WO2023060503A1 - 信息处理方法、装置、设备、介质、芯片、产品及程序 - Google Patents

信息处理方法、装置、设备、介质、芯片、产品及程序 Download PDF

Info

Publication number
WO2023060503A1
WO2023060503A1 PCT/CN2021/123777 CN2021123777W WO2023060503A1 WO 2023060503 A1 WO2023060503 A1 WO 2023060503A1 CN 2021123777 W CN2021123777 W CN 2021123777W WO 2023060503 A1 WO2023060503 A1 WO 2023060503A1
Authority
WO
WIPO (PCT)
Prior art keywords
information
processing
model
processing module
pilot
Prior art date
Application number
PCT/CN2021/123777
Other languages
English (en)
French (fr)
Inventor
肖寒
Original Assignee
Oppo广东移动通信有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Oppo广东移动通信有限公司 filed Critical Oppo广东移动通信有限公司
Priority to CN202180101608.1A priority Critical patent/CN117897931A/zh
Priority to PCT/CN2021/123777 priority patent/WO2023060503A1/zh
Publication of WO2023060503A1 publication Critical patent/WO2023060503A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines

Definitions

  • the embodiments of the present application relate to the field of communication technology, and specifically relate to an information processing method, device, equipment, medium, chip, product, and program.
  • AI artificial intelligence
  • Embodiments of the present application provide an information processing method, device, device, medium, chip, product, and program.
  • an embodiment of the present application provides an information processing method, the method comprising:
  • the first processing module of the information processing device receives the first information, uses the first model in the first processing module to process the first information to obtain second information, and sends the information to the second processing module of the information processing device said second information;
  • the second processing module processes the second information to obtain third information; wherein, the third information is estimated information of the first information;
  • the first processing module uses the first information and the third information to train the first model.
  • an information processing device comprising:
  • the first processing module is configured to receive the first information, use the first model in the first processing module to process the first information to obtain second information, and send the information to the second processing module of the information processing device. the second information;
  • a second processing module configured to process the second information to obtain third information; wherein the third information is estimated information of the first information;
  • the first processing module is further configured to use the first information and the third information to train the first model.
  • the embodiment of the present application provides an information processing device, including: a memory and a processor,
  • the memory stores a computer program executable on the processor
  • the above method is realized when the processor executes the program.
  • the embodiment of the present application provides a computer storage medium, where one or more programs are stored in the computer storage medium, and the one or more programs can be executed by one or more processors, so as to implement the foregoing method.
  • the embodiment of the present application provides a chip, including: a processor, configured to invoke and run a computer program from a memory, so that a device installed with the chip executes the above method.
  • the embodiment of the present application provides a computer program product
  • the computer program product includes a computer storage medium
  • the computer storage medium stores a computer program
  • the computer program includes instructions executable by at least one processor, when The instructions implement the above method when executed by the at least one processor.
  • the embodiment of the present application provides a computer program, the computer program causes a computer to execute the above method.
  • the first model in the first processing module processes the received first information to obtain the second information
  • the second processing module processes the second information to obtain the estimated information of the first information
  • the second A processing module uses the first information and the estimated information of the first information to train the first model, so that the information processing device can use the received first information and the estimated information of the first information determined by the first information
  • the first model in the information processing device is trained, so that the first information received during the use of the information processing device can realize the training of the first model, the training method is simple, and the trained model can accurately Subsequent processing of the input information avoids the situation in the related art that the information processing device cannot train the first model.
  • FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the present application
  • FIG. 2 is a schematic diagram of a communication flow of a wireless communication system provided by an embodiment of the present application
  • FIG. 3 is a schematic diagram of a channel estimation and recovery process in a wireless communication system provided by an embodiment of the present application
  • Fig. 4 is a schematic structural diagram of a neural network proposed by the related art
  • FIG. 5 is a schematic structural diagram of a convolutional neural network provided by related technologies
  • FIG. 6 is a schematic structural diagram of an autoencoder provided in the related art.
  • FIG. 7 is a schematic diagram of channel estimation and restoration using AI provided by an embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of a neural network-based channel feedback system provided in an embodiment of the present application.
  • FIG. 9 is a schematic flowchart of an information sending and receiving process provided by an embodiment of the present application.
  • FIG. 10 is a schematic flowchart of an information processing method provided in an embodiment of the present application.
  • FIG. 11 is a schematic diagram of a training framework of a first model provided in the embodiment of the present application.
  • FIG. 12 is a schematic diagram of a method for obtaining first information provided by an embodiment of the present application.
  • FIG. 13 is a schematic diagram of another way of obtaining first information provided by the embodiment of the present application.
  • FIG. 14 is a schematic diagram of another method for obtaining first information provided by the embodiment of the present application.
  • FIG. 15 is a schematic diagram of an update architecture in which a first model is a channel estimation model provided by an embodiment of the present application;
  • FIG. 16 is a schematic diagram of an updated architecture in which the first model is a decoder model provided by an embodiment of the present application;
  • FIG. 17 is a schematic diagram of an update architecture in which the first model is a receiver model provided by an embodiment of the present application.
  • FIG. 18 is a schematic diagram of the composition and structure of an information processing device provided by an embodiment of the present application.
  • FIG. 19 is a schematic structural diagram of an information processing device provided by an embodiment of the present application.
  • FIG. 20 is a schematic structural diagram of a chip according to an embodiment of the present application.
  • FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the present application.
  • a communication system 100 may include an information processing device 101 and an information sending device 102 .
  • the information sending device 102 may communicate with the information processing device 101 through an air interface. Multi-service transmission is supported between the information processing device 101 and the information sending device 102 .
  • the information processing device 101 may be called an information receiving device.
  • the information sending device 102 may also be referred to as another information processing device.
  • the information processing device 101 and the information sending device 102 may be in the same network or in different networks.
  • the embodiment of the present application is only described by using the communication system 100 as an example, but the embodiment of the present application is not limited thereto. That is to say, the technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Long Term Evolution (Long Term Evolution, LTE) system, LTE Time Division Duplex (Time Division Duplex, TDD), Universal Mobile Communication System (Universal Mobile Telecommunication System, UMTS), Internet of Things (Internet of Things, IoT) system, Narrow Band Internet of Things (NB-IoT) system, enhanced Machine-Type Communications (eMTC) system, 5G communication system (also known as New Radio (NR) communication system), or future communication systems (such as 6G, 7G communication systems), etc.
  • LTE Long Term Evolution
  • LTE Time Division Duplex Time Division Duplex
  • TDD Time Division Duplex
  • Universal Mobile Telecommunication System Universal Mobile Telecommunication System
  • UMTS Universal Mobile Communication System
  • Internet of Things Internet of Things
  • NB-IoT Narrow Band Internet of Things
  • the information processing device 101 and/or the information sending device 102 in the embodiment of the present application may be called User Equipment (User Equipment, UE), Mobile Station (Mobile Station, MS) or Mobile Terminal (Mobile Terminal, MT), etc.
  • the information processing device 101 and/or the information sending device 102 may include one or a combination of at least two of the following: server, mobile phone, tablet computer (Pad), computer with wireless transceiver function, palmtop computer, desktop computer , personal digital assistants, portable media players, smart speakers, navigation devices, smart watches, smart glasses, smart necklaces and other wearable devices, pedometers, digital TVs, virtual reality (Virtual Reality, VR) information processing equipment, enhanced Reality (Augmented Reality, AR) information processing equipment, wireless terminals in industrial control, wireless terminals in self driving, wireless terminals in remote medical surgery, smart grid wireless terminals in the grid, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, and vehicles in the Internet of Vehicles system, on-board equipment, Vehicle module, wireless modem
  • the information processing device 101 and/or the information sending device 102 may include one or a combination of at least two of the following: a base station in 6G, an evolved Base station (Evolutional Node B, eNB or eNodeB), next generation radio access network (Next Generation Radio Access Network, NG RAN) equipment, base station (gNB) in NR system, small station, micro station, cloud wireless access network ( Wireless controller in Cloud Radio Access Network (CRAN), wireless fidelity (Wireless-Fidelity, Wi-Fi) access point, transmission reception point (transmission reception point, TRP), relay station, access point, vehicle equipment, Wearable devices, hubs, switches, bridges, routers, information sending devices in the future evolution of the Public Land Mobile Network (PLMN), etc.
  • a base station in 6G an evolved Base station (Evolutional Node B, eNB or eNodeB), next generation radio access network (Next Generation Radio Access Network, NG RAN) equipment, base station (gNB) in NR system, small station, micro station,
  • the core network equipment can be 6G core network equipment or 5G core network (5G Core, 5GC) equipment, and the core network equipment can include one of the following or a combination of at least two: Access and Mobility Management Function (Access and Mobility Management Function, AMF), authentication server function (Authentication Server Function, AUSF), user plane function (User Plane Function, UPF), session management function (Session Management Function, SMF), location management function (Location Management Function, LMF).
  • the core information sending device may also be an Evolved Packet Core (EPC) device of the LTE network, for example, a data gateway (Session Management Function+Core Packet Gateway, SMF) of the session management function+core network +PGW-C) equipment.
  • EPC Evolved Packet Core
  • SMF+PGW-C can realize the functions of SMF and PGW-C at the same time.
  • the above-mentioned core network equipment may be called by other names, or a new network entity may be formed by dividing functions of the core network, which is not limited in this embodiment of the present application.
  • Various functional units in the communication system 100 may also establish a connection through a next generation network (next generation, NG) interface to implement communication.
  • NG next generation network
  • the information processing device establishes an air interface connection with the access network device through the NR interface for transmitting user plane data and control plane signaling; the information processing device can establish a control plane signaling connection with the AMF through the NG interface 1 (N1 for short); Access network devices such as next-generation wireless access base stations (gNB) can establish user plane data connections with UPF through NG interface 3 (referred to as N3); access network devices can establish control planes with AMF through NG interface 2 (referred to as N2) Signaling connection; UPF can establish a control plane signaling connection with SMF through NG interface 4 (referred to as N4); UPF can exchange user plane data with the data network through NG interface 6 (referred to as N6); AMF can communicate through NG interface 11 (referred to as N11) ) to establish a control plane signaling connection with the SMF; the SMF may establish a control plane signaling connection with the PCF through the NG interface 7 (N7 for short).
  • gNB next-generation wireless access base stations
  • N3 next-generation wireless
  • the embodiment of the present application does not limit the implementation manners of the information processing device 101 and the information sending device 102, and the information processing device 101 and the information sending device 102 may be any two devices capable of wireless communication.
  • FIG. 1 is only an illustration of a system applicable to this application, and of course, the method shown in the embodiment of this application may also be applicable to other systems.
  • system and “network” are often used interchangeably herein.
  • the term “and/or” in this article is just an association relationship describing associated objects, which means that there can be three relationships, for example, A and/or B can mean: A exists alone, A and B exist simultaneously, and there exists alone B these three situations.
  • the character "/" in this article generally indicates that the contextual objects are an "or” relationship.
  • the "indication” mentioned in the embodiments of the present application may be a direct indication, may also be an indirect indication, and may also mean that there is an association relationship.
  • A indicates B, which can mean that A directly indicates B, for example, B can be obtained through A; it can also indicate that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also indicate that there is an association between A and B relation.
  • the "correspondence" mentioned in the embodiments of the present application may mean that there is a direct correspondence or an indirect correspondence between the two, or that there is an association between the two, or that it indicates and is indicated. , configuration and configured relationship.
  • pre-defined may refer to defined in the protocol.
  • the "protocol” may refer to a standard protocol in the communication field, for example, it may include the LTE protocol, the NR protocol, and related protocols applied to future communication systems, and this application does not limit this .
  • FIG. 2 is a schematic diagram of a communication flow of a wireless communication system provided by an embodiment of the present application. As shown in FIG. are the information transmitting device 102 and the information processing device 101 described above.
  • the transmitter 201 performs channel coding and modulation on the source bit stream to obtain modulation symbols; inserts pilot symbols into the modulated symbols, and the inserted pilot symbols are used for channel estimation and symbol detection at the receiving end, and finally forms The signal is sent through the channel to the receiver. Wherein, the signal will be interfered by noise during the process of sending the signal to the receiving end through the channel.
  • the receiver 202 first receives the received signal, uses the pilot frequency in the received signal to perform channel estimation, and feeds back the channel state information (Channel State Information, CSI) to the sending end through the feedback link for the transmitter to adjust the channel Coding, modulation, precoding, etc. Finally, the receiver obtains the final restored bit stream through steps such as symbol detection, demodulation, and channel decoding.
  • CSI Channel State Information
  • Figure 2 is a simple illustration of the communication process of the wireless communication system.
  • modules not listed in the wireless communication system, such as resource mapping, precoding, interference cancellation, and CSI measurement. It is designed and implemented separately, and then each independent module can form a complete wireless communication system after integration.
  • the receiver's estimation and recovery of the wireless channel directly affects the final data recovery performance.
  • Figure 3 is a schematic diagram of the channel estimation and recovery process in a wireless communication system provided by the embodiment of the present application.
  • the signal sent by the transmitter on the time-frequency resource as shown in (a) will also send a series of specific pilot symbols known to the receiver (that is, reference signal symbols), such as Channel State Information-Reference Signal (CSI-RS) signal, demodulation reference signal (DeModulation Reference Signal , DMRS) signals, etc.
  • reference signal symbols such as Channel State Information-Reference Signal (CSI-RS) signal, demodulation reference signal (DeModulation Reference Signal , DMRS) signals, etc.
  • CSI-RS Channel State Information-Reference Signal
  • DMRS Demodulation Reference Signal
  • the signal sent by the transmitter is transmitted to the receiver through the channel.
  • the data symbols and reference signal symbols received by the receiver carry noise (that is, the data symbols and reference signal symbols carrying noise), and the receiver can carry noisy data symbols and reference signal symbols are used for channel estimation.
  • the receiver can use the least squares method (Least Squares, LS) or the minimum mean square error according to the real pilot and the received pilot (Minimum Mean Square Error, MMSE) and other methods estimate the channel information at the position of the reference signal. Then the receiver can perform channel recovery.
  • the receiver uses the interpolation algorithm to recover the channel information on the full time-frequency resource according to the channel information estimated at the pilot position, which is used for subsequent Channel information feedback or data recovery, etc.
  • the resource where the channel has been estimated/restored is the resource at the position of the reference signal symbol. It can be seen from (d) that the resources where the channel has been estimated/restored are the resources at the positions of the reference signal symbols and the data symbols.
  • the codebook-based scheme is mainly used to realize the extraction and feedback of channel features. That is, after the channel estimation is performed at the transmitting end, the precoding matrix that best matches the current channel is selected from the pre-set precoding codebook according to a certain optimization criterion according to the channel estimation result, and the matrix is transmitted through the feedback link of the air interface.
  • the index information of the precoding matrix indicator (Precoding Matrix Indicator, PMI) is fed back to the receiving end for the receiving end to implement precoding, and the channel quality indication (Channel Quality Indication, CQI) obtained according to the measurement is also fed back to the receiving end for the receiving end.
  • the receiving end implements adaptive modulation and coding, etc.
  • Fig. 4 is a schematic structural diagram of a neural network proposed by related technologies.
  • the structure of the neural network may include: an input layer, a hidden layer and an output layer.
  • the input layer is responsible for receiving data, hiding The layer processes the data, and the final result is generated in the output layer.
  • each node represents a processing unit, which can be regarded as simulating a neuron. Multiple neurons form a layer of neural network, and multi-layer information transmission and processing constructs an overall neural network.
  • neural network deep learning algorithms have been proposed in recent years, more hidden layers have been introduced, and feature learning is performed through layer-by-layer training of neural networks with multiple hidden layers, which greatly improves the learning of neural networks.
  • processing capabilities and are widely used in pattern recognition, signal processing, optimization combination, anomaly detection, etc.
  • CNN Convolutional Neural Networks
  • FIG. 5 is a schematic structural diagram of a convolutional neural network provided by related technologies.
  • the structure of the convolutional neural network may include: an input layer, multiple convolutional layers, multiple pooling layers, and a fully connected layer and the output layer.
  • the sharp increase of network parameters is effectively controlled, the number of parameters is limited, and the characteristics of local structures are mined, which improves the robustness of the algorithm.
  • Fig. 6 is a schematic structural diagram of a self-encoder provided in the related art.
  • the input of encoder 61 is image 62 to be compressed, and the output is code stream 63; decoder 64 inputs It is code stream 65, and the output is decompressed image 66.
  • the autoencoder is an end-to-end architecture with the same input and learning goals, that is, the image to be compressed can be used as the input (encoder input) and label (decoder output) of the autoencoder during training,
  • the encoder and decoder are jointly trained end-to-end.
  • FIG. 7 is a schematic diagram of channel estimation and recovery using AI provided by the embodiment of the present application.
  • the input information is the reference signal
  • the output information is the channel estimation result.
  • other auxiliary information can be added to improve the performance of the AI-based channel estimation and restoration module 71, for example, these Other auxiliary information may be information obtained by extracting features of the reference signal, energy levels, time delay features, noise features, and the like.
  • the neural network architecture commonly used in deep learning is non-linear and data-driven. It can extract features from the actual channel matrix data and restore the channel matrix information compressed and fed back by the UE as much as possible at the sending end. While ensuring the restoration of channel information, it also provides It is possible to reduce the CSI feedback overhead on the UE side.
  • AI-based CSI feedback regards channel information as an image to be compressed, uses deep learning autoencoder to compress and feed back channel information, and reconstructs the compressed channel image at the sending end, which can preserve channel information to a greater extent.
  • Fig. 8 is a schematic structural diagram of a neural network-based channel feedback system provided by the embodiment of the present application.
  • the channel feedback system is divided into an encoder and a decoder, which are respectively deployed at the sending end and the receiving end.
  • the sender obtains the channel information through channel estimation, it compresses and encodes the channel information matrix through the neural network of the encoder to obtain an M ⁇ 1-dimensional vector.
  • the compressed bit stream (that is, the M ⁇ 1 dimensional vector) is fed back to the receiving end through the air interface feedback link, and the receiving end restores the channel information through the decoder according to the feedback bit stream to obtain complete feedback channel information.
  • the structure shown in Figure 8 uses several fully connected layers for encoding at the encoder, and uses a residual network structure (such as RefineNet) for decoding at the decoder.
  • a residual network structure such as RefineNet
  • the network model structure inside the encoder and decoder can be flexibly designed.
  • FIG. 9 is a schematic flow diagram of an information sending and receiving process provided by an embodiment of the present application.
  • a neural network can be used to directly replace the signal processing flow of a traditional receiver.
  • the input of the end-to-end AI receiver is the signal received by the receiving end, and the output is the restored bit stream.
  • the network model structure inside the AI receiver can be flexibly designed.
  • the source bit stream undergoes a series of operations such as encoding and adjustment to obtain a sending signal, which is sent to the receiving end through a channel. During the sending process, it will be interfered by noise.
  • the receiving end may send the received signal to the AI receiver, so that the AI receiver outputs a restored bit stream.
  • Reinforcement learning is developed from theories of animal learning and parameter perturbation adaptive control. Its basic principle is: if a certain behavioral strategy of the agent leads to positive rewards in the environment, then the tendency of the agent to produce this behavioral strategy will be strengthened in the future. . The agent's goal is to discover the optimal policy at each discrete state to maximize the sum of expected rewards. Reinforcement learning regards learning as a trial and evaluation process.
  • the agent chooses an action to use in the environment. After the environment receives the action, the state changes, and at the same time generates a reinforcement signal (reward or punishment) to feed back to the agent. According to the reinforcement signal and the environment, the agent In the current state, choose the next action.
  • the principle of selection is to increase the probability of receiving positive reinforcement (reward).
  • the selected action not only affects the immediate strengthening value, but also affects the state of the environment at the next moment and the final strengthening value.
  • the channel modeling based on traditional communication system design can no longer describe the increasingly complex channel environment brought about by large-scale antennas, underwater communications, millimeter waves, etc.; at the same time, the increasingly diverse signal processing devices Combination utilization also brings certain nonlinear characteristics to the signal processing process.
  • Traditional signal processing methods based on mathematical modeling are increasingly unable to better meet the current high-reliability communication requirements; and traditional communication systems such as symbol detection and other iterative The algorithm also has high complexity, and it cannot meet the current high-speed communication requirements well.
  • AI-based end-to-end transmitter and receiver design can solve the above-mentioned problems of traditional communication systems to a certain extent.
  • this end-to-end design is often purely data-driven, requiring a large amount of data and a long time for training and construction, and the acquisition of massive data sets and long-term training are difficult in current wireless communication systems .
  • the design of the AI-based modular communication system proposed in the embodiment of this application considers the dual drive of data and model, which can make good use of the prior structure of the original communication system model, and can flexibly adjust and train each module, for example Independent neural network modules are designed for channel estimation, modulation coding, etc., which can greatly reduce the need for large data sets and long training time.
  • the encoder and decoder structures of autoencoders are naturally compatible with many architectures in communication systems, so there have been a series of studies on the application of autoencoders in communication systems.
  • the encoder and the decoder may respectively correspond to the sending end and the receiving end of the overall communication system, and may also respectively correspond to the channel compression module and the decompression module of the CSI feedback problem.
  • this is only a structural adaptation.
  • the existence of the air interface in the actual communication system makes it difficult for the training data to be synchronized between the sending end and the receiving end when the encoder and the decoder are deployed in a distributed manner at the sending end and the receiving end, and it is difficult to accurately transmit the gradient through the air interface, making this structure only online It is difficult to achieve online self-update after pre-training and deploying online.
  • the inherent disadvantage of the generalization problem of the neural network itself in practical applications makes the trained network only suitable for application scenarios with the same characteristics as the training set, that is, the training set is often difficult to cover all situations. When the scene characteristics change, the well-trained It is difficult for the model to continue to maintain good generalization performance.
