for determining models adopted by a receiver and a transmitter, and relates to the technical field of wireless communication. In the method, a terminal device receives first information from a network device, and the first information indicates a first model for local data processing and a second model for global data processing. The terminal device may also receive an output signal from the network device. It can be understood that the output signal is data processed by the network device through the first model and the second model. And the terminal equipment processes the output signal based on the first model and the second model to obtain corresponding data. Based on the scheme, the network equipment can determine the first model and the second model for data processing and indicate the first model and the second model to the terminal equipment, so that the terminal equipment can determine to adopt the first model and the second model to process signals according to the indication of the network equipment, and the performance of a receiver is improved.

Description

Information transmission method and device Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to an information transmission method and device. Background Art

[0002] Artificial intelligence technology has been very successfully applied in the fields of image processing and natural language processing. Its increasing maturity will significantly drive the evolution of mobile communication network technology. Currently, artificial intelligence technology is primarily applied to the network and physical layers.

[0003] Artificial intelligence technologies for the physical layer mostly replace physical layer modules, such as signal processing modules like coding, modulation, MIMO precoding, and beamforming. Their primary advantages are reduced computational latency and improved algorithm performance. However, the independent optimization algorithms for each physical layer module are already close to their upper performance limits, so simply replacing these modules offers limited gains.

[0004] To achieve better receiver performance, multiple modules can be jointly optimized to design an artificial intelligence (AI) model, which can then be used to process received signals. However, in wireless networks, varying channel conditions can significantly impact the performance of the AI ​​model, and thus the receiver. Consequently, it can be difficult for terminals to determine which AI model to use for signal reception.

[0005] Summary of the Invention

[0006] The present application provides an information transmission method and apparatus for determining the models adopted by a receiver and a transmitter.

[0007] In a first aspect, a method for information transmission is provided, which can be performed by a second device. The second device can be a terminal device, a network device, or a chip / chip system. In this method, the second device receives first information from the first device, where the first information indicates a first model for local data processing and a second model for global data processing. The second device can also receive an output signal from the first device. It is understood that the output signal is a signal obtained by the first device processing the first data using the first model and the second model. The second device processes the output signal based on the first model and the second model to obtain second data.

[0008] Based on this scheme, the first device can determine the first model and the second model for data processing, and indicate the first model and the second model to the second device, so that the second device can determine to use the first model and the second model to process the signal according to the indication of the first device, thereby improving the performance of the receiver.

[0009] In one possible implementation, a second device receives a downlink reference signal from a first device and measures the downlink reference signal to obtain measurement information of the downlink reference signal. The measurement information of the downlink reference signal is used to determine the first model and the second model. The second device sends the measurement information of the downlink reference signal to the first device.

[0010] Based on this solution, the second device can measure the downlink reference signal, determine the downlink reference signal measurement information, and report the downlink reference signal measurement information to the first device. In this way, the first device can determine the channel state based on the downlink reference signal measurement information, and thus determine whether to select the first model or the second model for data processing.

[0011] In one possible implementation, the second device sends an uplink reference signal to the first device, and the uplink reference signal is used by the first device to measure the uplink reference signal to obtain measurement information of the uplink reference signal, wherein the measurement information of the uplink reference signal is used to determine the first model and the second model.

[0012] Based on this solution, the second device can send an uplink reference signal to the first device for the first device to measure. In this way, the first device can determine the channel state and thus determine whether to select the first model or the second model for data processing.

[0013] In one possible implementation, the first information is used to indicate the index of one or more first models, and the first information is also used to indicate the index of the second model. Based on this solution, the first information can indicate the index of the first model and the index of the second model, occupying fewer bits.

[0014] In one possible implementation, the first information includes an antenna port field and a channel characteristic indication field. The antenna port field and the channel characteristic indication field can be used to indicate the first model and the second model. It will be appreciated that in related art, the antenna port field is used to indicate the antenna port number, and the channel characteristic indication field is used to indicate channel characteristics. Based on this solution, the first information can reuse existing fields.

[0015] In one possible implementation, the second device inputs the output signal into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors. The second device processes the N1 high-dimensional vectors using the first model to obtain N2 high-dimensional vectors. The first model is pre-trained for processing the N1 high-dimensional vectors. The second device processes the N2 high-dimensional vectors using the second model to obtain N3 high-dimensional vectors. The second model is pre-trained for processing the N2 high-dimensional vectors. The second device inputs the N3 high-dimensional vectors into the output layer of the neural network to obtain second data.

[0016] Based on the above solution, the first and second models optimize the N1 high-dimensional vectors obtained from the input signal. Since the first and second models optimize the high-dimensional vectors independently of their dimensions, the complexity of signal processing can be reduced. Furthermore, the number of high-dimensional vectors input to the first and second models can be varied, thereby achieving scalability for signals of varying dimensions.

[0017] In a possible implementation, the second device may input the output signals corresponding to the M data streams into the input layer for dimensionality increase processing to obtain N1 high-dimensional vectors, where one high-dimensional vector corresponds to one data stream.

[0018] Based on this solution, by processing local data through the first model, the neural network provided in this application can be applied to a multiple input multiple output (MIMO) scenario to realize the processing of data streams.

[0019] In one possible implementation, the output signal includes data received by the communication device, and the second data includes a log-likelihood ratio. Based on the above solution, when the output signal includes data received by the second device, and the second data includes the log-likelihood ratio of the data, the second device may decode the received data based on the log-likelihood ratio.

[0020] In one possible implementation, the first information further indicates whether the output signal includes a demodulation reference signal (DMRS). Based on this solution, the first information can indicate whether the output signal includes the DMRS, thereby allowing the second device to determine whether channel estimation and channel equalization based on the DMRS are required.

[0021] In a possible implementation, the first information further indicates the number of transmission streams of the output signal. Based on this solution, the first information can indicate the number of transmission streams, thereby allowing the second device to determine the number of data streams.

[0022] In a second aspect, an information transmission method is provided, which can be performed by a first device. The first device can be a network device, a terminal device, or a chip / chip system. In this method, the first device sends first information to a second device, where the first information indicates a first model for local data processing and a second model for global data processing. The first device processes the first data based on the first model and the second model to obtain an output signal. The first device then sends the output signal to the second device.

[0023] In one possible implementation, a first device sends a downlink reference signal to a second device. The first device receives measurement information of the downlink reference signal from the second device. The measurement information of the downlink reference signal is measured by the second device based on the downlink reference signal, and the measurement information of the downlink reference signal is used to determine the first model and the second model.

[0024] In a possible implementation, a first device receives an uplink reference signal from a second device, measures the uplink reference signal, and obtains measurement information of the uplink reference signal, which is used to determine the first model and the second model.

[0025] In a possible implementation, the first information is used to indicate an index of one or more first models, and the first information is also used to indicate an index of a second model.

[0026] In one possible implementation, the first information includes an antenna port field and a channel characteristic indication field. The antenna port field and the channel characteristic indication field may be used to indicate the first model and the second model. It is understood that in related art, the antenna port field is used to indicate the antenna port number, and the channel characteristic indication field is used to indicate channel characteristics.

[0027] In one possible implementation, a first device inputs the first data into the input layer of a neural network for dimensionality increase processing, thereby obtaining N1 high-dimensional vectors. A second device processes the N1 high-dimensional vectors using a first model, thereby obtaining N2 high-dimensional vectors. The first model is pre-trained for processing the N1 high-dimensional vectors. The second device processes the N2 high-dimensional vectors using a second model, thereby obtaining N3 high-dimensional vectors. The second model is pre-trained for processing the N2 high-dimensional vectors. The second device inputs the N3 high-dimensional vectors into the output layer of the neural network, thereby obtaining an output signal.

[0028] In a possible implementation, the first device may input the output signals corresponding to the M data streams into the input layer for dimensionality increase processing to obtain N1 high-dimensional vectors, where one high-dimensional vector corresponds to one data stream.

[0029] In one possible implementation, the first data includes modulation symbols, and the output signal includes transmission symbols. It is understood that the transmission symbols can generate a waveform through a waveform generation module. The first device can transmit a signal corresponding to the waveform via an antenna. Based on the above solution, when a neural network is used as a transmitter, the first data includes modulation symbols, and the output signal includes transmission symbols. The first device can achieve aliasing of the modulation symbols, allowing the transmitter to adapt to the channel.

[0030] In one possible implementation, the first data includes bits to be encoded, and the output signal includes encoded bits or transmitted symbols. Based on the above solution, when the first data includes bits to be encoded and the output signal includes encoded bits, the first device can encode the bits to be encoded based on the above neural network. When the first data includes coded bits and the output signal includes transmitted symbols, the first device can map the bits to be encoded to transmitted symbols based on the above neural network.

[0031] In a possible implementation manner, the first information further indicates whether the output signal includes a DMRS.

[0032] In a possible implementation, the first information further indicates the number of transmission streams of the output signal.

[0033] According to a third aspect, a communication device is provided, comprising a processing unit and a transceiver unit.

[0034] The transceiver unit is configured to receive first information from a first device, the first information indicating a first model for performing local data processing and a second model for performing global data processing. The transceiver unit is further configured to receive an output signal from the first device. The output signal is a signal obtained by the first device processing the first data using the first model and the second model. The processing unit is configured to process the output signal based on the first model and the second model to obtain second data.

[0035] In one possible implementation, the transceiver unit is further configured to receive a downlink reference signal from the first device. The processing unit is further configured to measure the downlink reference signal to obtain downlink reference signal measurement information. The downlink reference signal measurement information is used to determine the first model and the second model. The transceiver unit is further configured to send the downlink reference signal measurement information to the first device.

[0036] In one possible implementation, the transceiver unit is further configured to send an uplink reference signal to the first device, where the uplink reference signal is used by the first device to measure the uplink reference signal and obtain measurement information of the uplink reference signal, where the measurement information of the uplink reference signal is used to determine the first model and the second model.