  • FIG. 10 is a schematic flowchart of an information processing method provided in the embodiment of the present application. As shown in FIG. 10, the method is applied to an information processing device, and the method includes:
  • the first processing module of the information processing device receives first information, uses the first model in the first processing module to process the first information to obtain second information, and sends the information to the second processing module of the information processing device The module sends the second information.
  • the first information may be information received by the information processing device from the air interface, or the first information may be information obtained by processing the information received from the air interface by the information processing device.
  • the first processing device may input the first information to the first model in the first processing device, process the first information through the first model, and obtain and output the second information.
  • the first model in this embodiment of the present application may be an AI model.
  • the first information may be a matrix or one or more vectors.
  • the second information may be a matrix or one or more vectors.
  • the first processing module may send the second information to other processing modules in the information processing device, so that other processing modules process the information processing device.
  • the first processing module When it is determined that the internal first model needs to be updated, the first processing module not only sends the second information to other processing modules, but also sends the second information to the second processing module; In the case of updating the first model of , the first processing module only sends the second information to other processing modules.
  • the second processing module processes the second information to obtain third information; wherein the third information is estimated information of the first information.
  • the first information may be a reference signal (also referred to as a first pilot information set) sent by an information processing device at a predefined position
  • the first model may be a channel estimation recovery model (also referred to as a channel estimation recovery model). network)
  • the second information may be estimated channel state information (also called the first channel estimation matrix).
  • the third information may be estimation information of the first pilot information set.
  • the other modules may be the first encoding module and/or the symbol detection module.
  • the first encoding module may encode the estimated channel state information and feed it back to the information sending device.
  • the symbol detection module is used to detect the reference signal.
  • the first information includes a second set of bitstream vectors; the second information includes a second channel estimation matrix.
  • the first model may be a decoder model.
  • the third information includes estimation information of the second set of bitstream vectors.
  • the second processing module may be a second encoding module.
  • the other modules may be a processor and/or a channel coding module and/or a modulation module, the processor can obtain the second channel estimation matrix, and the channel coding module and/or the modulation module are used to estimate based on the second channel matrix for encoding and/or modulation.
  • the first information may include the received second pilot information set and the received first data information set; the second information includes information source estimation.
  • the first model may be a receiver model.
  • the third information includes estimation information of the second pilot information set and estimation information of the first data information set.
  • the second processing module may be a transmitter. In this case, the other module may be a processor capable of obtaining estimated information of the source.
  • the first processing module uses the first information and the third information to train the first model.
  • the first information and the third information may be used to determine the weight of each sample in each training sample, and the first model is trained based on the weight of each sample.
  • the first model may be trained based on each information included in the first information, each information included in the third information, and weights of each sample.
  • the weight of the sample may be the weight of each piece of information included in the first information during training. There can be one-to-one correspondence between each sample and each piece of information.
  • the purpose of the training may include making the degree of difference between the first information and the third information smaller than a target value.
  • training the first model based on the first information and the third information may include: training the first model based on the first information, second information and third information .
  • the first information and the second information may be determined as the training samples, and the first model is trained based on the training samples and the weights of the samples.
  • the weight of each sample is used to represent the importance of each sample in the training process. For example, the greater the weight of a sample, the higher the importance of the sample in model training.
  • the purpose of the training may include making the degree of difference between the second information and the information obtained by processing the third information through the first model smaller than a target value.
  • the first model in the first processing module processes the received first information to obtain the second information; the second processing module processes the second information to obtain the estimated information of the first information;
  • the first information and the estimated information of the first information train the first model, so that the information processing device can use the received first information and the estimated information of the first information determined by the first information to train the information processing device
  • the first model in the system is trained, so that the first information received during the use of the information processing device can realize the training of the first model, the training method is simple, and the trained model can accurately understand the subsequent input information
  • the processing is performed to avoid the situation that the information processing equipment cannot train the first model in the related art.
  • the processing the second information to obtain third information includes:
  • the first processing method and/or the first processing parameter are the same as the second processing method and/or the second processing parameter adopted by the target processing module in the information sending device for the received fourth information;
  • the target processing module processes the fourth information to obtain fifth information;
  • the second information is estimated information of the fourth information; the first information is obtained by transmitting the fifth information through an air interface.
  • the fifth information may be processed by one or more modules of the information sending device, and transmitted to the information processing device through the air interface, and processed by one or more modules of the information processing module to obtain the first information.
  • the operations performed by the target processing module and the first processing module are opposite.
  • the target processing module can process X information to obtain Y information
  • the first processing module can process Y information to obtain X information.
  • the first information may be information obtained by adding noise to the fifth information.
  • the format of the first information and the fifth information may be the same, for example, both are information obtained by encoding, or both are obtained by adding pilots.
  • the fourth information may be the first set of channel matrices received or acquired by the target processing module of the information sending device, and the second information may be the estimated first set of channel matrices (ie, the first channel estimation matrix).
  • the transmitted pilot signal can be obtained based on the pilot sequence set and the transmitted first channel matrix set
  • the information transmitting device transmits the pilot signal
  • the information processing device receives the pilot signal (that is, the received first pilot information set), restore the first channel estimation matrix through the channel estimation recovery model, and determine the estimated information of the first pilot information set based on the first channel estimation matrix and the pilot sequence set.
  • an implementation manner of obtaining the transmitted pilot signal based on the pilot sequence set and the transmitted first channel matrix set may be: for each element in the pilot sequence set and the transmitted first channel matrix set Each element of is multiplied correspondingly to obtain the transmitted pilot signal.
  • an implementation manner of determining the estimated information of the first pilot information set may be: for each element in the first channel estimation matrix and each element in the pilot sequence set The elements are multiplied correspondingly to obtain the estimated information of the first pilot information set.
  • the fourth information may be the second channel matrix set received or acquired by the target processing module of the information sending device, and the second information may be the estimated second channel matrix set.
  • the information sending device can obtain the second set of channel matrices, process the second set of channel matrices through an encoder or an encoder model, and obtain a set of transmitted bit stream vectors, and the information processing device receives the set of bit stream vectors (that is, the second bitstream vector set), the second bitstream vector set is decoded by the decoder model to obtain the estimated second channel matrix set (ie, the second channel estimation matrix), and the second channel estimation matrix is performed by the encoder or the encoder model After encoding, an estimated second set of bitstream vectors (that is, estimated information of the second set of bitstream vectors) is obtained.
  • the encoding parameters of the encoder or encoder model of the information sending device may be the same as the encoding parameters of the encoder or encoder model of the information processing device.
  • the fourth information may be information source information received or acquired by the target processing module of the information sending device, and the second information may be estimated information source information.
  • the information sending device can obtain the information source, use the transmitter or the transmitter model to process the information source, and obtain the transmitted pilot information set and the data information set, and the information processing equipment receives the pilot information set and the data information set (ie After receiving the second pilot information set and the received first data information set), the AI receiver processes the second pilot information set and the first data information set to obtain the estimated information source (that is, the estimated information of the information source ), the transmitter or transmitter model of the information processing device can process the estimated information of the information source to obtain an estimated pilot information set and an estimated data information set.
  • the channel estimation module in the information processing device can process the received second pilot information set and the received first data information set to obtain a third channel estimation matrix, based on the estimated pilot information set and the estimated data The information set, and the third channel estimation matrix, obtain estimated information for the received pilot information set and the received data information set.
  • the transmission parameters of the transmitter or the transmitter model of the information sending device may be the same as the transmission parameters of the transmitter or the transmitter model of the information processing device.
  • estimate information on the received pilot information set and data information set is obtained, which may include: an estimated pilot information set and an estimated data information set
  • Each element in the formed matrix is multiplied correspondingly with each element in the third channel estimation matrix to obtain estimated information on the received pilot information set and the received data information set.
  • the embodiment of the present application adopts the reinforcement learning method, and proposes a first model update scheme based on reinforcement learning and its communication application design.
  • Fig. 11 is a schematic diagram of a training framework of a first model provided by the embodiment of the present application.
  • the original signal First after processing and sending through the processing module on the side of the information sending device (that is, the above-mentioned target processing module), the output (first information), the information processing device receives the signal And output the restored original signal through the first model (second information).
  • a device for processing (such as an encoder) with the same structure and parameters constitutes a reinforcement learning problem.
  • a processing module such as an encoder
  • a reinforcement learning problem constitutes a reinforcement learning problem.
  • the signal After being processed by the first model g on the side of the information processing device, the signal is obtained further signal Send it to the processing module f deployed on the information processing device side for processing, and obtain the processed signal (third information), the processed signal as states in reinforcement learning problems;
  • the parameters of the processing module deployed on the information processing device side can be frozen, and the first model can be trained and updated online by using a reinforcement learning algorithm such as the policy gradient method.
  • using the first information and the third information to train the first model may include:
  • the first model is trained by using the cosine similarity and/or the mean square error.
  • the cosine similarity and/or the mean square error may be used to characterize the degree of difference between the first information and the third information.
  • the weight of each sample in the training samples may be determined based on the cosine similarity and/or the mean square error, and the first model may be trained based on the weight of each sample.
  • training the first model based on the weight of each sample may include: acquiring each sample in the training samples, and training the first model based on each sample in the training sample and the weight of each sample.
  • the first information includes one or more sub-information; the method further includes:
  • the first processing module acquires first indication information; the first indication information indicates that the first model is trained once in each training period; each training period includes one or more transmission periods;
  • the first processing module determines at least part of the information received in each transmission cycle included in each training cycle as sub-information, and obtains the one or more sub-information; the one or more sub-information and the one or multiple transmission cycles in one-to-one correspondence.
  • the information processing device When the information processing device acquires the first indication information, enter into the step of determining the first information.
  • the information processing device can determine a piece of first information in each training cycle.
  • the information processing device may receive data in each transmission period, and determine the data received in each transmission period as information corresponding to each transmission period. At least part of the information corresponding to each transmission period may be part or all of the information corresponding to each transmission period.
  • the first indication information also indicates at least one of the following:
  • the training period the size of the first information, the size of each of the sub-information, the ratio of the size of each of the sub-information to the size of the information acquired in each transmission cycle.
  • the first indication information may further include: a start time for starting to acquire the first data. At this start time, the information processing device acquires at least part of the information corresponding to each transmission cycle, so as to continuously obtain the first information. Exemplarily, the offset between the start time for acquiring the first data and the time slot where the indication signaling is located may be used as the indicated start time.
  • the information processing device may stop training the first model and collect sub-information when the number of training times reaches the preset number of times.
  • the information processing device may receive fourth instruction information sent by the network device, and the fourth instruction information may indicate to stop training the first model, so that the information processing device may no longer perform the first model training based on the fourth instruction information. training, and no longer collect sub-information.
  • FIG. 12 is a schematic diagram of a method for acquiring first information provided by an embodiment of the present application.
  • the transmitting end sends signals to the receiving end one or more times.
  • This transmission can be divided into online learning transmission and traditional transmission.
  • the first model needs online learning fine-tuning before inference, it is defined as online learning transmission (transmission period t); when the first model only performs inference while collecting data
  • the process is defined as legacy transfer. Defined as the training period r.
  • Each training period r can contain several transmission periods t, wherein, when the transmission period is traditional transmission, the information processing equipment receives the signal and stores the signal locally to realize online training data collection, and at the same time for the signal transmitted this time Using the first model to perform inference; when the transmission cycle is online learning transmission, the information processing device can use the collected online training data to perform online training of the first model, and use the trained first model to perform inference.
  • related parameters may be indicated by radio resource control (Radio Resource Control, RRC) signaling or control channel.
  • the relevant parameters include at least one of the following: training period, online training data set size. Among them, the size of the online training data set means that not all the collected data is necessarily used for each online training, but a subset of the training data set is taken as the online training data set.
  • the first information includes one or more sub-information; the method further includes:
  • the first processing module acquires second indication information; the second indication information indicates that the first model is trained once in each training period in N training periods; the training period includes one or more transmission periods; Said N is an integer greater than or equal to 1;
  • the first processing module determines at least part of the information received in each transmission cycle included in each training cycle as sub-information, and obtains the one or more sub-information; the one or more sub-information and the one or multiple transmission cycles in one-to-one correspondence.
  • the information processing device When the information processing device acquires the second indication information, enter into the step of determining the first information.
  • the information processing device determines a piece of first information in each training period of the N training periods.
  • the second indication information also indicates at least one of the following:
  • the start time of the N training periods, the training period, the N the size of the first information, the size of each of the sub-information, the size of each of the sub-information and the size of each of the sub-information.
  • the start time of the N training periods and the offset of the time slot where the indication signaling is located may be used as the start time of the indication.
  • the information processing device trains the first model once in each training period, and continuously trains N times indicated by the second indication information.
  • the information processing device no longer performs training of the first model, and no longer collects sub-information.
  • Figure 13 is a schematic diagram of another way to obtain the first information provided by the embodiment of the present application.
  • signaling that is, the second indication information
  • relevant parameters can be indicated by RRC signaling or control channel.
  • the parameters include at least one of the following: starting point of online training, online training period, total number of online training periods, and online training data set size.
  • the offset between the starting point of the online training and the time slot where the indication signaling is located can be used as the indication information of the starting point of the online training; the total number of cycles of the online training determines the end point of the online training period.
  • the first information includes one or more sub-information; the method further includes:
  • the first processing module acquires third indication information; the third indication information indicates to train the first model once;
  • the first processing module determines at least part of the information received in each of the one or more transmission cycles as sub-information, and obtains the one or more sub-information; the one or more sub-information and the one or more Multiple transmission cycles are in one-to-one correspondence.
  • the third indication information also indicates at least one of the following:
  • the start time of the one or more transmission periods, the number of periods of the transmission period, the size of the first information, the size of each of the sub-information, the size of each of the sub-information and the The ratio of the size of the information acquired in each transmission cycle is described.
  • the offset between the start time of one or more transmission cycles and the time slot where the indication signaling is located can be used as the start time of one or more transmission cycles indicated; the total number of online training cycles determines the end of the online training cycle point.
  • FIG. 14 is a schematic diagram of another method for acquiring first information provided by the embodiment of the present application.
  • signaling (third indication information) is used to indicate online learning transmission each time.
  • the parameters indicated by the signaling at least include: the size of the online training data set.
  • the content indicated by the indication information (including at least one of the first indication information, the second indication information, and the third indication information) can be flexibly configured according to the actual transmission environment, delay and complexity requirements.
  • the parameter value indicated by the indication information may be modified.
  • the embodiment of the present application can improve the overall adaptability of the AI-based communication system to the environment, and the higher the configuration ratio of the online learning transmission, the stronger the adaptability of the first model.
  • the information processing device when the instruction information is received, the information processing device first obtains one or more sub-information, determines the one or more sub-information as the first information, and then executes the training of the first model. Afterwards, the time corresponding to the first training of the first model is separated from the received time by at least one transmission cycle.
  • the information processing device may determine the obtained sub-information of the first transmission cycle as the first information, and use the first information to execute the first information of the first model. Training, and then after the first transmission cycle, the first information is acquired every training cycle, and the first model training is performed according to the first information obtained each time.
  • the information processing device may acquire several pieces of sub-information from the obtained sub-information of the first subsequent transmission cycle and information corresponding to at least one previous transmission cycle, so that All sub-information corresponding to one training cycle is determined, and all sub-information is determined as the first first information, and the first training of the first model is performed.
  • the indication information may be indication information pre-stored in the information processing device.
  • an information processing device may acquire indication information from itself.
  • the indication information may be indicated by another information processing device, for example, when the information processing device is a terminal device, the indication information may be indicated by a network device.
  • the network device may use RRC signaling or a control channel indication to indicate the indication information.
  • the information processing device or another information processing device may acquire the indication information when it is determined that the error between the estimated information of the first information and the first information is relatively large.
  • using the first information and the third information to train the first model includes: using the first information and the third information to determine the first mean square error of each sample; performing normalization processing on the first mean square error to obtain a first normalization result; determining the first information and the second information as the training samples, The first normalization result is determined as a first weight of each sample; and the first model is trained by using the training sample and the first weight.
  • the first information can be The third information can be Based on the first information B and the third information B', calculate the first mean square error of each sample in the corresponding training sample For the first mean square error Perform normalization to obtain the first normalization result.
  • the normalization step is to first calculate The mean e and standard deviation v of m1,...,mn included in , and then the first normalized result obtained is: the first mean square error After each element in is subtracted from the mean e, each element obtained is divided by the result of the standard deviation v.
  • the first normalization result can be the first weight of each sample, using express.
  • using the first information and the third information to train the first model includes: using the first information and the third information to determine The first standard mean square error of each sample; normalize the first standard mean square error to obtain a second normalization result; determine the first information and the second information as the training samples, the second normalization result is determined as the second weight of each sample; the first model is trained by using the training samples and the second weight.
  • the first information can be The third information can be Based on the first information B and the third information B', calculate the first standard mean square error of each sample in the corresponding training sample For s is the length of the bitstream, for example, the value of s can be the same as dimension related.
  • first standard mean square error Perform normalization to obtain the second normalization result.
  • the normalization step is to first calculate The mean e and standard deviation v of m1,...,mn included in , and then the second normalization result obtained is: the first mean square error After each element in is subtracted from the mean e, each element obtained is divided by the result of the standard deviation v.
  • the second normalization result can be the first weight of each sample, using express.
  • the training of the first model by using the first information and the third information includes: dequantizing the first information to obtain a first solution A quantization result; performing dequantization processing on the third information to obtain a second dequantization result; using the first dequantization result and the second dequantization result to train the first model.
  • using the first dequantization result and the second dequantization result to train the first model includes: using the first dequantization result and the second dequantization result Quantize the result, determine the second mean square error or the second standard mean square error of each sample in the corresponding training sample; carry out normalization processing to the second mean square error or the second standard mean square error to obtain the third A normalized result; determining the first information and the second information as the training samples, and determining the third normalized result as the third weight of each sample; using the training samples and the The third weight is used to train the first model.
  • the first information can be The third information can be Dequantization processing is performed on the first information, and the obtained first dequantization result may be performing dequantization processing on the third information, and the obtained second dequantization result may be deQ( ⁇ ) is dequantization calculation.
  • the second mean square error of each sample in the corresponding training sample as an example: based on Z and Z', each of the corresponding training samples determined
  • Sample second mean square error For the second mean square error Perform normalization to obtain the third normalization result.
  • the normalization step is to first calculate The mean e and standard deviation v of m1,...,mn included in , and then the third normalized result obtained is: the second mean square error After each element in is subtracted from the mean e, each element obtained is divided by the result of the standard deviation v.
  • the third normalized result can be the first weight of each sample, using express.
  • the first model includes a channel estimation model; the first information includes a first set of pilot information received; the second information includes a first channel estimation matrix; the third information includes The estimation information of the first pilot information set; the processing the second information to obtain the third information includes: combining each element in the first channel estimation matrix with each element in the pilot sequence matrix The elements are multiplied correspondingly to obtain the estimated information of the first pilot information set.
  • the first model is a channel estimation model: the sending end (information sending device) of the communication system can send a pilot sequence set P, and the pilot sequence set can be called a pilot vector, a pilot symbol vector, or a pilot information set , pilot matrix, etc.
  • H is channel information (or channel matrix)
  • N is additive noise.
  • H.*P means that each element in the matrix H is multiplied by each element in the matrix P.
  • the input of the channel estimation model is YP
  • the output is the estimation information of H (ie, the first channel estimation matrix).
  • the input of the channel estimation model is YP and the pilot sequence set P, and the output is the estimation information of H (ie, the first channel estimation matrix).
  • Fig. 15 is a schematic diagram of an update architecture of a first model provided by an embodiment of the present application as a channel estimation model.
  • the channel information to the first set of pilot information The process of is regarded as the processing process of the processing module in the information sending device at the sending end (for example, the encoding process of the encoder), and will convert from the first pilot information set YP to the first channel estimation matrix
  • the process is regarded as the processing process of the processing module in the information processing device at the receiving end (for example, the decoding process of the decoder), and the channel estimation model can be updated at the receiving end, that is, the information processing device side.
  • the information processing device After receiving the first set of pilot information YP, the information processing device uses the channel estimation model to perform channel estimation to obtain a first channel estimation matrix H'.
  • the estimation information YP' of the first pilot information set YP is generated by using the first channel estimation matrix H' and the pilot sequence set P.
  • each element in the first channel estimation matrix H' may be correspondingly multiplied by each element in the pilot sequence set P to obtain the estimated information YP' of the first pilot information set YP.
  • the first set of pilot information YP may be the first information
  • the first channel estimation matrix may be the second information
  • the estimation information YP' of the first set of pilot information YP may be the third information.
  • the channel information H may be fourth information.
  • the processing on the channel information H is H.*P
  • the processing on the first channel estimation matrix H' is also H'.*P.
  • the training samples can be can be N samples in the training samples, respectively.
  • the channel estimation model is trained by using the obtained training sample D and the above obtained weight m (including the first weight, the second weight or the third weight).
  • the training stops, and the condition is satisfied including but not limited to the number of training times reaching the set maximum number of iterations.
  • the method further includes: the first processing module sending the first channel estimation matrix to a first encoding module of the information processing device;
  • the estimated matrix is encoded to obtain a first set of bitstream vectors;
  • the first encoding module sends the first set of bitstream vectors to the information sending device.
  • the first encoding module may send the first set of bitstream vectors to the transceiver module of the information processing module, so that the transceiver module sends the first set of bitstream vectors to the information sending device.
  • the first channel estimation matrix may be encoded by an encoder or an encoder model to obtain a first set of bitstream vectors.
  • the information sending device can input the received bitstream information vectors into the decoder model, so that the decoder model can understand the received bitstream information vectors Processing to obtain a specific channel estimation matrix, where the specific channel estimation matrix is estimation information for the first channel estimation matrix.
  • the information sending device can also train the decoder model in the information sending device, and the training method for the decoder model in the information sending device can be compared with the information processing device for the decoder in the information processing device The model is trained in the same way.
  • the first model includes a decoder model; the second processing module includes a second encoding module; the first information includes a second set of bitstream vectors; the second information includes a second channel an estimation matrix; the third information includes estimation information of the second set of bitstream vectors; the second processing module processes the second information to obtain third information, including: the second encoding module processes the performing encoding processing on the second channel estimation matrix to obtain estimation information of the second bitstream vector set.
  • Fig. 16 is a schematic diagram of an updated architecture of a first model provided by the embodiment of the present application as a decoder model.
  • the sending end (information sending device) of the communication system can use the encoder or the encoder model to convert the channel gather
  • Each channel in the channel is compressed and coded into a bit stream vector to form a set of bit stream vectors, which is sent to the receiving end (information processing device) through a feedback link.
  • the information processing device collects the received second bitstream vectors through the decoder model Each bitstream in is restored to the corresponding second channel estimation matrix Using the encoder or encoder model on the information processing device side will Each channel in is re-encoded into the corresponding bitstream, and the estimated information of the second set of bitstream vectors is obtained
  • the encoding parameters of the encoder or encoder model at the information processing device end are the same as the encoding parameters of the encoder or encoder model at the information sending device end.
  • the second set of bitstream vectors Can be the first information
  • the second channel estimation matrix Can be the second information
  • estimated information of the second bitstream vector set It may be third information.
  • channel set It may be fourth information.
  • the training samples can be can be N samples in the training samples, respectively.
  • the decoder model is trained by using the obtained training sample D and the above obtained weight m (including the first weight, the second weight or the third weight).
  • the training stops, and the condition is satisfied including but not limited to the number of training times reaching the set maximum number of iterations.
  • the first model includes a receiver model; the second processing module includes a transmitter (or a transmitter model, the transmitter does not include a model that needs to be trained in this embodiment of the application); the The first information includes the received second pilot information set and the received first data information set; the second information includes information source estimation; the third information includes the second pilot information set Estimated information and estimated information of the first data information set; the second processing module processes the second information to obtain third information, including: the transmitter uses a transmitter for the estimated information of the information source Perform processing to obtain a processing result; the processing result is used to: determine the estimated information of the second pilot information set and the estimated information of the first data information set.
  • the processing result includes a transmission information estimation matrix;
  • the transmission information estimation matrix includes: the estimation information of the third pilot information sent by the information sending device and the estimation information of the second data information set sent ;
  • the method also includes:
  • the channel estimation module of the information processing device performs channel estimation processing on the second pilot information set and the first data information set to obtain a third channel estimation matrix; the third processing module of the information processing device converts the Each element in the transmission information estimation matrix is correspondingly multiplied by each element in the third channel estimation matrix to obtain the estimation information of the second pilot information set and the estimation information of the first data information set .
  • FIG. 17 is a schematic diagram of an updated architecture of a first model provided by an embodiment of the present application, which is a receiver model.
  • source bitstream vector Perform processing to obtain the symbol matrix after inserting the pilot Among them, the transmitter or the transmitter model can be used for the source bit stream vector Perform encoding and modulation processing.
  • the symbol matrix x is sent to the receiving end through the air interface, and the signal received by the receiving end is h.
  • *x represents the multiplication of each element in matrix h and each element in matrix x; matrix h is the channel matrix, N is additive noise; yp is the received data matrix, and yd is the received symbol matrix.
  • the information processing equipment at the receiving end aggregates the received signals (that is, the received second pilot information set and the received first data information set) are sent to the receiver or receiver model, and the bit stream vector set is restored by the receiver or receiver model (that is, the estimated information of the source), on the other hand, the channel set is estimated by the channel estimation module (ie the third channel estimation matrix).
  • the channel estimation module may be a channel estimation module of a traditional communication system, or may be an AI-based channel estimation module.
  • the information processing device can restore the bitstream vector set Each bit in is sent to the transmitter or transmitter model of the information processing device for processing to obtain the transmitted information estimation matrix.
  • the estimated matrix of the sent information and the estimated channel set Each element in is multiplied correspondingly to obtain the reconstructed received signal set (that is, the estimation information of the second pilot information set and the estimation information of the first data information set).
  • the transmission parameters of the transmitter or transmitter model at the information processing device end are the same as those of the transmitter or transmitter model at the information sending device end.
  • the received second set of pilot information and the received first set of data information may be the first information
  • the estimated information of the information source may be the second information
  • the estimation information of the first data information set may be third information.
  • Multiple Source Bitstream Vectors The formed set may be fourth information.
  • the training samples can be can be N samples in the training samples, respectively.
  • the receiver model is trained by using the obtained training sample D and the above obtained weight m (including the first weight, the second weight or the third weight).
  • the training stops, and the condition is satisfied including but not limited to the number of training times reaching the set maximum number of iterations.
  • the term "and/or" is only an association relationship describing associated objects, indicating that there may be three relationships. Specifically, A and/or B may mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" in this article generally indicates that the contextual objects are an "or" relationship.
  • FIG. 18 is a schematic diagram of the composition and structure of an information processing device provided by an embodiment of the present application.
  • the information processing device 1800 can be applied to the above-mentioned information processing equipment. As shown in FIG. 18 , the information processing device 1800 includes:
  • the first processing module 1801 is configured to receive the first information, use the first model in the first processing module to process the first information to obtain second information, and send it to the second processing module of the information processing device said second information;
  • the second processing module 1802 is configured to process the second information to obtain third information; wherein, the third information is estimated information of the first information;
  • the first processing module 1801 is further configured to use the first information and the third information to train the first model.
  • the second processing module 1802 is further configured to process the second information using a first processing method and/or a first processing parameter to obtain the third information;
  • the first processing method and/or the first processing parameter are the same as the second processing method and/or the second processing parameter adopted by the target processing module in the information sending device for the received fourth information;
  • the target processing module processes the fourth information to obtain fifth information;
  • the second information is estimated information of the fourth information; the first information is obtained by transmitting the fifth information through an air interface.
  • the first processing module 1801 is further configured to:
  • the first model is trained by using the cosine similarity and/or the mean square error.
  • the first information includes one or more sub-information; the first processing module 1801 further includes:
  • the first indication information indicates that the first model is trained once in each training period; each training period includes one or more transmission periods;
  • the first indication information also indicates at least one of the following:
  • the training period the size of the first information, the size of each of the sub-information, the ratio of the size of each of the sub-information to the size of the information acquired in each transmission cycle.
  • the first information includes one or more sub-information; the first processing module 1801 is further configured to:
  • the second instruction information indicates that the first model is trained once in each training cycle of N training cycles; the training cycle includes one or more transmission cycles; the N is greater than or equal to an integer of 1;
  • the second indication information also indicates at least one of the following:
  • the start time of the N training periods, the training period, the N the size of the first information, the size of each of the sub-information, the size of each of the sub-information and the size of each of the sub-information.
  • the first information includes one or more sub-information; the first processing module 1801 is further configured to:
  • the third instruction information indicates to train the first model once
  • the third indication information also indicates at least one of the following:
  • the start time of the one or more transmission periods, the number of periods of the transmission period, the size of the first information, the size of each of the sub-information, the size of each of the sub-information and the The ratio of the size of the information acquired in each transmission cycle is described.
  • the first processing module 1801 is further configured to:
  • the first model is trained by using the training samples and the first weights.
  • the first processing module 1801 is further configured to:
  • the first model is trained by using the training samples and the second weights.
  • the first processing module 1801 is further configured to:
  • the first model is trained by using the first dequantization result and the second dequantization result.
  • the first processing module 1801 is further configured to:
  • the first model is trained by using the training samples and the third weight.
  • said first model comprises a channel estimation model
  • the first information includes a first set of received pilot information
  • the second information includes a first channel estimation matrix
  • the third information includes estimation information of the first set of pilot information
  • the second processing module 1802 is further configured to: multiply each element in the first channel estimation matrix by each element in the pilot sequence matrix to obtain the estimated information of the first pilot information set .
  • the device also includes:
  • a first encoding module configured to encode the first channel estimation matrix to obtain a first set of bitstream vectors
  • the first coding module is further configured to send the first set of bitstream vectors to an information sending device.
  • said first model comprises a decoder model
  • said second processing module comprises a second encoding module
  • said first information includes a second set of bitstream vectors
  • the second information includes a second channel estimation matrix
  • said third information includes estimation information for said second set of bitstream vectors
  • the second encoding module is configured to perform encoding processing on the second channel estimation matrix to obtain estimation information of the second bitstream vector set.
  • said first model comprises a receiver model
  • said second processing module comprises a transmitter
  • the first information includes the received second set of pilot information and the received first set of data information
  • the second information includes estimated information of a source
  • the third information includes estimation information of the second pilot information set and estimation information of the first data information set;
  • the transmitter is configured to process the estimated information of the information source to obtain a processing result; the processing result is used to: determine the estimated information of the second pilot information set and the first data information set estimated information.
  • the processing result includes a transmission information estimation matrix;
  • the transmission information estimation matrix includes: the estimation information of the third pilot information sent by the information sending device and the estimation information of the second data information set sent;
  • the Said device also includes:
  • a channel estimation module configured to perform channel estimation processing on the second pilot information set and the first data information set to obtain a third channel estimation matrix
  • a third processing module configured to multiply each element in the transmission information estimation matrix by each element in the third channel estimation matrix to obtain the estimated information and the estimated information of the second pilot information set The estimated information of the first data information set.
  • Fig. 19 is a schematic structural diagram of an information processing device provided by an embodiment of the present application.
  • the information processing device may be a terminal device or a network device.
  • the information processing device 1900 shown in FIG. 19 includes a processor 1910 and a memory 1920.
  • the memory 1920 stores a computer program that can run on the processor 1910.
  • the processor 1910 executes the program, the embodiment of the present application is implemented. Methods.
  • the processor 1910 can invoke and run a computer program from the memory, so as to implement the method in the embodiment of the present application.
  • the information processing device 1900 may further include a memory 1920 .
  • the processor 1910 can invoke and run a computer program from the memory 1920, so as to implement the method in the embodiment of the present application.
  • the memory 1920 may be an independent device independent of the processor 1910 , or may be integrated in the processor 1910 .
  • the information processing device 1900 may further include a transceiver 1930, and the processor 1910 may control the transceiver 1930 to communicate with other devices, specifically, to send information or data to other devices, or receive information or data from other devices.
  • the transceiver 1930 may include a transmitter and a receiver.
  • the transceiver 1930 may further include antennas, and the number of antennas may be one or more.
  • FIG. 20 is a schematic structural diagram of a chip according to an embodiment of the present application.
  • the chip 2000 shown in FIG. 20 includes a processor 2010, and the processor 2010 can call and run a computer program from a memory, so as to implement the method in the embodiment of the present application.
  • the chip 2000 may further include a memory 2020 .
  • the processor 2010 can invoke and run a computer program from the memory 2020, so as to implement the method in the embodiment of the present application.
  • the memory 2020 may be a separate device independent of the processor 2010 , or may be integrated in the processor 2010 .
  • the chip 2000 may further include an input interface 2030 .
  • the processor 2010 can control the input interface 2030 to communicate with other devices or chips, specifically, can obtain information or data sent by other devices or chips.
  • the chip 2000 may further include an output interface 2040 .
  • the processor 2010 can control the output interface 2040 to communicate with other devices or chips, specifically, can output information or data to other devices or chips.
  • the chip can be applied to the information processing device in the embodiment of the present application, and the chip can implement the corresponding processes implemented by the information processing device in the various methods of the embodiment of the present application. For the sake of brevity, no further repeat.
  • the chip mentioned in the embodiment of the present application may also be called a system-on-chip, a system-on-chip, a system-on-a-chip, or a system-on-a-chip.
  • the embodiment of the present application also provides a computer storage medium, the computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors, so as to realize the Methods.
  • the computer-readable storage medium can be applied to the information processing device in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the information processing device in the methods of the embodiments of the present application, in order It is concise and will not be repeated here.
  • the embodiment of the present application also provides a computer program product, the computer program product includes a computer storage medium, the computer storage medium stores a computer program, and the computer program includes instructions executable by at least one processor, when the When the instructions are executed by the at least one processor, the methods in the embodiments of the present application are implemented.
  • the computer program product can be applied to the information processing device in the embodiments of the present application, and the computer program instructions cause the computer to execute the corresponding processes implemented by the information processing device in the various methods of the embodiments of the present application.
  • the computer program instructions cause the computer to execute the corresponding processes implemented by the information processing device in the various methods of the embodiments of the present application.
  • the embodiment of the present application also provides a computer program.
  • the computer program enables a computer to execute the method in the embodiment of the present application.
  • the computer program can be applied to the information processing device in the embodiment of the present application.
  • the computer program executes the corresponding functions implemented by the information processing device in the methods of the embodiment of the present application. For the sake of brevity, the process will not be repeated here.
  • the processor in the embodiment of the present application may be an integrated circuit chip, which has a signal processing capability.
  • each step of the above-mentioned method embodiments may be completed by an integrated logic circuit of hardware in a processor or instructions in the form of software.
  • processors may include any one or more of the following integrations: general-purpose processors, application-specific integrated circuits (Application Specific Integrated Circuit, ASIC), digital signal processors (Digital Signal Processor, DSP), digital signal processing devices (Digital Signal Processing Device, DSPD), Programmable Logic Device (Programmable Logic Device, PLD), Field Programmable Gate Array (Field Programmable Gate Array, FPGA), Central Processing Unit (Central Processing Unit, CPU), Graphics Processing Unit (Graphics Processing Unit, GPU), embedded neural-network processing units (NPU), controller, microcontroller, microprocessor, programmable logic device, discrete gate or transistor logic device, discrete hardware components.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • DSPD Digital Signal Processing Device
  • PLD Programmable Logic Device
  • Field Programmable Gate Array Field Programmable Gate Array
  • FPGA Field Programmable Gate Array
  • CPU Central Processing Unit
  • GPU Graphics Processing Unit
  • NPU embedded neural-network processing units
  • controller microcontroller, microprocess
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the steps of the method disclosed in connection with the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register.
  • the storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware.
  • the memory computer storage medium in the embodiments of the present application may be a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memories.
  • the non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electronically programmable Erase Programmable Read-Only Memory (Electrically EPROM, EEPROM) or Flash.
  • the volatile memory can be Random Access Memory (RAM), which acts as external cache memory.
  • RAM Static Random Access Memory
  • SRAM Static Random Access Memory
  • DRAM Dynamic Random Access Memory
  • Synchronous Dynamic Random Access Memory Synchronous Dynamic Random Access Memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM, DDR SDRAM enhanced synchronous dynamic random access memory
  • Enhanced SDRAM, ESDRAM synchronous connection dynamic random access memory
  • Synchlink DRAM, SLDRAM Direct Memory Bus Random Access Memory
  • Direct Rambus RAM Direct Rambus RAM
  • the memory in the embodiment of the present application may also be a static random access memory (static RAM, SRAM), a dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM) , Synchronous connection dynamic random access memory (synch link DRAM, SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DR RAM) and so on. That is, the memory in the embodiments of the present application is intended to include, but not be limited to, these and any other suitable types of memory.
  • the disclosed systems, devices and methods may be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the units is only a logical function division. In actual implementation, there may be other division methods.
  • multiple units or components can be combined or May be integrated into another system, or some features may be ignored, or not implemented.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • each functional unit in each embodiment of the present application may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit.
  • the functions described above are realized in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium.
  • the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory,) ROM, random access memory (Random Access Memory, RAM), magnetic disk or optical disc, etc., which can store program codes. .