[0037] In a possible implementation, the first information is used to indicate an index of one or more first models, and the first information is also used to indicate an index of a second model.

[0038] In one possible implementation, the first information includes an antenna port field and a channel characteristic indication field. The antenna port field and the channel characteristic indication field may be used to indicate the first model and the second model. It is understood that in related art, the antenna port field is used to indicate the antenna port number, and the channel characteristic indication field is used to indicate channel characteristics.

[0039] In one possible implementation, the processing unit is specifically configured to input the output signal into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors. The processing unit is specifically configured to process the N1 high-dimensional vectors using a first model to obtain N2 high-dimensional vectors. The first model is pre-trained for processing the N1 high-dimensional vectors. The processing unit is specifically configured to process the N2 high-dimensional vectors using a second model to obtain N3 high-dimensional vectors. The second model is pre-trained for processing the N2 high-dimensional vectors. The processing unit is specifically configured to input the N3 high-dimensional vectors into the output layer of the neural network to obtain second data.

[0040] In one possible implementation, the processing unit is specifically configured to input the output signals corresponding to the M data streams into the input layer for dimensionality increase processing to obtain N1 high-dimensional vectors, wherein one high-dimensional vector corresponds to one data stream.

[0041] In a possible implementation, the output signal includes a modulation symbol, a waveform signal, an air interface data symbol, an air interface pilot symbol, or a filtered air interface symbol.

[0042] In a possible implementation manner, the second data includes a channel estimation result, a channel equalization weight, noise power, a log-likelihood ratio, or a decoded bit.

[0043] In a possible implementation manner, the first information further indicates whether the output signal includes a DMRS.

[0044] In a possible implementation, the first information further indicates the number of transmission streams of the output signal.

[0045] According to a fourth aspect, a communication device is provided, comprising a processing unit and a transceiver unit.

[0046] The transceiver unit is configured to transmit first information to a second device, the first information indicating a first model for performing local data processing and a second model for performing global data processing. The processing unit is configured to process the first data based on the first model and the second model to obtain an output signal. The transceiver unit is further configured to transmit the output signal to the second device.

[0047] In one possible implementation, the transceiver unit is further configured to send a downlink reference signal to the second device. The transceiver unit is further configured to receive measurement information of the downlink reference signal from the second device. The measurement information of the downlink reference signal is measured by the second device based on the downlink reference signal, and the measurement information of the downlink reference signal is used to determine the first model and the second model.

[0048] In one possible implementation, the transceiver unit is further configured to receive an uplink reference signal from the second device. The processing unit is further configured to measure the uplink reference signal to obtain measurement information of the uplink reference signal, and the measurement information of the uplink reference signal is used to determine the first model and the second model.

[0049] In a possible implementation, the first information is used to indicate an index of one or more first models, and the first information is also used to indicate an index of a second model.

[0050] In one possible implementation, the first information includes an antenna port field and a channel characteristic indication field. The antenna port field and the channel characteristic indication field may be used to indicate the first model and the second model. It is understood that in related art, the antenna port field is used to indicate the antenna port number, and the channel characteristic indication field is used to indicate channel characteristics.

[0051] In one possible implementation, the processing unit is specifically configured to input the first data into the input layer of a neural network for dimensionality increase processing, thereby obtaining N1 high-dimensional vectors. The processing unit is specifically configured to process the N1 high-dimensional vectors using a first model, thereby obtaining N2 high-dimensional vectors. The first model is pre-trained for processing the N1 high-dimensional vectors. The processing unit is specifically configured to process the N2 high-dimensional vectors using a second model, thereby obtaining N3 high-dimensional vectors. The second model is pre-trained for processing the N2 high-dimensional vectors. The processing unit is specifically configured to input the N3 high-dimensional vectors into the output layer of the neural network, thereby obtaining an output signal.

[0052] In a possible implementation, the first device may input the output signals corresponding to the M data streams into the input layer for dimensionality increase processing to obtain N1 high-dimensional vectors, where one high-dimensional vector corresponds to one data stream.

[0053] In a possible implementation manner, the first data includes uncoded bits, coded bits, or modulation symbols.

[0054] In a possible implementation, the output signal includes a modulation symbol or a waveform signal.

[0055] In a possible implementation manner, the first information further indicates whether the output signal includes a DMRS.

[0056] In a possible implementation, the first information further indicates the number of transmission streams of the output signal.

[0057] In a fifth aspect, the present application provides a communication device comprising a processor coupled to a memory, the memory being configured to store computer programs or instructions, and the processor being configured to execute the computer programs or instructions to perform the respective implementation methods of the first and second aspects described above. The memory may be located within or outside the device. The number of processors may be one or more.

[0058] In a sixth aspect, the present application provides a communication device, comprising: a processor and an interface circuit, the interface circuit being used to communicate with other devices, and the processor being used to implement the various methods of the first and second aspects above.

[0059] In a seventh aspect, a communication device is provided, which includes a logic circuit and an input / output interface.

[0060] In an eighth aspect, the present application provides a communication system, comprising: a terminal device and a network device for executing each implementation method of the above-mentioned first and second aspects.

[0061] In a ninth aspect, the present application also provides a chip system, comprising: a processor for executing the various implementation methods of the first and second aspects above.

[0062] In a tenth aspect, the present application also provides a computer program product, including a computer program or instructions, which enables the implementation methods of the first and second aspects mentioned above to be executed when the computer program or instructions are run on a computer.

[0063] In the eleventh aspect, the present application also provides a computer-readable storage medium, in which a computer program or instruction is stored. When the instruction is executed on a computer, the implementation methods of the first and second aspects mentioned above are implemented.

[0064] The technical effects achieved in the above-mentioned second to eleventh aspects can refer to the technical effects in the first aspect and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] FIG1 is a schematic diagram of a communication system provided in an embodiment of the present application;

[0066] FIG2 is a schematic diagram of a signal processing method for a receiver based on a convolutional neural network structure;

[0067] FIG3 is an exemplary flow chart of a signal processing method provided in an embodiment of the present application;

[0068] FIG4 is a schematic diagram of a neural network provided in an embodiment of the present application;

[0069] FIG5A is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0070] FIG5B is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0071] FIG5C is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0072] FIG5D is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0073] FIG6 is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0074] FIG7A is a schematic diagram of a second model of a neural network provided in an embodiment of the present application;

[0075] FIG7B is a schematic diagram of a first matrix provided in an embodiment of the present application;

[0076] FIG7C is a schematic diagram of a first model provided in an embodiment of the present application;

[0077] FIG8 is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0078] FIG9 is a schematic diagram of a processing layer of a neural network according to an embodiment of the present application;

[0079] FIG10 is one of the exemplary flow charts of the information transmission method provided in an embodiment of the present application;

[0080] FIG11 is a schematic diagram of a DMRS provided in an embodiment of the present application;

[0081] FIG12 is one of the exemplary flow charts of the information transmission method provided in an embodiment of the present application;

[0082] FIG13 is a schematic diagram of a communication device according to an embodiment of the present application;

[0083] FIG14 is a schematic diagram of a communication device according to an embodiment of the present application;

[0084] FIG15 is a schematic diagram of a communication device according to an embodiment of the present application;

[0085] FIG16 is one of the schematic diagrams of a communication device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0086] The technical solutions of the embodiments of the present application can be applied to New Radio (NR) systems, Long Term Evolution (LTE) systems, LTE Frequency Division Duplex (FDD) systems, LTE Time Division Duplex (TDD) systems, Worldwide Interoperability for Microwave Access (WiMAX) communication systems, etc., without limitation herein.

[0087] FIG1 is a schematic diagram of the architecture of a communication system 1000 used in an embodiment of the present application. As shown in FIG1 , the communication system includes a wireless access network 100. The wireless access network 100 may include at least one network device (such as 110a and / or 110b in FIG1 ) and may also include at least one terminal device (such as at least one of 120a-120j in FIG1 ). The terminal device is connected to the access network device wirelessly, and the access network device is connected to the core network device wirelessly or by wire. Terminal devices and network devices may be connected to each other by wire or by wireless. FIG1 is only a schematic diagram, and the communication system may also include other network devices, such as wireless relay devices and wireless backhaul devices, which are not shown in FIG1 .

[0088] A network device is a network-side device with wireless transceiver capabilities. A network device can be a device in a radio access network (RAN) that provides wireless communication capabilities for terminal devices, and is called a RAN device. For example, a network device can be a base station, an evolved NodeB (eNodeB), a transmission reception point (TRP), a next-generation NodeB (gNB) in a fifth-generation (5G) mobile communication system, a next-generation base station in a sixth-generation (6G) mobile communication system, a base station in a future mobile communication system, or an access node in a WiFi system. It can also be a module or unit that performs some of the functions of a base station, such as a centralized unit (CU) or a distributed unit (DU). The CU here completes the functions of the radio resource control protocol and the packet data convergence protocol (PDCP) of the base station, and can also complete the function of the service data adaptation protocol (SDAP); the DU completes the functions of the radio link control layer and the medium access control (MAC) layer of the base station, and can also complete the functions of part of the physical layer or all of the physical layer. For the specific description of the above-mentioned various protocol layers, please refer to the relevant technical specifications of the 3rd Generation Partnership Project (3GPP). The network device can be a macro base station (such as 110a in Figure 1), a micro base station or an indoor station (such as 110b in Figure 1), or a relay node or a donor node, etc. The embodiments of the present application do not limit the specific technology and specific device form adopted by the network device. In the embodiments of the present application, the network device is taken as a base station as an example for explanation.

[0089] In another possible scenario, multiple RAN nodes collaborate to assist the terminal in achieving wireless access, and different RAN nodes implement part of the functions of the base station respectively. For example, the RAN node can be a CU, DU, CU-control plane (CP), CU-user plane (UP), or radio unit (RU). The CU and DU can be set separately, or they can be included in the same network element, such as a baseband unit (BBU). The RU can be included in a radio frequency device or radio frequency unit, such as a remote radio unit (RRU), an active antenna unit (AAU), or a remote radio head (RRH).

[0090] In different systems, CU (or CU-CP and CU-UP), DU or RU may also have different names, but those skilled in the art can understand their meanings. For example, in the ORAN system, CU may also be called O-CU (Open CU), DU may also be called O-DU, CU-CP may also be called O-CU-CP, CU-UP may also be called O-CU-UP, and RU may also be called O-RU. For the convenience of description, this application uses CU, CU-CP, CU-UP, DU and RU as examples for description. Any unit of CU (or CU-CP, CU-UP), DU and RU in this application can be implemented by a software module, a hardware module, or a combination of a software module and a hardware module.

[0091] A terminal device is a user-side device with wireless transceiver capabilities. A terminal device may also be referred to as user equipment (UE), a mobile station, a mobile terminal, etc. Terminal devices can be widely used in various scenarios, such as device-to-device (D2D), vehicle-to-everything (V2X) communication, machine-type communication (MTC), the Internet of Things (IoT), virtual reality, augmented reality, industrial control, autonomous driving, telemedicine, smart grids, smart furniture, smart offices, smart wearables, smart transportation, smart cities, etc. The terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver capabilities, a wearable device, a vehicle, a drone, a helicopter, an airplane, a ship, a robot, a robotic arm, a smart home appliance, etc. The embodiments of this application do not limit the specific technology and specific device form adopted by the terminal device. The embodiments of this application are described using the terminal device as an example.

[0092] Network devices and terminal devices can be fixed or mobile. They can be deployed on land, including indoors or outdoors, handheld or vehicle-mounted; on water; and in the air on aircraft, balloons, and satellites. The embodiments of this application do not limit the application scenarios of network devices and terminal devices.

[0093] The roles of network devices and terminal devices can be relative. For example, the helicopter or drone 120i in Figure 1 can be configured as a mobile network device. To terminal devices 120j accessing the wireless access network 100 via 120i, terminal device 120i is a network device. However, to network device 110a, 120i is a terminal device, meaning that communication between 110a and 120i occurs via a wireless air interface protocol. Of course, communication between 110a and 120i can also occur via an interface protocol between network devices. In this case, 120i is also a network device relative to 110a. Therefore, both network devices and terminal devices can be collectively referred to as communication devices. 110a and 110b in Figure 1 can be referred to as communication devices with network device functionality, while 120a-120j in Figure 1 can be referred to as communication devices with terminal device functionality.

[0094] In the embodiments of the present application, the functions of the network device may also be performed by a module (such as a chip) in the network device, or by a control subsystem that includes the network device functions. The control subsystem that includes the network device functions here may be a control center in the above-mentioned application scenarios such as smart grid, industrial control, smart transportation, and smart city. The functions of the terminal device may also be performed by a module (such as a chip or a modem) in the terminal device, or by a device that includes the terminal device functions.

[0095] In the embodiments of the present application, the device for implementing the function of the terminal device can be the terminal device; it can also be a device that can support the terminal device to implement the function, such as a chip system. The device can be installed in the terminal device or used in conjunction with the terminal device. In the embodiments of the present application, the chip system can be composed of a chip or include a chip and other discrete devices.

[0096] In the embodiments of the present application, the device for implementing the function of the network device can be the network device; or it can be a device that can support the network device to implement the function, such as a chip system. The device can be installed in the network device or used in conjunction with the network device.

[0097] In the embodiments of the present application, the terminal device may also have AI processing capabilities, and the network device may also have AI processing capabilities. For example, the terminal device may have neural network training capabilities, reasoning capabilities, etc. Optionally, the network device may also have neural network training capabilities, reasoning capabilities, etc.

[0098] Artificial intelligence technology has been very successfully applied in the fields of image processing and natural language processing. Its increasing maturity will significantly drive the evolution of mobile communication network technology. Currently, artificial intelligence technology is primarily applied to the network and physical layers.

[0099] Artificial intelligence (AI) technologies for the physical layer often replace physical layer modules, such as signal processing modules, using AI. The main advantages are reduced computational latency and improved algorithm performance. However, the independent optimization algorithms for each physical layer module are already close to the upper bound of performance, and simply replacing each module offers limited gains. Therefore, joint optimization of multiple modules is a way to improve their combined performance, and this is an area where AI technologies excel, such as in the joint design of receivers.

[0100] In one possible implementation, a convolutional neural network (CNN) structure can be used to design a receiver, as shown in Figure 2. CNNs in Figure 2 represent multi-layer convolutional neural networks. As can be seen, the convolutional neural network structure design still follows the current receiver structure and is primarily used for channel estimation, equalization, and demodulation. However, due to the limited range of correlations that CNN can extract, it can usually only extract local information, such as the correlation between adjacent signal slices. Therefore, the main disadvantages of designing a receiver with a CNN structure are the small number of CNN parameters and the poor generalization performance of the convolution operation principle for different scenarios. For example, for complex frequency-selective channels, even using a convolutional neural network to design a receiver in multiple-input multiple-output (MIMO) scenarios will not achieve good performance.

[0101] To achieve optimal receiver performance, multiple modules can be jointly optimized to design an AI model, which then processes received signals. However, in wireless networks, varying channel conditions can significantly impact the performance of the AI ​​model, and thus the receiver. Consequently, it can be difficult for terminals to determine which AI model to use for signal reception.

[0102] It is understandable that the technical solutions provided in the embodiments of the present application can be applied to a first device and a second device. The first device may be a terminal device or a network device. Similarly, the second device may be a terminal device or a network device. For example, if the first device is a terminal device, the second device may be a terminal device or a network device. For another example, if the first device is a network device, the second device may be a terminal device or a network device. Hereinafter, the first device is a network device, such as a base station, and the second device is a terminal device, such as a terminal.

[0103] In view of this, an embodiment of the present application provides an information transmission method. In this method, a terminal can receive first information from a base station. This first information can indicate a first model for local data processing of a data stream and a second model for global data processing of the data stream. In this way, the terminal can use the first model and the second model to process the output signal received from the base station to obtain second data. Based on this solution, the base station can determine the first model and the second model for data processing and indicate the first model and the second model to the terminal. In this way, the terminal can determine to use the first model and the second model for signal processing, thereby improving receiver performance.

[0104] In order to facilitate understanding of the technical solutions provided by the embodiments of the present application, the first model and the second model are introduced below. Among them, the first model can be used for local data processing. For example, local data processing may include processing one or more symbols, such as one or more orthogonal frequency division multiplexing (OFDM) symbols, processing one or more data streams, processing one or more resource block groups (RBGs), processing one or more RBG bundlings, processing one or more control channel elements (CCEs), processing one or more aggregation levels, processing one or more search spaces, processing one or more control resource sets (CORESETs), or processing one or more CORESET combinations (a CORESET combination contains one or more CORESETs).

[0105] The second model provided in the embodiments of the present application can be used for global data processing. For example, global data processing can include processing all input data, such as jointly processing the output of the first model. Optionally, the first model can process local data, and the second model can process global data.

[0106] In an embodiment of the present application, the base station may input data, such as uncoded bits, coded bits, etc., into the first model and the second model, thereby obtaining output signals, such as modulation symbols or waveform symbols, etc. Similarly, the transmitter and the receiver may have the same structure, and the terminal may input signals received from the base station, such as unprocessed signals, processed modulation symbols, waveform signals, air interface data symbols, air interface pilot symbols, or filtered air interface symbols, etc., into the first model and the second model, thereby obtaining second data, such as channel estimation results, channel equalization weights, noise power, log likelihood ratios, or decoded bits, etc.

[0107] Below, in conjunction with Figure 3, we describe the data processing methods for base stations and terminals according to embodiments of the present application. The method shown in Figure 3 can be applied to communication devices, such as terminals or base stations. Specifically, this method is applicable to the receiving end, or it can be applied only to the transmitting end. Referring to Figure 3, an exemplary flowchart of signal processing by a base station or terminal according to embodiments of the present application can include the following operations.

[0108] S301: Input M1 input signals into the input layer of the neural network for dimensionality increase processing to obtain M2 high-dimensional vectors, where M2 is a positive integer. It is understood that M2 and M1 can be the same or different, and M2 can be greater than or less than M1.

[0109] It is understood that the input signal in S301 can be a received signal, such as a signal received from a base station. Alternatively, the input signal in S301 can be a signal obtained from a higher layer of the communication device, such as the physical layer. Exemplarily, when the method is used on a transmitting end, the input signal can be bits or symbols. Exemplarily, when the method is used on a receiving end, the input signal can be a symbol received by the receiving end; alternatively, the input signal can be an input signal representing probability, such as a log-likelihood ratio of a symbol.

[0110] In S301, the input layer of the neural network can perform dimensionality-upgrading on the input signal to obtain a high-dimensional vector. For example, the input layer of the neural network can upscale the 16-dimensional input signal to 256 dimensions. In the following, for ease of description, the dimension of the high-dimensional vector is described as d k , d k Is a positive integer.

[0111] S302: Input the M2 high-dimensional vectors into the processing layer of the neural network to obtain M3 high-dimensional vectors.

[0112] The processing layer of the neural network may include a first model and a second model.

[0113] In one possible implementation, the processing layers of a neural network can be used to process high-dimensional vectors. For example, a first model can optimize a single high-dimensional vector, while a second model can optimize multiple high-dimensional vectors based on the relationships between the multiple high-dimensional vectors. It will be appreciated that the manner in which the first model processes high-dimensional vectors and the manner in which the second model processes high-dimensional vectors are pre-trained.