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本申请实施例提供一种信息处理方法、装置、设备、介质、芯片、产品及程序，该方法包括：信息处理设备的第一处理模块接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；所述第二处理模块对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；所述第一处理模块利用所述第一信息和所述第三信息，对所述第一模型进行训练。

Description

信息处理方法、装置、设备、介质、芯片、产品及程序 技术领域
本申请实施例涉及通信技术领域，具体涉及一种信息处理方法、装置、设备、介质、芯片、产品及程序。
背景技术
鉴于人工智能(Artificial Intelligence，AI)技术在计算机视觉、自然语言处理等方面取得了巨大的成功，通信领域开始尝试利用AI技术来寻求新的技术思路来解决传统方法受限的技术难题。
如何对通信领域中的AI模型进行训练，是本领域一直关注的问题。
发明内容
本申请实施例提供一种信息处理方法、装置、设备、介质、芯片、产品及程序。
第一方面，本申请实施例提供一种信息处理方法，所述方法包括：
信息处理设备的第一处理模块接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；
所述第二处理模块对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；
所述第一处理模块利用所述第一信息和所述第三信息，对所述第一模型进行训练。
第二方面，本申请实施例提供一种信息处理装置，所述装置包括：
第一处理模块，用于接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；
第二处理模块，用于对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；
所述第一处理模块，还用于利用所述第一信息和所述第三信息，对所述第一模型进行训练。
第三方面，本申请实施例提供一种信息处理设备，包括：存储器和处理器，
所述存储器存储有可在处理器上运行的计算机程序，
所述处理器执行所述程序时实现上述方法。
第四方面，本申请实施例提供一种计算机存储介质，所述计算机存储介质存储有一个或者多个程序，所述一个或者多个程序可被一个或者多个处理器执行，以实现上述方法。
第五方面，本申请实施例提供一种芯片，包括：处理器，用于从存储器中调用并运行计算机程序，使得安装有所述芯片的设备执行上述方法。
第六方面，本申请实施例提供一种计算机程序产品，所述计算机程序产品包括计算机存储介质，所述计算机存储介质存储计算机程序，所述计算机程序包括能够由至少一个处理器执行的指令，当所述指令由所述至少一个处理器执行时实现上述方法。
第七方面，本申请实施例提供一种计算机程序，所述计算机程序使得计算机执行上述方法。
在本申请实施例中，由于第一处理模块中的第一模型对接收的第一信息进行处理，得到第二信息；第二处理模块对第二信息进行处理得到第一信息的估计信息；第一处理模块利用第一信息和第一信息的估计信息，对第一模型进行训练，这样，信息处理设备能够利用接收到的第一信息，和通过第一信息确定的第一信息的估计信息，对信息处理设备中的第一模型进行训练，从而信息处理设备的使用过程中通过接收到的第一信息即可实现对第一模型的训练，训练方式简单，且训练后的模型能够准确地对后续输入的信息进行处理，避免了相关技术中信息处理设备无法对第一模型进行训练的情况发生。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例 及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1为本申请实施例提供的一个应用场景的示意图；
图2为本申请实施例提供的一种无线通信系统的通信流程示意图；
图3为本申请实施例提供的一种无线通信系统中信道估计及恢复过程示意图；
图4为相关技术提出的一种神经网络的结构示意图；
图5为相关技术提供的一种卷积神经网络的结构示意图；
图6为相关技术中提供的一种自编码器的结构示意图；
图7为本申请实施例提供的一种利用AI实现信道估计与恢复的示意图；
图8为本申请实施例提供的一种基于神经网络的信道反馈系统的结构示意图；
图9为本申请实施例提供的一种信息收发过程的流程示意图；
图10为本申请实施例提供的一种信息处理方法的流程示意图；
图11为本申请实施例提供的一种第一模型的训练框架示意图；
图12为本申请实施例提供的一种获取第一信息的方式示意图；
图13为本申请实施例提供的另一种获取第一信息的方式示意图；
图14为本申请实施例提供的又一种获取第一信息的方式示意图；
图15为本申请实施例提供的一种第一模型为信道估计模型的更新架构示意图；
图16为本申请实施例提供的一种第一模型为解码器模型的更新架构示意图；
图17为本申请实施例提供的一种第一模型为接收机模型的更新架构示意图；
图18为本申请实施例提供的一种信息处理装置的组成结构示意图；
图19为本申请实施例提供的一种信息处理设备示意性结构图；
图20为本申请实施例的芯片的示意性结构图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行描述，显然，所描述的实施例是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。本申请实施例中的多个或多种，分别指的是至少两个或至少两种。
图1为本申请实施例提供的一个应用场景的示意图。
如图1所示，通信系统100可以包括信息处理设备101和信息发送设备102。信息发送设备102可以通过空口与信息处理设备101通信。信息处理设备101和信息发送设备102之间支持多业务传输。在一些实施例中，信息处理设备101可以称为信息接收设备。信息发送设备102也可以称为另一信息处理设备。信息处理设备101和信息发送设备102可以处于同一个网络中或者不同的网络中。
应理解，本申请实施例仅以通信系统100进行示例性说明，但本申请实施例不限定于此。也就是说，本申请实施例的技术方案可以应用于各种通信系统，例如：长期演进(Long Term Evolution，LTE)系统、LTE时分双工(Time Division Duplex，TDD)、通用移动通信系统(Universal Mobile Telecommunication System，UMTS)、物联网(Internet of Things，IoT)系统、窄带物联网(Narrow Band Internet of Things，NB-IoT)系统、增强的机器类型通信(enhanced Machine-Type Communications，eMTC)系统、5G通信系统(也称为新无线(New Radio，NR)通信系统)，或未来的通信系统(例如6G、7G通信系统)等。
本申请实施例中的信息处理设备101和/或信息发送设备102，可以称为用户设备(User Equipment，UE)、移动台(Mobile Station，MS)或移动终端(Mobile Terminal，MT)等。信息处理设备101和/或信息发送设备102，可以包括以下之一或者至少两者的组合：服务器、手机(mobile phone)、平板电脑(Pad)、带无线收发功能的电脑、掌上电脑、台式计算机、个人数字助理、便捷式媒体播放器、智能音箱、导航装置、智能手表、智能眼镜、智能项链等可穿戴设备、计步器、数字TV、虚拟现实(Virtual Reality，VR)信息处理设备、增强现实(Augmented Reality，AR)信息处理设备、工业控制(industrial control)中的无线终端、无人驾驶(self driving)中的无线终端、远程手术(remote medical surgery)中的无线终端、智能电网(smart grid)中的无线终端、运输安全(transportation safety)中的无线终端、智慧城市(smart city)中的无线终端、智慧家庭(smart home)中的无线终端以及车联网系统中的车、车载设备、车载模块、无线调制解调器(modem)、手持设备(handheld)、客户信息处理设备(Customer Premise Equipment，CPE)、智能家电。
在另一些实施例中，信息处理设备101和/或信息发送设备102，可以包括以下之一或者至少两者的 组合：6G中的基站、长期演进(Long Term Evolution，LTE)系统中的演进型基站(Evolutional Node B，eNB或eNodeB)、下一代无线接入网(Next Generation Radio Access Network，NG RAN)设备、NR系统中的基站(gNB)、小站、微站、云无线接入网络(Cloud Radio Access Network，CRAN)中的无线控制器、无线保真(Wireless-Fidelity，Wi-Fi)的接入点、传输接收点(transmission reception point，TRP)、中继站、接入点、车载设备、可穿戴设备、集线器、交换机、网桥、路由器、未来演进的公共陆地移动网络(Public Land Mobile Network，PLMN)中的信息发送设备等。
核心网设备可以是6G核心网设备或者5G核心网(5G Core，5GC)设备，核心网设备可以包括以下之一或者至少两者的组合：接入与移动性管理功能(Access and Mobility Management Function，AMF)、认证服务器功能(Authentication Server Function，AUSF)、用户面功能(User Plane Function，UPF)、会话管理功能(Session Management Function，SMF)、位置管理功能(Location Management Function，LMF)。在另一些实施方式中，核心信息发送设备也可以是LTE网络的分组核心演进(Evolved Packet Core，EPC)设备，例如，会话管理功能+核心网络的数据网关(Session Management Function+Core Packet Gateway，SMF+PGW-C)设备。应理解，SMF+PGW-C可以同时实现SMF和PGW-C所能实现的功能。在网络演进过程中，上述核心网设备也有可能叫其它名字，或者通过对核心网的功能进行划分形成新的网络实体，对此本申请实施例不做限制。
通信系统100中的各个功能单元之间还可以通过下一代网络(next generation，NG)接口建立连接实现通信。
例如，信息处理设备通过NR接口与接入网设备建立空口连接，用于传输用户面数据和控制面信令；信息处理设备可以通过NG接口1(简称N1)与AMF建立控制面信令连接；接入网设备例如下一代无线接入基站(gNB)，可以通过NG接口3(简称N3)与UPF建立用户面数据连接；接入网设备可以通过NG接口2(简称N2)与AMF建立控制面信令连接；UPF可以通过NG接口4(简称N4)与SMF建立控制面信令连接；UPF可以通过NG接口6(简称N6)与数据网络交互用户面数据；AMF可以通过NG接口11(简称N11)与SMF建立控制面信令连接；SMF可以通过NG接口7(简称N7)与PCF建立控制面信令连接。
本申请实施例不限定信息处理设备101和信息发送设备102的实现方式，信息处理设备101和信息发送设备102可以是任何两个能够无线通信的设备。
需要说明的是，图1只是以示例的形式示意本申请所适用的系统，当然，本申请实施例所示的方法还可以适用于其它系统。此外，本文中术语“系统”和“网络”在本文中常被可互换使用。本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。还应理解，在本申请的实施例中提到的“指示”可以是直接指示，也可以是间接指示，还可以是表示具有关联关系。举例说明，A指示B，可以表示A直接指示B，例如B可以通过A获取；也可以表示A间接指示B，例如A指示C，B可以通过C获取；还可以表示A和B之间具有关联关系。还应理解，在本申请的实施例中提到的“对应”可表示两者之间具有直接对应或间接对应的关系，也可以表示两者之间具有关联关系，也可以是指示与被指示、配置与被配置等关系。还应理解，在本申请的实施例中提到的“预定义”、“协议约定”、“预先确定”或“预定义规则”可以通过在设备(例如，包括信息处理设备和信息发送设备)中预先保存相应的代码、表格或其他可用于指示相关信息的方式来实现，本申请对于其具体的实现方式不做限定。比如预定义可以是指协议中定义的。还应理解，本申请实施例中，所述"协议"可以指通信领域的标准协议，例如可以包括LTE协议、NR协议以及应用于未来的通信系统中的相关协议，本申请对此不做限定。
为便于理解本申请实施例的技术方案，以下对本申请实施例的相关技术进行说明，以下相关技术作为可选方案与本申请实施例的技术方案可以进行任意结合，其均属于本申请实施例的保护范围。
图2为本申请实施例提供的一种无线通信系统的通信流程示意图，如图2所示，该无线通信系统可以包括发射端和接收端，发射端的发射机201和接收端的接收机202可以分别是上述的信息发送设备102和信息处理设备101。
在发射端，发射机201对信源比特流进行信道编码、调制，获得调制符号；在调制后的符号中插入导频符号，插入的导频符号用于接收端的信道估计和符号检测，最后形成发送信号，经过信道到达接收端。其中，信号经过信道发送到接收端的过程中会受到噪声的干扰。
在接收端，接收机202首先接收到接收信号，利用接收信号中的导频进行信道估计，并通过反馈链路将信道状态信息(Channel State Information，CSI)反馈给发送端，供发射机调整信道编码、调制、预编码等方式，最后，接收机通过符号检测、解调以及信道解码等步骤，获得最终的恢复比特流。
需要说明的是，图2是对无线通信系统的通信流程进行了简单的示意，无线通信系统中还有其他未 列举的如资源映射、预编码、干扰消除、CSI测量等模块，这些模块也都是单独设计实现，然后各个独立模块整合后可构成一个完整的无线通信系统。
由于无线信道环境的复杂性和时变性，在无线通信系统中，接收机针对无线信道的估计及恢复直接影响着最终的数据恢复性能。
图3为本申请实施例提供的一种无线通信系统中信道估计及恢复过程示意图，如图3所示，如(a)所示发射机在时频资源上发送的信号，除了信息数据符号外，还会发送一系列接收机已知的特定导频符号(即参考信号符号)，如信道状态信息参考信号(Channel State Information-Reference Signal，CSI-RS)信号、解调参考信号(DeModulation Reference Signal，DMRS)信号等。
发射机发送的信号经过信道传输到接收机，如(b)所示，接收机接收的数据符号和参考信号符号携带有噪音(即携带噪音的数据符号和参考信号符号)，接收机可以基于携带噪音的数据符号和参考信号符号进行信道估计，针对信道估计阶段，如(c)所示，接收机可根据真实导频与接收导频利用最小二乘法(Least Squares，LS)或者最小均方误差(Minimum Mean Square Error，MMSE)等方法估计出该参考信号位置上的信道信息。然后接收机可以进行信道恢复，针对信道恢复阶段，如(d)所示，接收机根据导频位置上估计出的信道信息利用插值算法恢复出全时频资源上的信道信息，用于后续的信道信息反馈或数据恢复等。
从(c)中可以看出，信道已估计/恢复的资源为参考信号符号的位置上的资源。从(d)中可以看出，信道已估计/恢复的资源为参考信号符号和数据符号的位置上的资源。
对于5G NR系统来说，在当前的CSI信道状态信息反馈设计中，主要是利用基于码本的方案来实现信道特征的提取与反馈。即在发送端进行信道估计后，根据信道估计的结果按照某种优化准则，从预先设定的预编码码本中选择与当前信道最匹配的预编码矩阵，并通过空口的反馈链路将矩阵的索引信息预编码矩阵指示(Precoding Matrix Indicator，PMI)反馈给接收端，供接收端实现预编码，同时也将根据测量得出的信道质量指示(Channel Quality Indication，CQI)反馈给接收端，供接收端实现自适应调制编码等。
近年来，以神经网络为代表的人工智能研究在很多领域都取得了非常大的成果，其也将在未来很长一段时间内在人们的生产生活中起到重要的作用。
图4为相关技术提出的一种神经网络的结构示意图，如图4所示，神经网络的结构可以包括：输入层，隐藏层和输出层，如图4所示，输入层负责接收数据，隐藏层对数据的处理，最后的结果在输出层产生。在这其中，各个节点代表一个处理单元，可以认为是模拟了一个神经元，多个神经元组成一层神经网络，多层的信息传递与处理构造出一个整体的神经网络。
随着神经网络研究的不断发展，近年来又提出了神经网络深度学习算法，较多的隐层被引入，通过多隐层的神经网络逐层训练进行特征学习，极大地提升了神经网络的学习和处理能力，并在模式识别、信号处理、优化组合、异常探测等方面广泛被应用。
同样，随着深度学习的发展，卷积神经网络(Convolutional Neural Networks，CNN)也被进一步研究。
图5为相关技术提供的一种卷积神经网络的结构示意图，如图5所示，卷积神经网络的结构可以包括：输入层、多个卷积层、多个池化层、全连接层及输出层。通过卷积层和池化层的引入，有效地控制了网络参数的剧增，限制了参数的个数并挖掘了局部结构的特点，提高了算法的鲁棒性。
图6为相关技术中提供的一种自编码器的结构示意图，如图6所示，以图像压缩为例，编码器61的输入为待压缩图像62，输出为码流63；解码器64输入为码流65，输出为解压缩后的图像66。从整体上来看，自编码器是一个端到端的、输入和学习目标相同的架构，即在训练期间可将待压缩图像作为自编码器的输入(编码器输入)和标签(解码器输出)，将编码器和解码器进行端到端联合训练。
图7为本申请实施例提供的一种利用AI实现信道估计与恢复的示意图，如图7所述，参考信号经过基于AI的信道估计与恢复模块71，基于AI的信道估计与恢复模块71的输入信息是参考信号，输出信息是信道估计结果。这里需要注意的是，对于这里所述的基于AI的信道估计与恢复模块71的输入信息除了参考信号外还可以增加其他辅助信息用于提升基于AI的信道估计与恢复模块71性能，例如，这些其他辅助信息可以是对参考信号的特征提取得到的信息、能量水平、时延特征、噪声特征等。