[0114] For example, the training of the first model and the second model can be based on a loss function. When the first model and the second model are applied to different scenarios, the loss functions can be different. For example, for training to reduce the peak to average power ratio (PAPR) of the transmitter, the neural networks on both the transmitting and receiving sides can be trained jointly, or the training can be completed only on the transmitter side. Taking unilateral training as an example, the loss function can be the ratio of the maximum value and the average value of the transmitter output signal, and the training goal is to minimize the loss function. Therefore, after the training is completed, the role of the first model can be understood as reducing the PAPR of the local transmission signal, or reducing the PAPR of the local transmission signal in high-dimensional space. For MIMO scenarios, it can be to reduce the PAPR of the transmission signal within the stream. The role of the second model can be understood as reducing the difference between the maximum value / average value of the global output signal, or reducing the difference between the maximum value / average value of the global output signal in high-dimensional space. For MIMO scenarios, it can be to reduce the difference between the maximum value / average value of the output signals of different streams.

[0115] For example, to reduce the bit error rate, take the example of joint training between the transmitter and receiver: the training loss function is the cross entropy between the input symbol / bit and the output log-likelihood ratio (LLR) / bit, which is equivalent to maximizing the mutual information between the system input and output. The role of the first model of the transmitter is to increase the distance between different local data in the frequency domain and time domain, such as the distance between different symbols. The role of the first model of the receiver is to distinguish each local data as much as possible, such as distinguishing different symbols, in conjunction with the first model of the transmitter. The role of the second model of the transmitter is to further increase the distance between different local data, such as the distance between different symbols, by combining the spatial characteristics of the channel. The role of the second model of the receiver is to perform interference elimination and differentiation on each local data in the spatial domain, such as eliminating interference and differentiation on the symbol stream.

[0116] S303: Input the M3 high-dimensional vectors into the output layer of the neural network for operation to obtain an output signal.

[0117] In S303, the output layer of the neural network can perform operations based on M3 high-dimensional vectors to obtain an output input signal. It can be understood that the operation of the output layer of the neural network varies depending on the scenario. For example, in the channel decoding scenario, the operation of the output layer of the neural network can be to obtain the log-likelihood ratio based on M3 high-dimensional vectors, that is, the output signal is the log-likelihood ratio. For another example, in the channel coding scenario, the output layer of the neural network can be to obtain the encoded code block based on M3 high-dimensional vectors, that is, the output signal is the encoded code block. For another example, when the input signal of S301 is the log-likelihood ratio of symbols, the corresponding output signal can be a bit sequence.

[0118] Optionally, in S303 , the output layer of the neural network may further perform dimensionality reduction on the M3 high-dimensional vectors, and obtain an output signal based on the M3 high-dimensional vectors after dimensionality reduction.

[0119] Hereinafter, M1, M2, and M3 are set to N for convenience of description.

[0120] Referring to FIG. 4 , the structure of a neural network provided in an embodiment of the present application is described. FIG. 4 is an exemplary structural diagram of a neural network provided in an embodiment of the present application. It should be understood that FIG. 4 is merely an example of a neural network structure and does not constitute a limitation on the structure of a neural network. Those skilled in the art may design other neural networks for signal processing based on the neural network shown in FIG. 4 to implement the operations shown in FIG. 3 .

[0121] As shown in Figure 4, the input layer is the input signal x, which is x1 to x N The input layer performs dimensionality-raising processing on N input signals to obtain N high-dimensional vectors S, which are S1 to S N . Among them, the dimension of each high-dimensional vector is d k , that is, the dimension of the output of the input layer is N×d k The input of the processing layer is the output of the input layer. The processing layer can be used to process S1 to S N It can be understood that the processing of high-dimensional vectors can be considered as the processing layer in N×d k Dimensional d k Therefore, the output of the processing layer is also N×d k After the processing layer, N high-dimensional vectors h1 to h N The input of the output layer is the output of the processing layer. The output layer can obtain h1 to h N Output input signal o0 to o N .

[0122] It should be noted that the dimensions of the processing layer's input and output are the same, so the number of inputs and outputs is the same. The input layer can increase or decrease the number of input signals. Similarly, the output layer can also increase or decrease each high-dimensional vector of the input.

[0123] Based on the above scheme, each input signal is processed by the processing layer. Since the processing process is independent of the input signal's dimension, the complexity of signal processing can be reduced. In addition, the number of input signals input to the processing layer can be varied, thus achieving scalability for signals of different dimensions.

[0124] In one possible implementation, the processing layer may include at least one model group. Model group 1 may include a second model 1_1 and one or more first models 1_2. The second model 1_1 may be used to process N high-dimensional vectors, and the one or more first models 1_2 may be used to process each high-dimensional vector.

[0125] In one possible scenario, the first model l_2 included in the model group 1 can be one, that is, the N high-dimensional vectors can be processed using the same first model l_2. For example, in a MIMO scenario, the N high-dimensional vectors can correspond to one or more data streams, and one first model l_2 can process all of the data streams. In another possible scenario, the first model l_2 can be multiple, that is, the N high-dimensional vectors can be processed using different first models l_2. For example, in a MIMO scenario, the N high-dimensional vectors can correspond to one data stream, and the one data stream can be processed using multiple first models l_2, such as each first model l_2 can process a portion of a data stream. For another example, in a MIMO scenario, the N high-dimensional vectors can correspond to multiple data streams, and the multiple data streams can be processed using multiple first models l_2, such as each first model l_2 can process one data stream, or each first model l_2 can process a portion of the multiple data streams.

[0126] It is understood that the order of the second model and one or more first models in each model group is not specifically limited. Because the input and output dimensions of the processing layer are the same, L layers of iteration can be performed at the processing layer. In other words, L groups of model groups can be connected to each other to achieve the purpose of performing L layers of iteration at the processing layer. Where l is an integer greater than or equal to 1 and less than L.

[0127] Referring to Figure 5A , in model group 1, second model 1_1 precedes one or more first models 1_2. This means the output of second model 1_1 serves as the input to one or more first models 1_2. In model group 2, second model 2_1 precedes one or more first models 2_2. This means the output of one or more first models 2_2 serves as the input to second model 2_1. As can be seen in Figure 5A , the input to second model 1_1 can be a high-dimensional vector or the output of one or more first models 1_2. The input to one or more first models 1_2 can be the output of second model 1_1 or the output of one or more first models 1_2.

[0128] Referring to Figure 5B , in model group 1, second model 1_1 precedes one or more first models 1_2. This means the output of second model 1_1 serves as the input to one or more first models 1_2. In model group 2, second model 2_1 precedes one or more first models 2_2. This means the output of second model 2_1 serves as the input to one or more first models 2_2. As can be seen in Figure 5B , the input to second model 1_1 can be a high-dimensional vector or the output of one or more first models 1_2. The input to one or more first models 1_2 can be the output of second model 1_1.

[0129] Referring to Figure 5C , in model group 1, second model 1_1 comes after one or more first models 1_2, meaning the outputs of one or more first models 1_2 serve as inputs to second model 1_1. In model group 2, second model 2_1 comes before one or more first models 2_2, meaning the outputs of second model 2_1 serve as inputs to one or more first models 2_2. As can be seen in Figure 5C , the inputs to second model 1_1 can be either the outputs of one or more first models 1_2 or the outputs of second model 1_1. The inputs to one or more first models 1_2 can be either high-dimensional vectors or the outputs of second model 1_1.

[0130] Referring to Figure 5D , in model group 1, second model 1_1 comes after one or more first models 1_2. This means that the outputs of one or more first models 1_2 serve as inputs to second model 1_1. In model group 2, second model 2_1 comes after one or more first models 2_2. This means that the outputs of one or more first models 2_2 serve as inputs to second model 2_1. As can be seen in Figure 5D , the inputs to second model 1_1 can be the outputs of one or more first models 1_2. The inputs to one or more first models can be high-dimensional vectors or the outputs of second model 1_1.

[0131] In one possible implementation, when the number of dimensions of the high-dimensional vector is greater than 2, for example, the dimension of the high-dimensional vector is N×d k When [the value of the third model is y], the processing layer may further include a third model. The third model may be used to process high-dimensional vectors in the y dimension, or to transform high-dimensional vectors, such as converting a three-dimensional vector into a two-dimensional vector. It is understood that the order of the third model relative to the second model and one or more first models is not specifically limited. See the above description of the order of the second model relative to one or more first models.

[0132] Based on the above scheme, the performance gain of the neural network can be obtained by implementing multi-layer iteration of the processing layer through multiple groups of model groups.

[0133] Hereinafter, the processing layer mentioned in the embodiments of the present application will be further explained and introduced through methods one to three.

[0134] Method 1:

[0135] The second model may be an operation of the attention layer, and one or more first models may be an operation of the fully connected layer. In other words, the operation of the processing layer may include the operation of the attention layer and the operation of the fully connected layer. The operation of the attention layer may satisfy the following formula (1):

[0136] in W Q 、W K 、W V are the trained parameters. is the input of the attention layer, where N can be understood as the number of high-dimensional vectors S, d k is the dimension of each high-dimensional vector S. Among them, ATT(Q,K,V) is the output of the attention layer, which represents the correlation between any two high-dimensional vectors. Q, K, V represent the results of the three linear transformations of the input S, Q represents the query vector, W Q represents the query vector weight, K represents the keyword vector, W K represents the keyword vector weight, V represents the value vector, W V Represents the value vector weight. It can be seen that the input of the attention layer is N×d k Dimensional.

[0137] Optionally, if the attention layer is extended to multi-head attention, then W Q , W K With W V There are multiple sets of values. It is understandable that the parameters of the attention layer of each model group can be different.

[0138] It should be noted that W Q 、W K 、W V Training can be performed by gradient backpropagation. For example, by Q 、W K 、W V The initial parameters of are set to random numbers. The high-dimensional vectors during training are processed through the attention layer to obtain the output and input signals during training. The training gradient is obtained by comparing the known output and input signals with the output and input signals during training, and the gradient is back-propagated to the attention layer to adjust W Q 、W K 、W V Through the above training method, W Q 、W K 、W V Perform multiple trainings to obtain the trained W Q 、W K 、W V .

[0139] The fully connected layer operates on each high-dimensional vector independently, and each high-dimensional vector shares the parameters of the fully connected layer. The operation of the fully connected layer can satisfy the following formula (2): y=f(xW M +b) Formula (2)

[0140] Where y represents the output of the fully connected layer, x represents the input of the fully connected layer, and f is the activation function. For example, the activation function can be one of the functions such as linear, pseudo-inverse, inverse, sigmoid, softmax, relu, gelu, etc. M W represents the weight of the fully connected layer, and b represents the bias of the fully connected layer. M and b are trained parameters, and the training method can refer to the above W Q 、W K 、W V It is understandable that the parameters of the fully connected layer of each model group can be different.

[0141] Refer to Figure 6 for an example of a processing layer. As shown in Figure 6, the normalized dashed box can represent an optional position of the normalization operation. As can be seen from Figure 6, the normalization operation can be set before the attention layer, after the attention layer, before the fully connected layer and / or after the fully connected layer. It can be understood that the processing layer may include one or more normalization operations. The normalization operation can normalize the dimensions of a batch of data, that is, the normalization operation can perform batch normalization on high-dimensional vectors. Alternatively, the normalization operation can also normalize the dimensions of a set of data, that is, the normalization operation can perform layer normalization on high-dimensional vectors. Exemplarily, the normalization operation can satisfy the following formula (3):

[0142] Where x represents the input of the normalization operation, y represents the output of the normalization operation, μ represents the mean calculated based on x, σ represents the variance calculated based on x, ∈ is the preset minimum value to prevent the denominator from being zero, and γ and β are trained parameters.

[0143] Taking the normalization operation before the attention layer and before the fully connected layer as an example, the input of the first normalization operation can be the N high-dimensional vectors output by the input layer. The first normalization operation can refer to the aforementioned formula (3), and the output of the first normalization operation can be the normalized N high-dimensional vectors. The input of the attention layer can be the output of the first normalization operation. The operation of the attention layer can refer to the aforementioned formula (1), and the attention layer can process N high-dimensional vectors. The output of the attention layer can be N high-dimensional vectors. As can be seen from Figure 6, the input of the second normalization operation can be the output of the attention layer, and optionally N high-dimensional vectors. Adding N high-dimensional vectors to the input of the second normalization operation can accelerate convergence and solve the problem of gradient dispersion. The second normalization operation can normalize the N high-dimensional vectors output by the attention layer, and the output of the second normalization operation can be the normalized N high-dimensional vectors. Among them, the dimension of the output of the normalization operation is the same as the dimension of the output of the fully connected layer. The input of the fully connected layer can be the output of the second normalization operation. The operation of the fully connected layer can be referred to in the above formula (2). The fully connected layer can process a high-dimensional vector. The output of the fully connected layer can be understood as a high-dimensional vector processed by the second model.

[0144] It should be noted that in the second model group, the input of the third normalization operation can be N high-dimensional vectors processed by the first model and the second model. Similarly, the processing layer shown in Figure 6 can be iterated L times.

[0145] In another possible scenario, the iterative operation can be stopped based on actual conditions. For example, a convergence condition can be set, such as convergence when the average error between the outputs of the lth iteration and the l+1th iteration is less than a threshold. Where l is an integer greater than or equal to 1 and less than L.

[0146] In one example, the input and output of each operation in the above processing layer can be viewed as a matrix, which can be N×d k Then the input of the second normalization layer can be the matrix obtained by adding the output of the attention layer and N high-dimensional vectors, that is, the input of the second normalization operation is still N×d k dimensional matrix.

[0147] It should be noted that in the embodiments of the present application, each operation of the neural network, such as the input layer, the second model in the processing layer, one or more first models, and the output layer, can be set as an operation of a fully connected neural network. The operation of the second model in the processing layer can be to apply a self-attention mechanism to the outputs of multiple fully connected neural networks.

[0148] By setting up the processing layer based on the first approach described above, the input signal is independently transformed. This means that the neural network parameters are independent of the input signal dimensions, thereby making the neural network easily scalable. Furthermore, the use of the attention layer to extract the correlation between any two input signals allows for a greater number of parameters, allowing the neural network to be applied to a wider range of communication scenarios.

[0149] Method 2:

[0150] The operations performed by the second model may be based on the first matrix, and the operations performed by one or more first models may be based on one or more second matrices. In other words, the operations of the processing layer may include operations based on the first matrix and operations based on the second matrix.

[0151] Referring to FIG. 7A , the first matrix may be an N×N matrix.

[0152] In one possible case, the first matrix can be a trained matrix. The training method of the first matrix can refer to the above W Q 、W K 、W V As shown in FIG7B , the elements of the main diagonal of the first matrix are all U, and the elements of the non-main diagonal are all V. The elements U on the main diagonal are used to represent the characteristics of the input signal itself, and the elements V on the non-main diagonal represent the characteristics of a certain input signal and other input signals. For example, when the dimension of the input high-dimensional vector is 3×d k, the size of the first matrix is ​​a 3×3 matrix. The element in the first row and first column of the first matrix is ​​U, which is used to process the first high-dimensional vector. The element in the first row and second column is V, which is used to process the first high-dimensional vector and the second high-dimensional vector. The element in the first row and third column is V, which is used to process the first high-dimensional vector and the third high-dimensional vector, and so on. Since the first matrix is ​​an N×N matrix, that is, the size of the first matrix increases as the number of input high-dimensional vectors increases, the first matrix is ​​scalable.

[0153] As shown in FIG7A , assuming the first output of the second model, in other words, the first row of the matrix of the output of the second model is h1=f(S1U, S2V, S3V, S4V…). Here, f can be a sum, representing the multiplication of S by the first matrix. It is understood that f can also be a maximum, minimum, or average value, etc., which is not specifically limited in this application. S can represent an input high-dimensional vector.

[0154] In another possible case, the first matrix is ​​calculated based on the input N high-dimensional vectors. The first matrix can be f(R S ), f(C S ), f(cos S ) etc., R S represents the autocorrelation matrix of S, C S represents the autocovariance matrix of S, cos S Indicates the calculation of the cosine similarity between two high-dimensional vectors S, f represents a normalization or nonlinear operation. S can represent the input high-dimensional vector.

[0155] In another possible case, the first matrix can be in the form of a fully connected neural network. For example, the output h of the second model S =f(SW S +b). Where W S and b are trained parameters, and the training method can refer to the above W Q 、W K 、W V The training method is implemented, and f is the activation function. Among them, W S It can be a weight and b can be a bias.

[0156] In the embodiment of the present application, b is used to represent the bias of the fully connected neural network. The value of b in different modules may be different. The specific value of b depends on the actual training results. For example, h S The value of b in and the value of b in the above formula (2) may be different.

[0157] Referring to FIG. 7C , the second matrix may be d k ×d kThe second matrix can be a trained parameter, and the training method can refer to the above W Q 、W K 、W V The training method is implemented. Assume that the first output of one or more first models h′1=f(h1W). Wherein, f can be a sum, representing the multiplication of h1 and the first matrix, h1 represents the first output of the second model, and W represents the weight. It is understandable that f can also be a maximum value, a minimum value, or an average value, etc., which is not specifically limited in this application.

[0158] It is understandable that the second matrix can also be expanded into the form of a fully connected neural network, such as h′=f(hW+b), where W and b are trained parameters, and the training method can refer to the above W Q 、W K 、W V The training method is implemented, and f is the activation function.

[0159] Optionally, the first matrix and the second matrix may be cascaded multiple times in one iteration, that is, the operation of the second model may be based on the operation of multiple first matrices, and the operation of one or more first models may be based on the operation of one or more second matrices.

[0160] Refer to FIG8 for an example of a processing layer. As shown in FIG8 , the normalized dotted box can represent the optional position of the normalization operation. The setting of the normalization operation can refer to the relevant description in FIG6 , which will not be repeated here. As shown in FIG8 , before the first normalization operation, the matrix transposition operation can be performed on the input high-dimensional vector. This is because the input of the processing layer is N×d k dimensional, and the second model is an operation on the dimension N, so the matrix transposition operation can be performed on the input high-dimensional vector, that is, N×d k The high-dimensional vector of dimension d is transposed to d k ×N dimensions. Similarly, a matrix transposition operation can be added after the output of the second model, that is, the matrix transposition operation can be performed on the output of the second model. This is because the output of the second model is d k ×N-dimensional, and one or more first models are d k The operation on this dimension can therefore perform a matrix transposition operation on the output of the second model, that is, d k N high-dimensional vectors of dimension N×N are transposed into N×d k Dimensional.

[0161] By setting up the processing layer based on the second approach described above, independent transformation of the input signal is achieved, and the input signal can be processed using the first matrix and the second matrix. Therefore, by expanding the size of the matrix, the neural network becomes more scalable. Furthermore, using the first matrix, the correlation between any two input signals can be extracted, resulting in a larger number of parameters, allowing the neural network to be applied to a wider range of communication scenarios.

[0162] Method 3:

[0163] In method 3, the operations of the processing layer can be those of a graph neural network. A detailed description is given below.

[0164] The input of graph neural network can be high-dimensional vector The 0 in can represent the initial state of the high-dimensional vector. The communication device can use each high-dimensional vector as a node of the graph neural network, that is, the graph neural network can have N nodes. In other words, the communication device can set each high-dimensional vector as the initial state of a node of the graph neural network, such as the state of the 0th iteration, and one high-dimensional vector can correspond to one node. In the graph neural network, there is no explicit structural distinction between the second model and one or more first models, but there is a distinction between the second model and one or more first models in the operation steps of the graph neural network.