深度学习中常用的神经网络架构是非线性且是数据驱动的，可以对实际信道矩阵数据进行特征提取并在发送端尽可能还原UE端压缩反馈的信道矩阵信息，在保证还原信道信息的同时也为UE侧降低CSI反馈开销提供了可能性。
基于AI的CSI反馈将信道信息视作待压缩图像，利用深度学习自编码器对信道信息进行压缩反馈，并在发送端对压缩后的信道图像进行重构，可以更大程度地保留信道信息。
图8为本申请实施例提供的一种基于神经网络的信道反馈系统的结构示意图，如图8所示，信道反 馈系统分为编码器及解码器部分，分别部署在发送端与接收端。发送端通过信道估计得到信道信息后，通过编码器的神经网络对信道信息矩阵进行压缩编码，得到M×1维向量。并将压缩后的比特流(即M×1维向量)通过空口反馈链路反馈给接收端，接收端通过解码器根据反馈比特流对信道信息进行恢复，以获得完整的反馈信道信息。图8所示的结构在编码器使用若干全连接层进行编码，在解码器使用残差网络结构(例如RefineNet)进行解码。然而，在编解码框架不变的情况下，编码器和解码器内部的网络模型结构可进行灵活设计。
鉴于AI技术在计算机视觉、自然语言处理等方面取得了巨大的成功，通信领域开始尝试利用AI技术来寻求新的技术思路来解决传统方法受限的技术难题，例如深度学习。通过在接收机设计中引入基于人工智能的解决方案，利用神经网络实现整体模型设计，可以获得更好的接收机性能增益。
图9为本申请实施例提供的一种信息收发过程的流程示意图，如图9所示，可以利用神经网络直接替代传统接收机的信号处理流程。端到端的AI接收机输入为接收端接收的信号，输出为恢复的比特流，同时，AI接收机内部的网络模型结构可进行灵活设计。在发送端，信源比特流经过编码、调整等一系列操作后得到发送信号，通过信道发送至接收端，在发送过程中，会受到噪声的干扰。接收端接收到信号之后，可以将接收的信号发送至AI接收机，以使AI接收机输出恢复的比特流。
强化学习是从动物学习、参数扰动自适应控制等理论发展而来，其基本原理是：如果智能体的某个行为策略导致环境正的奖赏，那么智能体以后产生这个行为策略的趋势便会加强。智能体的目标是在每个离散状态发现最优策略以使期望的奖赏和最大。强化学习把学习看作试探评价过程，智能体选择一个动作用于环境，环境接收该动作后状态发生变化，同时产生一个强化信号(奖或惩)反馈给智能体，智能体根据强化信号和环境当前状态再选择下一个动作，选择的原则是使受到正强化(奖)的概率增大。选择的动作不仅影响即时的强化值，而且影响环境下一时刻的状态及最终的强化值。
传统的通信系统设计所基于的信道建模不再能很好地刻画由大规模天线、水下通信、毫米波等带来的日益复杂的信道环境；同时，信号处理器件越来越多样性的组合利用也给信号处理流程带来了一定的非线性特征，基于数学建模的传统信号处理方法越来越不能较好地满足当下高可靠的通信要求；以及传统通信系统中例如符号检测等迭代算法也具有较高复杂度，针对目前高速率的通信要求也不能很好地满足。
基于AI的端到端发射机、接收机设计可以在一定程度上解决传统通信系统的上述问题。但是，这种端到端的设计常常是单纯地数据驱动的，需要大量的数据和较长的时间来进行训练构建，而海量数据集的获取和长时间的训练在目前的无线通信系统中较为困难。本申请实施例提出的基于AI的模块化通信系统设计考虑数据与模型双驱动，可以很好地利用原有通信系统模型的先验结构，同时可以针对每个模块进行灵活地调整和训练，例如针对信道估计、调制编码等分别设计独立的神经网络模块，这样可以大大减少了大数据集和长训练时间的需求。
自编码器的编码器、解码器结构与通信系统中的很多架构天然适配，因此也出现了一系列自编码器在通信系统种的应用的研究。例如，编码器与解码器可分别对应整体通信系统的发送端和接收端，也可分别对应CSI反馈问题的信道压缩模块与解压缩模块。然而，这仅是结构上的适配。实际通信系统中空口的存在使得当编码器与解码器在发送端与接收端分布式部署时，训练数据很难在发送端与接收端同步，梯度难以通过空口准确传递，使得该结构只能在线下预先训练好，再部署到线上，难以实现线上的自更新。神经网络本身在实际应用中泛化问题的先天劣势导致训练好的网络仅针对与训练集具有相同特征的应用场景适用，即训练集常常难以囊括所有的情况，当场景特征发生变化时，训练好的模型就很难继续维持较好的泛化性能。
为便于理解本申请实施例的技术方案，以下通过具体实施例详述本申请的技术方案。以上提供的技术作为可选方案与本申请实施例的技术方案可以进行任意结合，其均属于本申请实施例的保护范围。本申请实施例包括以下内容中的至少部分内容。
图10为本申请实施例提供的一种信息处理方法的流程示意图，如图10所示，该方法应用于信息处理设备，该方法包括：
S1001、信息处理设备的第一处理模块接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息。
第一信息可以是信息处理设备从空口接收到的信息，或者，第一信息可以是信息处理设备对从空口接收到的信息进行处理得到的信息。
在一些实施例中，第一处理设备可以将第一信息输入至第一处理设备中的第一模型，通过第一模型对第一信息进行处理，得到并输出第二信息。本申请实施例中的第一模型可以是AI模型。
第一信息可以是矩阵或者是一个或多个向量。第二信息可以是矩阵或者一个或多个向量。
第一处理模块在得到第二信息后，可以将第二信息发送至信息处理设备中的其它处理模块，以使其 它处理模块对信息处理设备进行处理。
在确定需要对内部的第一模型进行更新的情况下，第一处理模块不仅向其它处理模块发送第二信息，还向第二处理模块发送第二信息；第一处理模块在确定不需要对内部的第一模型进行更新的情况下，第一处理模块仅向其它处理模块发送第二信息。
S1002、所述第二处理模块对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息。
在一些实施例中，第一信息可以是信息处理设备在预先定义的位置上发送的参考信号(也称第一导频信息集合)，第一模型可以是信道估计恢复模型(也称信道估计恢复网络)，第二信息可以是估计的信道状态信息(也称第一信道估计矩阵)。第三信息可以为所述第一导频信息集合的估计信息。在这种情况下，其它模块是可以第一编码模块和/或符号检测模块。第一编码模块可以将估计的信道状态信息进行编码后，向信息发送设备反馈。符号检测模块用于检测参考信号。
在另一些实施例中，所述第一信息包括第二比特流向量集合；所述第二信息包括第二信道估计矩阵。第一模型可以是解码器模型。所述第三信息包括所述第二比特流向量集合的估计信息。第二处理模块可以是第二编码模块。在这种情况下，其它模块可以是处理器和/或信道编码模块和/或调制模块，处理器能够获取到第二信道估计矩阵，信道编码模块和/或调制模块用于基于第二信道估计矩阵进行编码和/或调制。
在又一些实施例中，第一信息可以包括接收到的第二导频信息集合和接收到的第一数据信息集合；所述第二信息包括信源的估计信息。第一模型可以是接收机模型。所述第三信息包括所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。第二处理模块可以是发射机。在这种情况下，其它模块可以是处理器，处理器能够获取到信源的估计信息。
S1003、所述第一处理模块利用所述第一信息和所述第三信息，对所述第一模型进行训练。
在一些实施例中，可以利用第一信息和第三信息，确定每个训练样本中各个样本的权重，基于各个样本的权重，对所述第一模型进行训练。例如，可以基于第一信息中包括的各个信息、第三信息中包括的各个信息，以及各个样本的权重，对所述第一模型进行训练。其中，样本的权重可以是第一信息中包括的各个信息中每个信息在训练时的权重。各个样本和各个信息可以一一对应。在一些实施例中，训练的目的可以包括使得第一信息与第三信息之间的差异程度小于目标值。
在一些实施例中，基于所述第一信息和所述第三信息，对所述第一模型进行训练，可以包括：基于第一信息、第二信息以及第三信息，对第一模型进行训练。例如，可以将所述第一信息和所述第二信息确定为所述训练样本，基于训练样本和各个样本的权重，对所述第一模型进行训练。其中，各个样本的权重用于表征各个样本在训练过程中的重要程度。例如，某一个样本的权重越大，则该样本在模型训练时的重要程度越高。在一些实施例中，训练的目的可以包括使得第二信息，与通过第一模型对第三信息进行处理后得到的信息之间的差异程度小于目标值。
在本申请实施例中，由于第一处理模块中的第一模型对接收的第一信息进行处理，得到第二信息；第二处理模块对第二信息进行处理得到第一信息的估计信息；利用第一信息和第一信息的估计信息，对第一模型进行训练，这样，信息处理设备能够利用接收到的第一信息，和通过第一信息确定的第一信息的估计信息，对信息处理设备中的第一模型进行训练，从而信息处理设备的使用过程中通过接收到的第一信息即可实现对第一模型的训练，训练方式简单，且训练后的模型能够准确地对后续输入的信息进行处理，避免了相关技术中信息处理设备无法对第一模型进行训练的情况发生。
在一些实施例中，所述对所述第二信息进行处理得到第三信息，包括：
对所述第二信息采用第一处理方式和/或第一处理参数进行处理，得到所述第三信息；
所述第一处理方式和/或所述第一处理参数，与所述信息发送设备中的目标处理模块对接收到的第四信息所采用的第二处理方式和/或第二处理参数相同；所述目标处理模块对所述第四信息处理得到第五信息；
所述第二信息是所述第四信息的估计信息；所述第一信息是所述第五信息经过空口传输得到的。
在另一些实施例中，第五信息可以是经过信息发送设备的一个或多个模块进行处理，并通过空口传输到信息处理设备，经过信息处理模块的一个或多个模块进行处理，得到第一信息。
目标处理模块和第一处理模块执行的操作相反，例如，目标处理模块可以执行对X信息进行处理，得到Y信息，第一处理模块可以执行对Y信息进行处理，得到X信息。
第一信息可以是第五信息添加噪声后的信息。第一信息可以与第五信息的格式相同，例如，都是编码得到的信息，或者，都是添加导频得到的信息。
在一些实施例中，第四信息可以是信息发送设备的目标处理模块接收或获取到的第一信道矩阵集合，第二信息可以是估计的第一信道矩阵集合(即第一信道估计矩阵)。其中，可以基于导频序列集合和发 送的第一信道矩阵集合得到发送的导频信号，信息发送设备发送该导频信号，信息处理设备接收到导频信号(即接收到的第一导频信息集合)，通过信道估计恢复模型恢复出第一信道估计矩阵，基于第一信道估计矩阵和导频序列集合，确定第一导频信息集合的估计信息。
示例性地，基于导频序列集合和发送的第一信道矩阵集合得到发送的导频信号的一种实施方式可以为：对导频序列集合中的每个元素和发送的第一信道矩阵集合中的每个元素对应相乘，得到发送的导频信号。基于第一信道估计矩阵和导频序列集合，确定第一导频信息集合的估计信息的一种实施方式可以为：对第一信道估计矩阵中的每个元素和导频序列集合中的每个元素对应相乘，得到第一导频信息集合的估计信息。
在另一些实施例中，第四信息可以是信息发送设备的目标处理模块接收或获取到的第二信道矩阵集合，第二信息可以是估计的第二信道矩阵集合。其中，信息发送设备可以得到第二信道矩阵集合，通过编码器或者编码器模型对第二信道矩阵集合进行处理，得到发送的比特流向量集合，信息处理设备接收到比特流向量集合(即第二比特流向量集合)，通过解码器模型对第二比特流向量集合进行解码，得到估计的第二信道矩阵集合(即第二信道估计矩阵)，第二信道估计矩阵通过编码器或者编码器模型进行编码后得到估计的第二比特流向量集合(即第二比特流向量集合的估计信息)。
示例性地，信息发送设备的编码器或者编码器模型中的编码参数，与信息处理设备的编码器或者编码器模型的编码参数可以相同。
在又一些实施例中，第四信息可以是信息发送设备的目标处理模块接收或获取到的信源信息，第二信息可以是估计的信源信息。其中，信息发送设备可以得到信源，采用发射机或者发射机模型对信源进行处理，得到发送的导频信息集合和数据信息集合，信息处理设备接收到导频信息集合和数据信息集合(即接收到第二导频信息集合和接收到的第一数据信息集合)，AI接收机对第二导频信息集合和第一数据信息集合进行处理，得到估计的信源(即信源的估计信息)，信息处理设备的发射机或者发射机模型可以对信源的估计信息进行处理，得到估计的导频信息集合和估计的数据信息集合。另外，信息处理设备中的信道估计模块可以对接收到第二导频信息集合和接收到的第一数据信息集合进行处理，得到第三信道估计矩阵，基于估计的导频信息集合和估计的数据信息集合，以及第三信道估计矩阵，得到对接收到的导频信息集合和接收到的数据信息集合的估计信息。
示例性地，信息发送设备的发射机或发射机模型中的发射参数，与信息处理设备的发射机或者发射机模型的发射参数可以相同。基于估计的导频信息集合和数据信息集合，以及第三信道估计矩阵，得到对接收到导频信息集合和数据信息集合的估计信息，可以包括：估计的导频信息集合和估计的数据信息集合所形成的矩阵中的每个元素，与第三信道估计矩阵中的每个元素对应相乘，得到对接收到的导频信息集合和接收到的数据信息集合的估计信息。
本申请实施例采用强化学习的方法，提出一种基于强化学习的第一模型更新方案及其通信应用设计。
图11为本申请实施例提供的一种第一模型的训练框架示意图，如图11所示，原始信号
Figure PCTCN2021123777-appb-000001
首先通过信息发送设备侧的处理模块(即上述的目标处理模块)处理并发送后，输出
Figure PCTCN2021123777-appb-000002
(第一信息)，信息处理设备接收到信号
Figure PCTCN2021123777-appb-000003
并通过第一模型输出恢复后的原始信号
Figure PCTCN2021123777-appb-000004
(第二信息)。
本申请实施例在信息处理设备侧事先部署一个与信息发送设备侧的用于处理
Figure PCTCN2021123777-appb-000005
的处理模块(例如编码器)结构和参数相同的处理模块，构成一个强化学习问题，问题中的几个要素分别为：
将信息处理设备侧的第一模型(例如是解码器或解码器模型)视作强化学习中的智能体；
将信息处理设备侧的第一模型的解码过程视作智能体的动作；
将信息处理设备侧的处理模块视作强化学习中的环境；
信号
Figure PCTCN2021123777-appb-000006
经过信息处理设备侧的第一模型g处理后，获得信号
Figure PCTCN2021123777-appb-000007
进一步将信号
Figure PCTCN2021123777-appb-000008
送入信息处理设备侧部署的处理模块f进行处理，获得处理后的信号
Figure PCTCN2021123777-appb-000009
(第三信息)，将处理后的信号
Figure PCTCN2021123777-appb-000010
视作强化学习问题中的状态；
基于某种度量规则h，例如余弦相似度或均方误差等，利用信号
Figure PCTCN2021123777-appb-000011
与估计信号
Figure PCTCN2021123777-appb-000012
计算
Figure PCTCN2021123777-appb-000013
视作强化学习问题中的奖赏，即利用信号
Figure PCTCN2021123777-appb-000014
与估计信号
Figure PCTCN2021123777-appb-000015
进行奖赏计算，以使
Figure PCTCN2021123777-appb-000016
的值最小。
当完成问题构建后，可将信息处理设备侧部署的处理模块参数冻结，利用如策略梯度法等强化学习算法对第一模型进行在线训练更新。
在一些实施例中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，可以包括：
利用所述第一信息和所述第三信息，确定余弦相似度和/或均方误差；
利用所述余弦相似度和/或所述均方误差，对所述第一模型进行训练。
其中，余弦相似度和/或均方误差可以用于表征第一信息和第三信息之间的差异程度。
示例性地，可以基于所述余弦相似度和/或所述均方误差，确定训练样本中各个样本的权重，基于各个样本的权重，对所述第一模型进行训练。例如，基于各个样本的权重，对所述第一模型进行训练，可以包括：获取训练样本中各个样本，基于训练样本中各个样本和各个样本的权重，对第一模型进行训练。
在一些实施例中，所述第一信息包括一个或多个子信息；所述方法还包括：
所述第一处理模块获取第一指示信息；所述第一指示信息指示在每个训练周期训练一次所述第一模型；所述每个训练周期包括一个或多个传输周期；
所述第一处理模块将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在信息处理设备获取到第一指示信息的情况下，进入确定第一信息的步骤。信息处理设备在每个训练周期都可以确定一个第一信息。
信息处理设备在每个传输周期内可以接收数据，将每个传输周期内接收的数据确定为每个传输周期对应的信息。每个传输周期对应的至少部分信息可以是每个传输周期对应的信息中的部分信息或者全部信息。
在一些实施例中，所述第一指示信息还指示以下至少之一：
训练周期、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
第一指示信息还可以包括：开始获取第一数据的起始时间，在这个起始时间上，信息处理设备获取每个传输周期对应的至少部分信息，从而不断地得到第一信息。示例性地，开始获取第一数据的起始时间与指示信令所在时隙的偏移，可作为指示的起始时间。
在这种实施方式中，只要信息处理设备获取第一指示信息，就在每个训练周期训练一次第一模型。在一些实施例中，信息处理设备可以在训练次数到到达预设次数的情况下，不再进行第一模型的训练，以及不再采集子信息。在另一些实施例中，信息处理设备可以接收网络设备发送的第四指示信息，第四指示信息可以指示停止训练第一模型，从而信息处理设备可以基于第四指示信息，不再进行第一模型的训练，以及不再采集子信息。
图12为本申请实施例提供的一种获取第一信息的方式示意图，如图12所示，在每一个传输周期中，发射端向接收端发送一次或多次信号。该传输可分为在线学习型传输及传统传输，当第一模型在推理前需要进行在线学习微调时，定义为在线学习型传输(传输周期t)；当第一模型仅进行推理，同时收集数据的过程定义为传统传输。定义为训练周期r。每一个训练周期r中可包含若干个传输周期t，其中，当传输周期为传统传输时，信息处理设备接收信号并将信号进行本地存储，实现在线训练数据收集，并同时针对本次传输的信号利用第一模型进行推理；当传输周期为在线学习型传输时，信息处理设备可利用已经收集好的在线训练数据进行第一模型在线训练，并利用训练后的第一模型进行推理。在整个传输进程中，相关参数可考虑利用无线资源控制(Radio Resource Control，RRC)信令或控制信道指示。相关参数至少包括如下的一种：训练周期、在线训练数据集大小。其中，在线训练数据集大小指每次在线训练时并不一定使用全部已收集的数据，而是从训练数据集中取一部分子集作为在线训练的数据集。
在一些实施例中，所述第一信息包括一个或多个子信息；所述方法还包括：
所述第一处理模块获取第二指示信息；所述第二指示信息指示在N个训练周期中每个训练周期训练一次所述第一模型；所述训练周期包括一个或多个传输周期；所述N为大于或等于1的整数；
所述第一处理模块将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在信息处理设备获取到第二指示信息的情况下，进入确定第一信息的步骤。信息处理设备在N个训练周期的每个训练周期确定一个第一信息。
在一些实施例中，述第二指示信息还指示以下至少之一：
所述N个训练周期的起始时间、训练周期、所述N、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
其中，N个训练周期的起始时间与指示信令所在时隙的偏移，可作为指示的起始时间。
在这种实施方式中，只要信息处理设备获取第二指示信息，就在每个训练周期训练一次第一模型，并连续地训练第二指示信息指示的N次。信息处理设备在训练次数达到N次的情况下，不再进行第一模型的训练，以及不再采集子信息。
图13为本申请实施例提供的另一种获取第一信息的方式示意图，如图13所示，为了系统的更灵活配置，在周期性的基础上，可利用信令(即第二指示信息)指示周期性训练的开始与结束。在整个传输 进程中，相关参数可考虑利用RRC信令或控制信道指示。参数至少包括如下的一种：在线训练起点、在线训练周期、在线训练总周期数、在线训练数据集大小。
其中，在线训练起点与指示信令所在时隙的偏移，可作为在线训练起点的指示信息；在线训练总周期数决定了在线训练周期的结束点。
在一些实施例中，所述第一信息包括一个或多个子信息；所述方法还包括：
所述第一处理模块获取第三指示信息；所述第三指示信息指示训练一次所述第一模型；
所述第一处理模块将一个或多个传输周期中每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在一些实施例中，所述第三指示信息还指示以下至少之一：
所述一个或多个传输周期的起始时间、所述传输周期的周期数、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
其中，一个或多个传输周期的起始时间与指示信令所在时隙的偏移，可作为指示的一个或多个传输周期的起始时间；在线训练总周期数决定了在线训练周期的结束点。
图14为本申请实施例提供的又一种获取第一信息的方式示意图，如图14所示，每次都通过信令(第三指示信息)指示在线学习型传输。信令指示的参数至少包括：在线训练数据集大小。
在整个传输进程中，指示信息(包括第一指示信息、第二指示信息、第三指示信息中的至少一者)指示的内容，可以根据实际传输环境、时延及复杂度要求等灵活配置。示例性地，在实际传输环境、时延及复杂度要求等至少之一发生变化的情况下，可以修改指示信息指示的参数值。
本申请实施例可以提高基于AI的通信系统对环境的整体自适应性，在线学习型传输的配置占比越高，第一模型的自适应性越强。
在一些实施例中，在接收到指示信息的情况下，信息处理设备先得到一个或多个子信息，并将一个或多个子信息确定为第一信息，然后执行第一模型的训练，在接收到之后，第一次执行第一模型的训练所对应时间，与接收到的时间间隔至少一个传输周期。
在另一些实施例中，在接收到的情况下，信息处理设备可以将获取的之后的第一个传输周期的子信息，确定为第一信息，采用第一信息执行第一次第一模型的训练，接着在第一个传输周期之后，每隔训练周期获取一次第一信息，并根据每次得到的第一信息进行一次第一模型的训练。
在又一些实施例中，在接收到的情况下，信息处理设备可以将获取的之后的第一个传输周期的子信息，以及在之前的至少一个传输周期对应的信息中获取若干个子信息，从而确定一个训练周期对应的全部子信息，将该全部子信息确定为第一个第一信息，执行第一次第一模型的训练。
在一些实施方式中，指示信息可以是信息处理设备中预先存储的指示信息。例如，信息处理设备可以从自身获取指示信息。在另一些实施方式中，指示信息可以是另一信息处理设备指示的，例如，信息处理设备为终端设备的情况下，指示信息可以是网络设备指示的。网络设备可以利用RRC信令或控制信道指示来指示该指示信息。在一些实施例中，信息处理设备或者另一信息处理设备可以在确定第一信息的估计信息与第一信息之间误差较大的情况下，获取指示信息。
在一些实施例中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一均方误差；对所述第一均方误差进行归一化处理，得到第一归一化结果；将所述第一信息和所述第二信息确定为所述训练样本，所述第一归一化结果确定为所述各个样本的第一权重；利用所述训练样本和所述第一权重，对所述第一模型进行训练。
第一信息可以为
Figure PCTCN2021123777-appb-000017
第三信息可以为
Figure PCTCN2021123777-appb-000018
基于第一信息B和第三信息B'，计算对应训练样本中各个样本的第一均方误差
Figure PCTCN2021123777-appb-000019
对第一均方误差
Figure PCTCN2021123777-appb-000020
进行归一化，得到第一归一化结果，归一化的步骤是首先计算
Figure PCTCN2021123777-appb-000021
中包括的m1,...,mn的均值e和标准差v，进而得到的第一归一化结果为：将第一均方误差
Figure PCTCN2021123777-appb-000022
中的每个元素与均值e相减后，再将得到的每个元素除以标准差v的结果。第一归一化结果可以是各个样本的第一权重，用
Figure PCTCN2021123777-appb-000023
表示。
在一些实施例中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一标准均方误差；对所述第一标准均方误差进行归一化处理，得到第二归一化结果；将所述第一信息和所述第二信息确定为所述训练样本，所述第二归一化结果确定为所述各个样本的第二权重；利用所述训练样本和所述第二权重，对所述第一模型进行训练。
第一信息可以为
Figure PCTCN2021123777-appb-000024
第三信息可以为
Figure PCTCN2021123777-appb-000025
基于第一信息B和第三信息B'，计算对应训练样本中各个样本的第一标准均方误差
Figure PCTCN2021123777-appb-000026
对s为比特流长度，例如，s的值可以与
Figure PCTCN2021123777-appb-000027
的维度有关。对第一标准均方误差
Figure PCTCN2021123777-appb-000028
进行归一化，得到第二归一化结果，归一化的步骤是首先计算
Figure PCTCN2021123777-appb-000029
中包括的m1,...,mn的均值e和标准差v，进而得到的第二归一化结果为：将第一均方误差
Figure PCTCN2021123777-appb-000030
中的每个元素与均值e相减后，再将得到的每个元素除以标准差v的结果。第二归一化结果可以是各个样本的第一权重，用
Figure PCTCN2021123777-appb-000031
表示。
在一些实施例中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：对所述第一信息进行解量化(Dequantized)处理，得到第一解量化结果；对所述第三信息进行解量化处理，得到第二解量化结果；利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练。
在一些实施例中，所述利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练，包括：利用所述第一解量化结果和所述第二解量化结果，确定对应训练样本中各个样本的第二均方误差或者第二标准均方误差；对所述第二均方误差或者所述第二标准均方误差进行归一化处理，得到第三归一化结果；将所述第一信息和所述第二信息确定为所述训练样本，所述第三归一化结果确定为所述各个样本的第三权重；利用所述训练样本和所述第三权重，对所述第一模型进行训练。
第一信息可以为
Figure PCTCN2021123777-appb-000032
第三信息可以为
Figure PCTCN2021123777-appb-000033
对所述第一信息进行解量化处理，得到的第一解量化结果可以为
Figure PCTCN2021123777-appb-000034
对所述第三信息进行解量化处理，得到的第二解量化结果可以为
Figure PCTCN2021123777-appb-000035
deQ(·)为解量化计算。
此处以基于所述第一解量化结果和所述第二解量化结果，确定对应训练样本中各个样本的第二均方误差为例进行说明：基于Z和Z'，确定的对应训练样本中各个样本的第二均方误差
Figure PCTCN2021123777-appb-000036
对第二均方误差
Figure PCTCN2021123777-appb-000037
进行归一化，得到第三归一化结果，归一化的步骤是首先计算
Figure PCTCN2021123777-appb-000038
中包括的m1,...,mn的均值e和标准差v，进而得到的第三归一化结果为：将第二均方误差
Figure PCTCN2021123777-appb-000039
中的每个元素与均值e相减后，再将得到的每个元素除以标准差v的结果。第三归一化结果可以是各个样本的第一权重，用
Figure PCTCN2021123777-appb-000040
表示。
在一些实施例中，所述第一模型包括信道估计模型；所述第一信息包括接收到的第一导频信息集合；所述第二信息包括第一信道估计矩阵；所述第三信息包括所述第一导频信息集合的估计信息；所述对所述第二信息进行处理得到第三信息，包括：将所述第一信道估计矩阵中每个元素，与导频序列矩阵中每个元素对应相乘，得到所述第一导频信息集合的估计信息。
以下对第一模型为信道估计模型进行说明：通信系统的发送端(信息发送设备)可以发送导频序列集合P，导频序列集合可以称为导频向量、导频符号向量、导频信息集合、导频矩阵等。接收端(信息处理设备)接收到的第一导频信息集合为YP＝H.*P+N。其中，H为信道信息(或者称信道矩阵)，N为加性噪声。其中，H.*P表示矩阵H中的每个元素和矩阵P中的每个元素对应相乘。在一些实施例中，信道估计模型的输入为YP，输出为H的估计信息(即第一信道估计矩阵)。在另一些实施例中，信道估计模型的输入为YP和导频序列集合P，输出为H的估计信息(即第一信道估计矩阵)。
图15为本申请实施例提供的一种第一模型为信道估计模型的更新架构示意图，如图15所示，将从信道信息
Figure PCTCN2021123777-appb-000041
到第一导频信息集合
Figure PCTCN2021123777-appb-000042
的过程，视为发送端的信息发送设备中处理模块的处理过程(例如，编码器的编码过程)，将从第一导频信息集合YP到第一信道估计矩阵
Figure PCTCN2021123777-appb-000043
的过程视为接收端的信息处理设备中处理模块的处理过程(例如，解码器的解码过程)，可以在接收端，即信息处理设备侧实现对信道估计模型的更新。
信息处理设备在接收到第一导频信息集合YP之后，利用信道估计模型进行信道估计，得到第一信道估计矩阵H'。利用第一信道估计矩阵H'和导频序列集合P生成第一导频信息集合YP的估计信息YP'。示例性地，可以将第一信道估计矩阵H'中的每个元素，与导频序列集合P中的每个元素对应相乘，得到第一导频信息集合YP的估计信息YP'。
在一些实施例中，第一导频信息集合YP可以是第一信息，第一信道估计矩阵可以是第二信息，第一导频信息集合YP的估计信息YP'可以是第三信息。信道信息H可以是第四信息。对信道信息H进行的处理为H.*P，而对第一信道估计矩阵H'进行的处理也为H'.*P。
在一些实施例中，训练样本可以为
Figure PCTCN2021123777-appb-000044
可以分别为训练样本中的N个样本。
通过得到的训练样本D和上述得到的权重m(包括第一权重、第二权重或者第三权重)，对信道估计模型进行训练。在条件满足的情况下，训练停止，条件满足包括但不限于训练次数达到设置的最大迭代次数。
在一些实施例中，所述方法还包括：所述第一处理模块向所述信息处理设备的第一编码模块发送所述第一信道估计矩阵；所述第一编码模块对所述第一信道估计矩阵进行编码处理，得到第一比特流向量集合；所述第一编码模块向信息发送设备发送所述第一比特流向量集合。
在一些实施例中，所述第一编码模块可以向信息处理模块的收发模块发送第一比特流向量集合，以使收发模块向信息发送设备发送第一比特流向量集合。
可以通过编码器或者编码器模型对所述第一信道估计矩阵进行编码处理，得到第一比特流向量集合。
在信息处理设备向信息发送设备发送第一比特流向量集合的情况下，信息发送设备可以将接收到的比特流信息向量输入至解码器模型，以使解码器模型对接收到的比特流信息向量处理，得到特定信道估计矩阵，其中，特定信道估计矩阵是对第一信道估计矩阵的估计信息。在一些实施例中，信息发送设备也可以对信息发送设备中的解码器模型进行训练，信息发送设备中的解码器模型进行训练的训练方式，可以与信息处理设备对信息处理设备中的解码器模型进行训练的训练方式相同。
在一些实施例中，所述第一模型包括解码器模型；所述第二处理模块包括第二编码模块；所述第一信息包括第二比特流向量集合；所述第二信息包括第二信道估计矩阵；所述第三信息包括所述第二比特流向量集合的估计信息；所述第二处理模块对所述第二信息进行处理得到第三信息，包括：所述第二编码模块对所述第二信道估计矩阵进行编码处理，得到所述第二比特流向量集合的估计信息。
图16为本申请实施例提供的一种第一模型为解码器模型的更新架构示意图，如图16所示，通信系统的发送端(信息发送设备)可以通过编码器或编码器模型，将信道集合
Figure PCTCN2021123777-appb-000045
中的每个信道进行压缩编码为比特流向量，构成比特流向量集合，通过反馈链路发送给接收端(信息处理设备)。信息处理设备通过解码器模型将接收到的第二比特流向量集合
Figure PCTCN2021123777-appb-000046
中的每个比特流分别恢复为对应的第二信道估计矩阵
Figure PCTCN2021123777-appb-000047
利用信息处理设备端的编码器或编码器模型将
Figure PCTCN2021123777-appb-000048
中的每个信道重新编码为对应的比特流，得到第二比特流向量集合的估计信息
Figure PCTCN2021123777-appb-000049
信息处理设备端的编码器或编码器模型的编码参数，与信息发送设备端的编码器或编码器模型的编码参数相同。
在一些实施方式中，第二比特流向量集合
Figure PCTCN2021123777-appb-000050
可以是第一信息，第二信道估计矩阵
Figure PCTCN2021123777-appb-000051
可以是第二信息，第二比特流向量集合的估计信息
Figure PCTCN2021123777-appb-000052
可以是第三信息。信道集合
Figure PCTCN2021123777-appb-000053
可以是第四信息。
在一些实施例中，训练样本可以为
Figure PCTCN2021123777-appb-000054
可以分别为训练样本中的N个样本。
通过得到的训练样本D和上述得到的权重m(包括第一权重、第二权重或者第三权重)，对解码器模型进行训练。在条件满足的情况下，训练停止，条件满足包括但不限于训练次数达到设置的最大迭代次数。
在一些实施例中，所述第一模型包括接收机模型；所述第二处理模块包括发射机(或者称发射机模型，发射机在本申请实施例中不包括需要训练的模型)；所述第一信息包括接收到的第二导频信息集合和接收到的第一数据信息集合；所述第二信息包括信源的估计信息；所述第三信息包括所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息；所述第二处理模块对所述第二信息进行处理得到第三信息，包括：所述发射机对所述信源的估计信息采用发射机进行处理，得到处理结果；所述处理结果用于：确定所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
在一些实施例中，所述所述处理结果包括发送信息估计矩阵；所述发送信息估计矩阵包括：信息发送设备发送的第三导频信息的估计信息和发送的第二数据信息集合的估计信息；
所述方法还包括：
所述信息处理设备的信道估计模块对所述第二导频信息集合和所述第一数据信息集合进行信道估计处理，得到第三信道估计矩阵；所述信息处理设备的第三处理模块将所述发送信息估计矩阵中每个元 素，与所述第三信道估计矩阵中的每个元素对应相乘，得到所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
图17为本申请实施例提供的一种第一模型为接收机模型的更新架构示意图，如图17所示，通信系统的发送端(信息发送设备)可以通过发射机或者发射机模型，对信源比特流向量