[0165] Refer to Figure 9, which is an example of a processing layer provided in an embodiment of the present application. In Figure 9, S1, S2, S3 and S4 can be nodes of a graph neural network. Among them, the operation of the second model can be S N Get the aggregation status of other adjacent nodes. N Represents any node of the graph neural network. For example, S1 can obtain the aggregated state of S2, S3, and S4. The operation of the second model can satisfy the following formula (4):

[0166] The aggregation function (AGGREGATE) can be used to find the maximum value, minimum value, average value, or sum, etc.

[0167] In the above formula (4), Z represents the number of iterations, represents other nodes adjacent to node v, represents the aggregation state of other nodes connected to node v at the Zth iteration, represents the state of the Z-1th iteration of other nodes adjacent to node v, where Z is an integer greater than or equal to 1. Z can be a preset number of iterations. It is understood that the state of the 0th node can be a high-dimensional vector corresponding to the node.

[0168] The operation of one or more first models may be S NUpdate its own state according to the aggregated state of other adjacent nodes for the Zth time and its own state for the Zth time. The operation of one or more first models can satisfy the following formula (5):

[0169] Where σ is the activation function, Z represents the number of iterations, and Z is an integer greater than or equal to 1. Z can be a preset number of iterations. K represents the weight of the Zth iteration, represents the state of node v at the Z-1th iteration, Represents the aggregation state of other nodes connected to node v at the Zth iteration. The merge function (concat) can represent the cascade, that is, and Concatenation. It can be understood that concatenation can be understood as joining two vectors into one vector.

[0170] After K iterations, the graph neural network can output the state of each node at the Zth time and the aggregated state of the adjacent nodes obtained by each node at the Zth time. It can be understood that the dimension of the state of each node at the Zth time and the aggregated state of the adjacent nodes obtained by each node at the Zth time is d k The dimension of the state of each node output by N nodes at the Zth time and the aggregated state of the adjacent nodes obtained by each node at the Zth time is N×d k The state of each node at the Zth time can be understood as a high-dimensional vector optimized by the first model. The aggregated state of the adjacent nodes obtained by each node at the Zth time can be understood as a high-dimensional vector optimized by the second model.

[0171] In conjunction with Figure 10, the manner in which the base station or terminal performs data processing in an embodiment of the present application is introduced. Referring to Figure 10, the base station can input data into a neural network, and the processing layer of the neural network can include a first model and a second model. The base station can obtain an output signal based on the neural network, and the processing method for the base station to obtain the output signal can be implemented with reference to the embodiment shown in Figure 3, which will not be repeated here. The base station can send the output signal through an air interface (Uu). Optionally, the base station can perform layer mapping, port mapping, insert DMRS symbols, precoding, and generate orthogonal frequency division multiplexing (OFDM) and other operations on the output signal and then send it through the air interface.

[0172] The terminal can receive the output signal sent by the base station and input the output signal into the neural network to obtain the second data. The processing method for the terminal to obtain the output signal can be implemented with reference to the embodiment shown in Figure 3 and will not be described in detail here. Optionally, the terminal can perform operations such as OFDM waveform decoding and physical resource mapping on the output signal before inputting it into the neural network. Optionally, the second data can be a log-likelihood ratio, and the terminal can perform a decoding operation on the log-likelihood ratio output by the neural network to obtain information bits.

[0173] If the base station output signal includes DMRS, the terminal can perform channel estimation and channel equalization based on DMRS. The terminal can input the channel estimation results and channel equalization results into the neural network to obtain channel estimation results and channel equalization results with better performance.

[0174] It is understood that the DMRS in the embodiment of the present application can be a sparse and frequency-domain dispersed DMRS, as shown in Figure 11. The resources of the same color in Figure 11 use code division multiplexing to multiple ports, and each port occupies 1 resource element (RE) / resource block (RB). For this sparse and frequency-domain dispersed DMRS, the occupation of RE / RB can be reduced, saving resources. However, due to its poor channel estimation or channel equalization results, the channel estimation results and channel equalization results of the terminal can be input into the neural network to improve performance.

[0175] The following describes an information transmission method provided by an embodiment of the present application through the accompanying drawings. Referring to FIG12 , which is an exemplary flow chart of an information transmission method provided by an embodiment of the present application, the method may include the following operations.

[0176] S1101: The base station processes first data based on the first model and the second model to obtain an output signal.

[0177] For example, the base station may process the first data in the manner shown in FIG3 to obtain an output signal. When the base station processes the first data in the manner shown in FIG3 , the input signal in FIG3 may be the first data.

[0178] S1102: The base station sends an output signal to the terminal.

[0179] Correspondingly, the terminal can receive the output signal sent from the base station.

[0180] It is understandable that the output signal sent by the base station in S1102 can be the output of the aforementioned neural network, or it can be a signal obtained by processing the output of the neural network, such as a signal after inserting DMRS and precoding to generate an OFDM waveform.

[0181] S1103: The base station sends first information to the terminal.

[0182] Correspondingly, the terminal receives the first information from the base station.

[0183] The first information may indicate a first model and a second model.

[0184] S1104: The terminal processes the output signal based on the first model and the second model to obtain second data.

[0185] For example, the terminal can process the output signal based on the method shown in Figure 3 to obtain the second data. When the terminal processes the output signal based on the method shown in Figure 3, the input signal in Figure 3 can be the output signal, and the input signal in Figure 3 can be the second data.

[0186] In one possible implementation, the base station and the terminal may be configured with multiple pre-trained models, such as multiple first models and multiple second models, for adapting to multiple channel states or scenarios. In one possible scenario, the base station may determine the first model and the second model based on the channel state. For example, the base station may send a downlink reference signal, such as a channel state information (CSI)-reference signal (RS). The terminal may measure the received CSI-RS to obtain CSI, such as channel quality indicator (CQI), rank indicator (RI) and precoding indicator (PMI). The terminal may feed back the CSI to the base station. In this way, the base station may determine the first model and the second model based on the CSI reported by the terminal.

[0187] For another example, the terminal may send an uplink sounding reference signal (SRS), and the base station may measure the received SRS to determine the channel state. For example, the base station may measure the SRS signal to interference plus noise ratio (SNR), reference signal received power (RSRP), or reference signal received quality (RSRQ). In this way, the base station may determine the first model and the second model based on the SRS measurement results.

[0188] The measurement results of the above-mentioned CSI or SRS can be used to determine the channel state, so that they can correspond to the first model and the second model. For example, the base station and the terminal can be configured with a correspondence between the first model and the channel state, and a correspondence between the second model and the channel state. Among them, the first model and the channel state can be many-to-one, that is, multiple first models can correspond to one channel state, or the first model and the channel state can be one-to-many, that is, one first model can correspond to multiple channel states, or the first model and the channel state can correspond one-to-one, that is, one first model can correspond to one channel state. Similarly, the second model and the channel state can be many-to-one, one-to-many or one-to-one.

[0189] In another possible scenario, the first model and / or the second model may be determined based on the scenario. For example, time-frequency domain resources, throughput, peak to average power ratio (PAPR), adjacent channel leakage ratio (ACLR), phase noise suppression or terminal mobility. For example, different time-frequency domain resources may correspond to different first models and / or second models. For another example, high throughput and low throughput may correspond to different first models and / or second models. For another example, different PAPRs may correspond to different first models and / or second models, such as low PAPR and high PAPR may correspond to different first models and / or second models. For another example, terminals with different mobilities may correspond to different first models and / or second models, such as terminals with high mobility and terminals with low mobility may correspond to different first models and / or second models.

[0190] In the above solution, the base station can determine the first model and the second model based on the channel state or scenario, and thus indicate the first model and the second model used for data processing to the terminal according to the first information.

[0191] In one possible implementation, the first information may indicate the structure of the first model and the structure of the second model. For example, the first information may indicate the number of layers of the first model and the second model, the number of nodes in each layer, the connection relationship between nodes in each layer, the connection relationship between each layer, and the weight of each node, thereby indicating the structure of the first model and the structure of the second model to the terminal.

[0192] In another possible implementation, the first information may indicate an index of the first model and an index of the second model. For example, the terminal and the base station may preconfigure multiple first models and multiple second models, and assign an index to each first model and an index to each second model. In this way, the first information may indicate the index of the first model and the index of the second model. The terminal may determine the corresponding first model based on the index of the first model, and the corresponding second model based on the index of the second model.

[0193] The following introduces an implementation method in which the first information indicates the index of the first model and the index of the second model.

[0194] Case 1: A new field is added in the signaling to carry the first information, indicating the index of the first model and the index of the second model.

[0195] For example, a new field may be added to downlink control information (DCI), radio resource control (RRC) signaling, or other control signaling to carry the first information. To facilitate distinction, the field indicating the index of the first model is referred to as a local processing module index (single input single output block index, SBI), and the index indicating the second model is referred to as a global processing module index (cross layer block index, CBI).

[0196] Taking the newly added SBI and CBI in DCI as an example, the format of the newly added fields in DCI may be as shown in Table 1.

[0197] Table 1: Example of the format of SBI and CBI in a DCI

[0198] Among them, SBI can be shown in Table 2, and CBI can be shown in Table 4.

[0199] Table 2: Example of an SBI configuration table

[0200] In the example shown in Table 2, the SBI may be 3 bits, 4 bits, or 6 bits, and may indicate an index of the first model, such as an index of one or more first models.

[0201] In one possible case, in a MIMO scenario, the SBI provided in the embodiment of the present application may indicate the number of transmitted streams, as shown in Table 3.