Figure PCTCN2021123777-appb-000055
进行处理，得到插入导频后的符号矩阵
Figure PCTCN2021123777-appb-000056
其中，发射机或者发射机模型可以用于对信源比特流向量
Figure PCTCN2021123777-appb-000057
进行编码和调制等处理。
Figure PCTCN2021123777-appb-000058
为数据向量，
Figure PCTCN2021123777-appb-000059
为导频符号向量。符号矩阵x通过空口发送至接收端，接收端接收到的信号为
Figure PCTCN2021123777-appb-000060
h.*x表示矩阵h中的每个元素和矩阵x中的每个元素对应相乘；矩阵h为信道矩阵，N为加性噪声；yp为接收到的数据矩阵，yd为接收到的符号矩阵。
接收端的信息处理设备一方面将接收到的信号集合
Figure PCTCN2021123777-appb-000061
(即接收到的第二导频信息集合和接收到的第一数据信息集合)送入接收机或者接收机模型，通过接收机或者接收机模型还原出比特流向量集合
Figure PCTCN2021123777-appb-000062
(即信源的估计信息)，另一方面通过信道估计模块估计出信道集合
Figure PCTCN2021123777-appb-000063
(即第三信道估计矩阵)。其中，信道估计模块可为传统通信系统的信道估计模块，也可为基于AI的信道估计模块。
信息处理设备可以将还原出比特流向量集合
Figure PCTCN2021123777-appb-000064
中的每个比特送入信息处理设备的发射机或发射机模型进行处理，得到发送信息估计矩阵。将发送信息估计矩阵与估计出信道集合
Figure PCTCN2021123777-appb-000065
中每个元素对应相乘，得到重构的接收信号集合
Figure PCTCN2021123777-appb-000066
(即第二导频信息集合的估计信息和所述第一数据信息集合的估计信息)。
信息处理设备端的发射机或发射机模型的发射参数，与信息发送设备端的发射机或发射机模型的发射参数相同。
在一些实施例中，接收到的第二导频信息集合和接收到的第一数据信息集合可以是第一信息，信源的估计信息可以是第二信息，第二导频信息集合的估计信息和所述第一数据信息集合的估计信息可以是第三信息。多个信源比特流向量
Figure PCTCN2021123777-appb-000067
所构成的集合可以是第四信息。
在一些实施例中，训练样本可以为
Figure PCTCN2021123777-appb-000068
可以分别为训练样本中的N个样本。
通过得到的训练样本D和上述得到的权重m(包括第一权重、第二权重或者第三权重)，对接收机模型进行训练。在条件满足的情况下，训练停止，条件满足包括但不限于训练次数达到设置的最大迭代次数。
在本申请实施例中，不同实施例中相同的字母代表的含义不同，不同的含义可以根据上下文通过简单的推理得出。
以上结合附图详细描述了本申请的优选实施方式，但是，本申请并不限于上述实施方式中的具体细节，在本申请的技术构思范围内，可以对本申请的技术方案进行多种简单变型，这些简单变型均属于本申请的保护范围。例如，在上述具体实施方式中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合，为了避免不必要的重复，本申请对各种可能的组合方式不再另行说明。又例如，本申请的各种不同的实施方式之间也可以进行任意组合，只要其不违背本申请的思想，其同样应当视为本申请所公开的内容。又例如，在不冲突的前提下，本申请描述的各个实施例和/或各个实施例中的技术特征可以和现有技术任意的相互组合，组合之后得到的技术方案也应落入本申请的保护范围。
另外，本申请实施例中，术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系。具体地，A和/或B可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。
图18为本申请实施例提供的一种信息处理装置的组成结构示意图，该信息处理装置1800可以应用于上述的信息处理设备，如图18所示，所述信息处理装置1800包括：
第一处理模块1801，用于接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；
第二处理模块1802，用于对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；
所述第一处理模块1801，还用于利用所述第一信息和所述第三信息，对所述第一模型进行训练。
在一些实施例中，所述第二处理模块1802，还用于对所述第二信息采用第一处理方式和/或第一处理参数进行处理，得到所述第三信息；
所述第一处理方式和/或所述第一处理参数，与所述信息发送设备中的目标处理模块对接收到的第四信息所采用的第二处理方式和/或第二处理参数相同；所述目标处理模块对所述第四信息处理得到第五信息；
所述第二信息是所述第四信息的估计信息；所述第一信息是所述第五信息经过空口传输得到的。
在一些实施例中，所述第一处理模块1801，还用于：
利用所述第一信息和所述第三信息，确定余弦相似度和/或均方误差；
利用所述余弦相似度和/或所述均方误差，对所述第一模型进行训练。
在一些实施例中，所述第一信息包括一个或多个子信息；第一处理模块1801还包括：
获取第一指示信息；所述第一指示信息指示在每个训练周期训练一次所述第一模型；所述每个训练周期包括一个或多个传输周期；
将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在一些实施例中，所述第一指示信息还指示以下至少之一：
训练周期、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
在一些实施例中，所述第一信息包括一个或多个子信息；第一处理模块1801还用于：
获取第二指示信息；所述第二指示信息指示在N个训练周期中每个训练周期训练一次所述第一模型；所述训练周期包括一个或多个传输周期；所述N为大于或等于1的整数；
将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在一些实施例中，所述第二指示信息还指示以下至少之一：
所述N个训练周期的起始时间、训练周期、所述N、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
在一些实施例中，所述第一信息包括一个或多个子信息；第一处理模块1801还用于：
获取第三指示信息；所述第三指示信息指示训练一次所述第一模型；
将一个或多个传输周期中每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
在一些实施例中，所述第三指示信息还指示以下至少之一：
所述一个或多个传输周期的起始时间、所述传输周期的周期数、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
在一些实施例中，所述第一处理模块1801，还用于：
利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一均方误差；
对所述第一均方误差进行归一化处理，得到第一归一化结果；
将所述第一信息和所述第二信息确定为所述训练样本，所述第一归一化结果确定为所述各个样本的第一权重；
利用所述训练样本和所述第一权重，对所述第一模型进行训练。
在一些实施例中，所述第一处理模块1801，还用于：
利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一标准均方误差；
对所述第一标准均方误差进行归一化处理，得到第二归一化结果；
将所述第一信息和所述第二信息确定为所述训练样本，所述第二归一化结果确定为所述各个样本的第二权重；
利用所述训练样本和所述第二权重，对所述第一模型进行训练。
在一些实施例中，所述第一处理模块1801，还用于：
对所述第一信息进行解量化处理，得到第一解量化结果；
对所述第三信息进行解量化处理，得到第二解量化结果；
利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练。
在一些实施例中，所述第一处理模块1801，还用于：
利用所述第一解量化结果和所述第二解量化结果，确定对应训练样本中各个样本的第二均方误差或者第二标准均方误差；
对所述第二均方误差或者所述第二标准均方误差进行归一化处理，得到第三归一化结果；
将所述第一信息和所述第二信息确定为所述训练样本，所述第三归一化结果确定为所述各个样本的第三权重；
利用所述训练样本和所述第三权重，对所述第一模型进行训练。
在一些实施例中，所述第一模型包括信道估计模型；
所述第一信息包括接收到的第一导频信息集合；
所述第二信息包括第一信道估计矩阵；
所述第三信息包括所述第一导频信息集合的估计信息；
所述第二处理模块1802，还用于：将所述第一信道估计矩阵中每个元素，与导频序列矩阵中每个元素对应相乘，得到所述第一导频信息集合的估计信息。
在一些实施例中，所述装置还包括：
第一编码模块，用于对所述第一信道估计矩阵进行编码处理，得到第一比特流向量集合；
所述第一编码模块，还用于向信息发送设备发送所述第一比特流向量集合。
在一些实施例中，所述第一模型包括解码器模型；所述第二处理模块包括第二编码模块；
所述第一信息包括第二比特流向量集合；
所述第二信息包括第二信道估计矩阵；
所述第三信息包括所述第二比特流向量集合的估计信息；
所述第二编码模块，用于对所述第二信道估计矩阵进行编码处理，得到所述第二比特流向量集合的估计信息。
在一些实施例中，所述第一模型包括接收机模型；所述第二处理模块包括发射机；
所述第一信息包括接收到的第二导频信息集合和接收到的第一数据信息集合；
所述第二信息包括信源的估计信息；
所述第三信息包括所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息；
所述发射机，用于对所述信源的估计信息进行处理，得到处理结果；所述处理结果用于：确定所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
在一些实施例中，所述处理结果包括发送信息估计矩阵；所述发送信息估计矩阵包括：信息发送设备发送的第三导频信息的估计信息和发送的第二数据信息集合的估计信息；所述装置还包括：
信道估计模块，用于对所述第二导频信息集合和所述第一数据信息集合进行信道估计处理，得到第三信道估计矩阵；
第三处理模块，用于将所述发送信息估计矩阵中每个元素，与所述第三信道估计矩阵中的每个元素对应相乘，得到所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
本领域技术人员应当理解，本申请实施例的上述信息处理装置1800的相关描述可以参照本申请实施例的信息处理方法的相关描述进行理解。
图19是本申请实施例提供的一种信息处理设备示意性结构图。该信息处理设备可以是终端设备，也可以是网络设备。图19所示的信息处理设备1900包括处理器1910和存储器1920，所述存储器1920存储有可在处理器1910上运行的计算机程序，所述处理器1910执行所述程序时实现本申请实施例中的方法。
处理器1910可以从存储器中调用并运行计算机程序，以实现本申请实施例中的方法。在一些实施例中，如图19所示，信息处理设备1900还可以包括存储器1920。其中，处理器1910可以从存储器1920中调用并运行计算机程序，以实现本申请实施例中的方法。
其中，存储器1920可以是独立于处理器1910的一个单独的器件，也可以集成在处理器1910中。
在一些实施例中，如图19所示，信息处理设备1900还可以包括收发器1930，处理器1910可以控制该收发器1930与其他设备进行通信，具体地，可以向其他设备发送信息或数据，或接收其他设备发送的信息或数据。
其中，收发器1930可以包括发射机和接收机。收发器1930还可以进一步包括天线，天线的数量可以为一个或多个。
图20是本申请实施例的芯片的示意性结构图。图20所示的芯片2000包括处理器2010，处理器2010可以从存储器中调用并运行计算机程序，以实现本申请实施例中的方法。
在一些实施例中，如图20所示，芯片2000还可以包括存储器2020。其中，处理器2010可以从存储器2020中调用并运行计算机程序，以实现本申请实施例中的方法。
其中，存储器2020可以是独立于处理器2010的一个单独的器件，也可以集成在处理器2010中。
在一些实施例中，该芯片2000还可以包括输入接口2030。其中，处理器2010可以控制该输入接口2030与其他设备或芯片进行通信，具体地，可以获取其他设备或芯片发送的信息或数据。
在一些实施例中，该芯片2000还可以包括输出接口2040。其中，处理器2010可以控制该输出接口2040与其他设备或芯片进行通信，具体地，可以向其他设备或芯片输出信息或数据。
在一些实施例中，该芯片可应用于本申请实施例中的信息处理设备，并且该芯片可以实现本申请实施例的各个方法中由信息处理设备实现的相应流程，为了简洁，在此不再赘述。
应理解，本申请实施例提到的芯片还可以称为系统级芯片，系统芯片，芯片系统或片上系统芯片等。
本申请实施例还提供了一种计算机存储介质，所述计算机存储介质存储有一个或者多个程序，所述一个或者多个程序可被一个或者多个处理器执行，以实现本申请实施例中的方法。
在一些实施例中，该计算机可读存储介质可应用于本申请实施例中的信息处理设备，并且该计算机程序使得计算机执行本申请实施例的各个方法中由信息处理设备实现的相应流程，为了简洁，在此不再赘述。
本申请实施例还提供了一种计算机程序产品，所述计算机程序产品包括计算机存储介质，所述计算机存储介质存储计算机程序，所述计算机程序包括能够由至少一个处理器执行的指令，当所述指令由所述至少一个处理器执行时实现本申请实施例中的方法。
在一些实施例中，该计算机程序产品可应用于本申请实施例中的信息处理设备，并且该计算机程序指令使得计算机执行本申请实施例的各个方法中由信息处理设备实现的相应流程，为了简洁，在此不再赘述。
本申请实施例还提供了一种计算机程序所述计算机程序使得计算机执行本申请实施例中的方法。
在一些实施例中，该计算机程序可应用于本申请实施例中的信息处理设备，当该计算机程序在计算机上运行时，使得计算机执行本申请实施例的各个方法中由信息处理设备实现的相应流程，为了简洁，在此不再赘述。
本申请实施例的处理器可能是一种集成电路芯片，具有信号的处理能力。在实现过程中，上述方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器可以包括以下任一个或多个的集成：通用处理器、特定用途集成电路(Application Specific Integrated Circuit，ASIC)、数字信号处理器(Digital Signal Processor，DSP)、数字信号处理装置(Digital Signal Processing Device，DSPD)、可编程逻辑装置(Programmable Logic Device，PLD)、现场可编程门阵列(Field Programmable Gate Array，FPGA)、中央处理器(Central Processing Unit，CPU)、图形处理器(Graphics Processing Unit，GPU)、嵌入式神经网络处理器(neural-network processing units，NPU)、控制器、微控制器、微处理器、可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法的步骤。
可以理解，本申请实施例中的存储器计算机存储介质可以是易失性存储器或非易失性存储器，或可包括易失性和非易失性存储器两者。其中，非易失性存储器可以是只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable ROM，PROM)、可擦除可编程只读存储器(Erasable PROM，EPROM)、电可擦除可编程只读存储器(Electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(Static RAM，SRAM)、动态随机存取存储器(Dynamic RAM，DRAM)、同步动态随机存取存储器(Synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(Synchlink DRAM，SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM，DR RAM)。应注意，本文描述的系统和方法的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
应理解，上述存储器计算机存储介质为示例性但不是限制性说明，例如，本申请实施例中的存储器还可以是静态随机存取存储器(static RAM，SRAM)、动态随机存取存储器(dynamic RAM，DRAM)、同步动态随机存取存储器(synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(double data rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synch link DRAM，SLDRAM)以及直接内存总线随机存取存储器(Direct Rambus RAM，DR RAM)等等。也就是说，本申请实施例中的存储器旨在包括但不限于这些和任意其它适合类型的存储器。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能 够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(Read-Only Memory，)ROM、随机存取存储器(Random Access Memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应所述以权利要求的保护范围为准。