[0202] Table 3: Example of an SBI configuration table

[0203] As shown in Table 3, when the SBI value is "0", it can indicate that the number of transmission streams is 2, and can indicate the model with index "0" and the model with index "8". Optionally, in this case, one model can process the data stream of one layer, such as the model with index "0" can be used to process the first layer data stream, the first model with index "8" can be used to process the second layer data stream, the model with index "8" can be used to process the first layer data stream, the first model with index "0" can be used to process the second layer data stream, and so on.

[0204] Table 4: Example of a CBI configuration table

[0205] In the example shown in Table 4, the CBI may indicate the index of the second model.

[0206] In one possible scenario, the second model may correspond to the number of transmission streams in the MIMO scenario, and different second models may be allocated according to the association relationship between layers, as shown in Table 5.

[0207] Table 5: Example of a CBI configuration table

[0208] As shown in Table 5, when the number of transmission streams is 1, the base station can use the second model with an index of "0" and indicate this to the terminal through the CBI. For another example, when the number of transmission streams is 2 and the correlation between layers is strong, the base station can use the second model with an index of "1" and indicate this to the terminal through the CBI, and so on.

[0209] Optionally, in an embodiment of the present application, the output signal sent by the base station to the terminal may or may not include a DMRS. The first information may also indicate whether the output signal includes a DMRS. For example, the SBI may indicate whether the output signal includes a DMRS. For another example, the SBI may indicate whether the output signal includes a DMRS.

[0210] Similarly, taking the newly added SBI and CBI in DCI as an example, the format of the newly added fields in DCI may be as shown in Table 6.

[0211] Table 6: Example of the format of SBI and CBI in a DCI

[0212] As shown in Table 6, the SBI and / or CBI may indicate whether the output signal includes a DMRS. If the SBI indicates whether the output signal includes a DMRS, it may be as shown in Table 7.

[0213] Table 7: Example of an SBI configuration table

[0214] As shown in Table 7, SBI can indicate the presence or absence of DMRS in the output signal.

[0215] If CBI indicates whether the output signal includes DMRS, the CBI may be as shown in Table 8.

[0216] Table 8: Example of a CBI configuration table

[0217] As shown in Table 8, CBI can indicate the presence or absence of DMRS in the output signal.

[0218] Based on the above solution, the index of the first model and the index of the second model can be indicated by SBI and CBI, and the number of transmission streams and the presence or absence of DMRS can be optionally indicated.

[0219] It is understandable that the above-mentioned SBI and CBI can be two different fields. Those skilled in the art can design the SBI and CBI into a single field to jointly indicate the index of the first model and the index of the second model. For example, the first x bits of a field are used to indicate the index of the first model, and the remaining bits are used to indicate the index of the second model, etc. This application does not make specific limitations.

[0220] Case 2: An existing field in the multiplexing signaling carries the first information, indicating the index of the first model and the index of the second model.

[0221] For example, the first information may be carried in a reserved field in the signaling to indicate the index of the first model and the index of the second model.

[0222] For another example, the index of the first model and / or the index of the second model can be indicated by the antenna port field in the DCI. For example, the antenna port can be associated with the first model and / or the second model, so that the antenna port field can indicate the antenna port number while indicating the index of the first model and / or the index of the second model.

[0223] For another example, the index of the first model and the index of the second model can be jointly indicated by multiplexing different fields to carry the first information. For example, the antenna ports field in the DCI can be 4 bits, so the index of the first model can be jointly indicated by the antenna port field and other fields, such as part of the reserved field or the channel characteristic indication field. Taking the channel characteristic indication field as an example, the channel characteristic indication field in the related art can be used to indicate the channel state, such as the channel state mentioned above, which can be 6 bits. Exemplarily, the index of the first model can be indicated by the 4 bits of the antenna port field plus the 2 bits of the channel state indication, such as the first 2 bits or the last 2 bits. The index of the second model is indicated by the remaining bits of the channel state indication.

[0224] Optionally, in case 2, the first information may indicate the presence or absence of DMRS, and may optionally indicate the number of transmitted streams. This may be implemented with reference to case 1 and will not be described in detail here.

[0225] It should be noted that the names of the above fields and the number of bits of the fields are shown only as examples. Those skilled in the art can change their names and number of bits to indicate the index of the first model and the index of the second model. This application does not make specific limitations.

[0226] In one possible scenario, after the measurement of the reference signal is completed, such as after the measurement of the SRS is completed or after the terminal sends CSI to the base station, the base station can send DCI to the terminal, and thus the first information can be carried in the DCI. In another possible scenario, the first information can also be carried in an RRC configuration message or an RRC reconfiguration message, which is not specifically limited in this application.

[0227] In an embodiment of the present application, if the processing layer of the neural network further includes a third model, the base station may send second information to the terminal to indicate the third model, such as indicating the structure of the third model or indicating the index of the third model. The implementation of the second information may refer to the implementation of the first information. Optionally, the second information may also be carried in the first information.

[0228] Based on the concepts of the above embodiments, referring to FIG13 , an embodiment of the present application provides a communication device 1300, which includes a processing unit 1301 and a transceiver unit 1302. The device 1300 can be a communication device, or a device applied to a communication device that can support the communication device to perform a signal processing method.

[0229] The transceiver unit may also be referred to as a transceiver module, transceiver, transceiver, transceiver device, etc. The processing unit may also be referred to as a processor, processing board, processing unit, processing device, etc. Optionally, the device used to implement the receiving function in the transceiver unit may be considered a receiving unit. It should be understood that the transceiver unit is used to perform the sending and receiving operations of the communication device in the above method embodiments, and the device used to implement the sending function in the transceiver unit is considered a sending unit, that is, the transceiver unit includes a receiving unit and a sending unit.

[0230] In addition, it should be noted that if the device is implemented using a chip / chip circuit, the transceiver unit can be an input and output circuit and / or a communication interface, performing input operations (corresponding to the aforementioned receiving operations) and output operations (corresponding to the aforementioned sending operations); the processing unit is an integrated processor or microprocessor or integrated circuit.

[0231] The following describes in detail the implementation of applying the apparatus 1300 to a terminal device and a network device.

[0232] For example, when the apparatus 1300 is applied to a terminal device, operations performed by each unit thereof are described in detail.

[0233] In an optional embodiment, the communication device 1300 can be applied to a terminal device to execute the method executed by the terminal device described above, such as the method executed by the terminal device in the embodiment shown in FIG. 12 . The transceiver unit 1302 is configured to receive first information from a network device, where the first information indicates a first model for performing local data processing and a second model for performing global data processing. The transceiver unit 1302 is also configured to receive an output signal from the network device. The output signal is data processed by the first model and the second model. The processing unit 1301 is configured to process the output signal based on the first model and the second model to obtain second data.

[0234] For example, when the apparatus 1300 is applied to a network device, operations performed by each unit thereof are described in detail.

[0235] In an optional embodiment, the communication device 1300 can be applied to a network device to execute the method executed by the aforementioned network device, specifically, the method executed by the network device in the embodiment shown in FIG. 12 . The transceiver unit 1302 is configured to send first information to a terminal device, where the first information indicates a first model for local data processing and a second model for global data processing. The processing unit 1301 is configured to process data based on the first model and the second model to obtain an output signal. The transceiver unit 1302 is also configured to send the output signal to the terminal device.

[0236] Based on the concepts of the embodiments, as shown in FIG14 , the embodiments of the present application provide a communication device 1400. The communication device 1400 includes a processor 1410. Optionally, the communication device 1400 may further include a memory 1420 for storing instructions executed by the processor 1410, or storing input data required by the processor 1410 to execute instructions, or storing data generated after the processor 1410 executes instructions. The processor 1410 can implement the method described in the above method embodiment using the instructions stored in the memory 1420.

[0237] Based on the concept of the embodiment, as shown in Figure 15, the embodiment of the present application provides a communication device 1500, which can be a chip or a chip system. Optionally, in the embodiment of the present application, the chip system can be composed of a chip, or can include a chip and other discrete devices.

[0238] Communication device 1500 may include at least one processor 1510 coupled to a memory. Optionally, the memory may be located within or outside the device. For example, communication device 1500 may also include at least one memory 1520. Memory 1520 stores the necessary computer programs, configuration information, computer programs or instructions, and / or data for implementing any of the aforementioned embodiments. Processor 1510 may execute the computer programs stored in memory 1520 to perform the methods of any of the aforementioned embodiments. Optionally, the memory may be integrated with the processor.

[0239] The coupling in the embodiments of the present application is an indirect coupling or communication connection between devices, units, or modules, which can be electrical, mechanical, or other forms, and is used for information exchange between devices, units, or modules. The processor 1510 may operate in conjunction with the memory 1520. The specific connection medium between the transceiver 1530, the processor 1510, and the memory 1520 is not limited in the embodiments of the present application.

[0240] The communication device 1500 may also include a transceiver 1530, and the communication device 1500 can exchange information with other devices through the transceiver 1530. The transceiver 1530 can be a circuit, a bus, a transceiver or any other device that can be used for information exchange, or it can be called a signal transceiver unit. As shown in Figure 15, the transceiver 1530 includes a transmitter 1531, a receiver 1532 and an antenna 1533. In addition, when the communication device 1500 is a chip-type device or circuit, the transceiver in the communication device 1500 can also be an input and output circuit and / or a communication interface, which can input data (or receive data) and output data (or send data). The processor is an integrated processor or microprocessor or integrated circuit, and the processor can determine the output data based on the input data.

[0241] In one possible implementation, the communication device 1500 can be applied to a communication device. Specifically, the communication device 1500 can be a communication device or a device capable of supporting a communication device to implement the functions of the terminal device or network device in any of the above-mentioned embodiments. The memory 1520 stores the necessary computer programs, computer programs, instructions, and / or data to implement the functions of the terminal device or network device in any of the above-mentioned embodiments. The processor 1510 can execute the computer program stored in the memory 1520 to perform the method performed by the terminal device or network device in any of the above-mentioned embodiments.

[0242] Since the communication device 1500 provided in this embodiment can be applied to a terminal device or a network device to implement the method executed by the terminal device or the network device, the technical effects that can be obtained can be referred to the above method embodiments and will not be repeated here.

[0243] In the embodiments of the present application, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present application. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present application may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0244] In an embodiment of the present application, the memory may be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory (volatile memory), such as a random-access memory (RAM). The memory may also be any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in an embodiment of the present application may also be a circuit or any other device that can implement a storage function, for storing computer programs, computer programs or instructions and / or data.

[0245] Based on the above embodiments, referring to FIG16 , an embodiment of the present application also provides another communication device 1600, including: an input / output interface 1610 and a logic circuit 1620; the input / output interface 1610 is used to receive code instructions and transmit them to the logic circuit 1620; the logic circuit 1620 is used to run code instructions to execute the method executed by the terminal device or network device in any of the above embodiments.

[0246] The following describes in detail the operations performed by the apparatus 1600 when applied to a terminal device or a network device.

[0247] In an optional embodiment, the communication device 1600 can be applied to a terminal device to execute the method executed by the terminal device described above, such as the method executed by the terminal device in the embodiment shown in FIG. 12 . The input / output interface 1610 is used to input first information from the network device, where the first information indicates a first model for local data processing and a second model for global data processing. The input / output interface 1610 is also used to input an output signal from the network device. The output signal is data processed by the first model and the second model. The logic circuit 1620 is used to process the output signal based on the first model and the second model to obtain second data.

[0248] In an optional embodiment, the communication device 1600 can be applied to a network device to execute the method executed by the aforementioned network device, specifically, the method executed by the network device in the embodiment shown in FIG. 12 . The input / output interface 1610 is configured to output first information to a terminal device, where the first information indicates a first model for local data processing and a second model for global data processing. The logic circuit 1620 is configured to process data based on the first model and the second model to generate an output signal. The input / output interface 1610 is also configured to output an output signal to the terminal device.

[0249] Since the communication device 1600 provided in this embodiment can be applied to a terminal device or a network device to execute the method executed by the above-mentioned terminal device or network device, the technical effects that can be obtained can be referred to the above-mentioned method embodiments and will not be repeated here.

[0250] Based on the above embodiments, the present application also provides a communication system, which includes at least one terminal device and at least one network device. The technical effects that can be obtained can be referred to the above method embodiments, which will not be repeated here.

[0251] Based on the above embodiments, embodiments of the present application further provide a computer-readable storage medium storing a computer program or instructions. When the instructions are executed, the method performed by the communication device in any of the above embodiments is implemented. The computer-readable storage medium may include any medium capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory, a random access memory, a magnetic disk, or an optical disk.

[0252] To implement the functions of the communication devices of Figures 13 to 16 above, embodiments of the present application further provide a chip, including a processor, for supporting the communication device in implementing the functions of the terminal device or network device in the above method embodiments. In one possible design, the chip is connected to or includes a memory, and the memory is used to store computer programs, instructions, and data necessary for the terminal device or network device.

[0253] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can 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 can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0254] The present application is described with reference to the flow chart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each flow process and / or box in the flow chart and / or block diagram and the combination of the flow process and / or box in the flow chart and / or block diagram can be realized by a computer program or instruction. These computer programs or instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processing machine or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for realizing the function specified in one flow chart flow or multiple flows and / or one box or multiple boxes of the block diagram.

[0255] These computer programs or instructions may also be stored in a computer-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 computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0256] These computer programs or instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0257] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include these modifications and variations.

Claims

1. An information transmission method, characterized in that: include: receiving first information from a first device, the first information indicating a first model for performing local data processing and a second model for performing global data processing; receiving an output signal from the first device; the output signal is a signal obtained by the first device processing the first data through the first model and the second model; Based on the first model and the second model, the output signal is processed to obtain second data.

2. The method according to claim 1, characterized in that Also includes: receiving a downlink reference signal from the first device; Measuring the downlink reference signal to obtain measurement information of the downlink reference signal; The measurement information of the downlink reference signal is used to determine the first model and the second model; Sending measurement information of the downlink reference signal to the first device.

3. The method according to claim 1, characterized in that Also includes: sending an uplink reference signal to the first device, where the uplink reference signal is used by the first device to measure the uplink reference signal to obtain measurement information of the uplink reference signal; The measurement information of the uplink reference signal is used to determine the first model and the second model.

4. The method according to claim 1, characterized in that: The first information is used to indicate an index of one or more of the first models, and the first information is also used to indicate an index of the second model.

5. The method according to any one of claims 1 to 4, characterized in that: The first information includes an antenna port field and a channel characteristic indication field; wherein the antenna port field and the channel characteristic indication field are used to indicate the first model and the second model.

6. The method according to any one of claims 1 to 5, characterized in that: The step of processing the output signal based on the first model and the second model to obtain second data includes: Inputting the output signal into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors; The N1 high-dimensional vectors are processed by the first model to obtain N2 high-dimensional vectors; the first model processes the N1 high-dimensional vectors in a pre-trained manner; The N2 high-dimensional vectors are processed by the second model to obtain N3 high-dimensional vectors; the second model processes the N2 high-dimensional vectors in a pre-trained manner; The N3 high-dimensional vectors are input into the output layer of the neural network to obtain the second data.

7. The method according to claim 6, characterized in that The output signal is input into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors, including: The output signals corresponding to the M data streams are input into the input layer for dimensionality upgrading to obtain the N1 high-dimensional vectors; wherein one high-dimensional vector corresponds to one data stream.

8. The method according to claim 6 or 7, characterized in that: The output signal includes a modulation symbol, a waveform signal, an air interface data symbol, an air interface pilot symbol or a filtered air interface symbol.

9. The method according to any one of claims 6 to 8, characterized in that: The second data includes a channel estimation result, a channel equalization weight, a noise power, a log-likelihood ratio or a decoded bit.

10. The method according to any one of claims 1 to 9, characterized in that: The first information further indicates whether the output signal includes a demodulation reference signal DMRS.

11. The method according to any one of claims 1 to 10, characterized in that: The first information further indicates the number of transmission streams of the output signal.

12. An information transmission method, characterized in that: include: Sending first information to a second device, wherein the first information indicates a first model for performing local data processing and a second model for performing global data processing; Processing the first data based on the first model and the second model to obtain an output signal; The output signal is sent to the second device.

13. The method according to claim 12, characterized in that Also includes: sending a downlink reference signal to the second device; receiving measurement information of the downlink reference signal from the second device; The measurement information of the downlink reference signal is measured by the second device based on the downlink reference signal, and the measurement information of the downlink reference signal is used to determine the first model and the second model.

14. The method according to claim 12, characterized in that Also includes: receiving an uplink reference signal from the second device; The uplink reference signal is measured to obtain measurement information of the uplink reference signal, and the measurement information of the uplink reference signal is used to determine the first model and the second model.

15. The method according to any one of claims 12 to 14, characterized in that: The first information is used to indicate an index of one or more first models, and the first information is also used to indicate an index of the second model.

16. The method according to any one of claims 12 to 15, characterized in that: The first information includes an antenna port field and a channel characteristic indication field; wherein the antenna port field and the channel characteristic indication field are used to indicate the first model and the second model.

17. The method according to any one of claims 13 to 16, characterized in that: The processing of the first data based on the first model and the second model to obtain an output signal includes: Inputting the first data into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors; The N1 high-dimensional vectors are processed by the first model to obtain N2 high-dimensional vectors; the first model processes the N1 high-dimensional vectors in a pre-trained manner; The N2 high-dimensional vectors are processed by the second model to obtain N3 high-dimensional vectors; the second model processes the N2 high-dimensional vectors in a pre-trained manner; The N3 high-dimensional vectors are input into the output layer of the neural network to obtain the output signal.

18. The method according to claim 17, characterized in that The first data is input into the input layer of the neural network for dimensionality increase processing to obtain N1 high-dimensional vectors, including: The first data corresponding to the M data streams are input into the input layer for dimensionality upgrading to obtain the N1 high-dimensional vectors; wherein one high-dimensional vector corresponds to one data stream.

19. The method according to claim 17 or 18, characterized in that The first data includes uncoded bits, coded bits or modulation symbols.

20. The method according to any one of claims 17 to 19, characterized in that: The output signal includes a modulation symbol or a waveform signal.

21. The method according to any one of claims 17 to 20, characterized in that: The first information further indicates whether the output signal includes a demodulation reference signal DMRS.

22. The method according to any one of claims 12 to 21, characterized in that: The first information further indicates the number of transmission streams of the output signal.

23. A communication device, characterized in that: The method comprises a unit for executing the method according to any one of claims 1 to 11, or comprises a unit for executing the method according to any one of claims 12 to 22.

24. A communication device, characterized in that: include: processor; The processor is used to execute the computer program or instruction in the memory, so that the device executes the method according to any one of claims 1 to 11, or the device executes the method according to any one of claims 12 to 22.

25. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when called by an electronic device, enable the electronic device to execute the method as claimed in any one of claims 1 to 11, or enable the electronic device to execute the method as claimed in any one of claims 12 to 22.

26. A computer program product, characterized in that The method comprises a computer program or an instruction, which, when executed on a computer, enables the computer to execute the method according to any one of claims 1 to 11, or enables the computer to execute the method according to any one of claims 12 to 22.

27. A chip system, characterized in that: The chip system comprises: Communication interface; A processor, used to call and run a computer program or instruction through the communication interface, so that a device equipped with the chip system executes the method as described in any one of claims 1 to 11, or so that a device equipped with the chip system executes the method as described in any one of claims 12 to 22.

28. A communication system, characterized in that: The chip system includes a second device for executing the method according to any one of claims 1 to 11 and a first device for executing the method according to any one of claims 12 to 22.

Images