Claims (35)

  1. 一种信息处理方法，所述方法包括：
    信息处理设备的第一处理模块接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；
    所述第二处理模块对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；
    所述第一处理模块利用所述第一信息和所述第三信息，对所述第一模型进行训练。
  2. 根据权利要求1所述的方法，其中，所述对所述第二信息进行处理得到第三信息，包括：
    对所述第二信息采用第一处理方式和/或第一处理参数进行处理，得到所述第三信息；
    所述第一处理方式和/或所述第一处理参数，与所述信息发送设备中的目标处理模块对接收到的第四信息所采用的第二处理方式和/或第二处理参数相同；所述目标处理模块对所述第四信息处理得到第五信息；
    所述第二信息是所述第四信息的估计信息；所述第一信息是所述第五信息经过空口传输得到的。
  3. 根据权利要求1或2所述的方法，其中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：
    利用所述第一信息和所述第三信息，确定余弦相似度和/或均方误差；
    利用所述余弦相似度和/或所述均方误差，对所述第一模型进行训练。
  4. 根据权利要求1至3任一项所述的方法，其中，所述第一信息包括一个或多个子信息；所述方法还包括：
    所述第一处理模块获取第一指示信息；所述第一指示信息指示在每个训练周期训练一次所述第一模型；所述每个训练周期包括一个或多个传输周期；
    所述第一处理模块将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
  5. 根据权利要求4所述的方法，其中，所述第一指示信息还指示以下至少之一：
    训练周期、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
  6. 根据权利要求1至5任一项所述的方法，其中，所述第一信息包括一个或多个子信息；所述方法还包括：
    所述第一处理模块获取第二指示信息；所述第二指示信息指示在N个训练周期中每个训练周期训练一次所述第一模型；所述训练周期包括一个或多个传输周期；所述N为大于或等于1的整数；
    所述第一处理模块将所述每个训练周期包括的每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
  7. 根据权利要求6所述的方法，其中，所述第二指示信息还指示以下至少之一：
    所述N个训练周期的起始时间、训练周期、所述N、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
  8. 根据权利要求1至7任一项所述的方法，其中，所述第一信息包括一个或多个子信息；所述方法还包括：
    所述第一处理模块获取第三指示信息；所述第三指示信息指示训练一次所述第一模型；
    所述第一处理模块将一个或多个传输周期中每个传输周期接收到的至少部分信息确定为子信息，得到所述一个或多个子信息；所述一个或多个子信息和所述一个或多个传输周期一一对应。
  9. 根据权利要求8所述的方法，其中，所述第三指示信息还指示以下至少之一：
    所述一个或多个传输周期的起始时间、所述传输周期的周期数、所述第一信息的大小、每个所述子信息的大小，所述每个所述子信息的大小与所述每个传输周期中获取的信息的大小的比值。
  10. 根据权利要求1至9任一项所述的方法，其中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：
    利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一均方误差；
    对所述第一均方误差进行归一化处理，得到第一归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第一归一化结果确定为所述各个样本的第一权重；
    利用所述训练样本和所述第一权重，对所述第一模型进行训练。
  11. 根据权利要求1至9任一项所述的方法，其中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：
    利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一标准均方误差；
    对所述第一标准均方误差进行归一化处理，得到第二归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第二归一化结果确定为所述各个样本的第二权重；
    利用所述训练样本和所述第二权重，对所述第一模型进行训练。
  12. 根据权利要求1至9任一项所述的方法，其中，所述利用所述第一信息和所述第三信息，对所述第一模型进行训练，包括：
    对所述第一信息进行解量化处理，得到第一解量化结果；
    对所述第三信息进行解量化处理，得到第二解量化结果；
    利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练。
  13. 根据权利要求12所述的方法，其中，所述利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练，包括：
    利用所述第一解量化结果和所述第二解量化结果，确定对应训练样本中各个样本的第二均方误差或者第二标准均方误差；
    对所述第二均方误差或者所述第二标准均方误差进行归一化处理，得到第三归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第三归一化结果确定为所述各个样本的第三权重；
    利用所述训练样本和所述第三权重，对所述第一模型进行训练。
  14. 根据权利要求1至13任一项所述的方法，其中，所述第一模型包括信道估计模型；
    所述第一信息包括接收到的第一导频信息集合；
    所述第二信息包括第一信道估计矩阵；
    所述第三信息包括所述第一导频信息集合的估计信息；
    所述对所述第二信息进行处理得到第三信息，包括：将所述第一信道估计矩阵中每个元素，与导频序列矩阵中每个元素对应相乘，得到所述第一导频信息集合的估计信息。
  15. 根据权利要求14所述的方法，其中，所述方法还包括：
    所述第一处理模块向所述信息处理设备的第一编码模块发送所述第一信道估计矩阵；
    所述第一编码模块对所述第一信道估计矩阵进行编码处理，得到第一比特流向量集合；
    所述第一编码模块向信息发送设备发送所述第一比特流向量集合。
  16. 根据权利要求1至13任一项所述的方法，其中，所述第一模型包括解码器模型；所述第二处理模块包括第二编码模块；
    所述第一信息包括第二比特流向量集合；
    所述第二信息包括第二信道估计矩阵；
    所述第三信息包括所述第二比特流向量集合的估计信息；
    所述第二处理模块对所述第二信息进行处理得到第三信息，包括：所述第二编码模块对所述第二信道估计矩阵进行编码处理，得到所述第二比特流向量集合的估计信息。
  17. 根据权利要求1至13任一项所述的方法，其中，所述第一模型包括接收机模型；所述第二处理模块包括发射机；
    所述第一信息包括接收到的第二导频信息集合和接收到的第一数据信息集合；
    所述第二信息包括信源的估计信息；
    所述第三信息包括所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息；
    所述第二处理模块对所述第二信息进行处理得到第三信息，包括：所述发射机对所述信源的估计信息采用发射机进行处理，得到处理结果；所述处理结果用于：确定所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
  18. 根据权利要求17所述的方法，其中，所述处理结果包括发送信息估计矩阵；所述发送信息估计矩阵包括：信息发送设备发送的第三导频信息的估计信息和发送的第二数据信息集合的估计信息；
    所述方法还包括：
    所述信息处理设备的信道估计模块对所述第二导频信息集合和所述第一数据信息集合进行信道估计处理，得到第三信道估计矩阵；
    所述信息处理设备的第三处理模块将所述发送信息估计矩阵中每个元素，与所述第三信道估计矩阵中的每个元素对应相乘，得到所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
  19. 一种信息处理装置，所述装置包括：
    第一处理模块，用于接收第一信息，利用所述第一处理模块中的第一模型对所述第一信息进行处理得到第二信息，向所述信息处理设备的第二处理模块发送所述第二信息；
    第二处理模块，用于对所述第二信息进行处理得到第三信息；其中，所述第三信息是所述第一信息的估计信息；
    所述第一处理模块，还用于利用所述第一信息和所述第三信息，对所述第一模型进行训练。
  20. 根据权利要求19所述的装置，其中，
    所述第二处理模块，还用于对所述第二信息采用第一处理方式和/或第一处理参数进行处理，得到所述第三信息；
    所述第一处理方式和/或所述第一处理参数，与所述信息发送设备中的目标处理模块对接收到的第四信息所采用的第二处理方式和/或第二处理参数相同；所述目标处理模块对所述第四信息处理得到第五信息；
    所述第二信息是所述第四信息的估计信息；所述第一信息是所述第五信息经过空口传输得到的。
  21. 根据权利要求19或20所述的装置，其中，所述第一处理模块，还用于：
    利用所述第一信息和所述第三信息，确定余弦相似度和/或均方误差；
    利用所述余弦相似度和/或所述均方误差，对所述第一模型进行训练。
  22. 根据权利要求19至21任一项所述的装置，其中，所述第一处理模块，还用于：
    利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一均方误差；
    对所述第一均方误差进行归一化处理，得到第一归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第一归一化结果确定为所述各个样本的第一权重；
    利用所述训练样本和所述第一权重，对所述第一模型进行训练。
  23. 根据权利要求19至21任一项所述的装置，其中，所述第一处理模块，还用于：
    利用所述第一信息和所述第三信息，确定对应训练样本中各个样本的第一标准均方误差；
    对所述第一标准均方误差进行归一化处理，得到第二归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第二归一化结果确定为所述各个样本的第二权重；
    利用所述训练样本和所述第二权重，对所述第一模型进行训练。
  24. 根据权利要求19至21任一项所述的装置，其中，所述第一处理模块，还用于：
    对所述第一信息进行解量化处理，得到第一解量化结果；
    对所述第三信息进行解量化处理，得到第二解量化结果；
    利用所述第一解量化结果和所述第二解量化结果，对所述第一模型进行训练。
  25. 根据权利要求24所述的装置，其中，所述第一处理模块，还用于：
    利用所述第一解量化结果和所述第二解量化结果，确定对应训练样本中各个样本的第二均方误差或者第二标准均方误差；
    对所述第二均方误差或者所述第二标准均方误差进行归一化处理，得到第三归一化结果；
    将所述第一信息和所述第二信息确定为所述训练样本，所述第三归一化结果确定为所述各个样本的第三权重；
    利用所述训练样本和所述第三权重，对所述第一模型进行训练。
  26. 根据权利要求19至25任一项所述的装置，其中，所述第一模型包括信道估计模型；
    所述第一信息包括接收到的第一导频信息集合；
    所述第二信息包括第一信道估计矩阵；
    所述第三信息包括所述第一导频信息集合的估计信息；
    所述第二处理模块，还用于：将所述第一信道估计矩阵中每个元素，与导频序列矩阵中每个元素对应相乘，得到所述第一导频信息集合的估计信息。
  27. 根据权利要求26所述的装置，其中，所述装置还包括：
    第一编码模块，用于对所述第一信道估计矩阵进行编码处理，得到第一比特流向量集合；
    所述第一编码模块，还用于向信息发送设备发送所述第一比特流向量集合。
  28. 根据权利要求19至25任一项所述的装置，其中，所述第一模型包括解码器模型；所述第二处理模块包括第二编码模块；
    所述第一信息包括第二比特流向量集合；
    所述第二信息包括第二信道估计矩阵；
    所述第三信息包括所述第二比特流向量集合的估计信息；
    所述第二编码模块，用于对所述第二信道估计矩阵进行编码处理，得到所述第二比特流向量集合的估计信息。
  29. 根据权利要求19至25任一项所述的装置，其中，所述第一模型包括接收机模型；所述第二处理模块包括发射机；
    所述第一信息包括接收到的第二导频信息集合和接收到的第一数据信息集合；
    所述第二信息包括信源的估计信息；
    所述第三信息包括所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息；
    所述发射机，用于对所述信源的估计信息进行处理，得到处理结果；所述处理结果用于：确定所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
  30. 根据权利要求29所述的装置，其中，所述处理结果包括发送信息估计矩阵；所述发送信息估计矩阵包括：信息发送设备发送的第三导频信息的估计信息和发送的第二数据信息集合的估计信息；
    所述装置还包括：
    信道估计模块，用于对所述第二导频信息集合和所述第一数据信息集合进行信道估计处理，得到第三信道估计矩阵；
    第三处理模块，用于将所述发送信息估计矩阵中每个元素，与所述第三信道估计矩阵中的每个元素对应相乘，得到所述第二导频信息集合的估计信息和所述第一数据信息集合的估计信息。
  31. 一种信息处理设备，包括：存储器和处理器，
    所述存储器存储有可在处理器上运行的计算机程序，
    所述处理器执行所述程序时实现权利要求1至18任一项所述方法。
  32. 一种计算机存储介质，所述计算机存储介质存储有一个或者多个程序，所述一个或者多个程序可被一个或者多个处理器执行，以实现权利要求1至18任一项所述方法。
  33. 一种芯片，包括：处理器，用于从存储器中调用并运行计算机程序，使得安装有所述芯片的设备执行如权利要求1至18任一项所述方法。
  34. 一种计算机程序产品，所述计算机程序产品包括计算机存储介质，所述计算机存储介质存储计算机程序，所述计算机程序包括能够由至少一个处理器执行的指令，当所述指令由所述至少一个处理器执行时实现权利要求1至18任一项所述方法。
  35. 一种计算机程序，所述计算机程序使得计算机执行如权利要求1至18任一项所述方法。
PCT/CN2021/123777 2021-10-14 2021-10-14 信息处理方法、装置、设备、介质、芯片、产品及程序 WO2023060503A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202180101608.1A CN117897931A (zh) 2021-10-14 2021-10-14 信息处理方法、装置、设备、介质、芯片、产品及程序
PCT/CN2021/123777 WO2023060503A1 (zh) 2021-10-14 2021-10-14 信息处理方法、装置、设备、介质、芯片、产品及程序

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2021/123777 WO2023060503A1 (zh) 2021-10-14 2021-10-14 信息处理方法、装置、设备、介质、芯片、产品及程序

Publications (1)

Publication Number Publication Date
WO2023060503A1 true WO2023060503A1 (zh) 2023-04-20

Family

ID=85987222

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/123777 WO2023060503A1 (zh) 2021-10-14 2021-10-14 信息处理方法、装置、设备、介质、芯片、产品及程序

Country Status (2)

Country Link
CN (1) CN117897931A (zh)
WO (1) WO2023060503A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111669344A (zh) * 2020-06-01 2020-09-15 西北工业大学 一种基于深度学习的时变ofdm系统信号检测方法
CN112054863A (zh) * 2019-06-06 2020-12-08 华为技术有限公司 一种通信方法及装置
CN112152948A (zh) * 2019-06-28 2020-12-29 华为技术有限公司 一种无线通信处理的方法和装置
WO2021041862A1 (en) * 2019-08-30 2021-03-04 Idac Holdings, Inc. Deep learning aided mmwave mimo blind detection schemes
WO2021041551A2 (en) * 2019-08-26 2021-03-04 Board Of Regents, The University Of Texas System Autoencoder-based error correction coding for low-resolution communication
CN113422745A (zh) * 2021-06-17 2021-09-21 中山大学 一种基于深度学习的空海无线信道估计方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112054863A (zh) * 2019-06-06 2020-12-08 华为技术有限公司 一种通信方法及装置
CN112152948A (zh) * 2019-06-28 2020-12-29 华为技术有限公司 一种无线通信处理的方法和装置
WO2021041551A2 (en) * 2019-08-26 2021-03-04 Board Of Regents, The University Of Texas System Autoencoder-based error correction coding for low-resolution communication
WO2021041862A1 (en) * 2019-08-30 2021-03-04 Idac Holdings, Inc. Deep learning aided mmwave mimo blind detection schemes
CN111669344A (zh) * 2020-06-01 2020-09-15 西北工业大学 一种基于深度学习的时变ofdm系统信号检测方法
CN113422745A (zh) * 2021-06-17 2021-09-21 中山大学 一种基于深度学习的空海无线信道估计方法

Also Published As

Publication number Publication date
CN117897931A (zh) 2024-04-16

Similar Documents

Publication Publication Date Title
WO2021217519A1 (zh) 用于调整神经网络的方法和装置
WO2022001822A1 (zh) 获取神经网络的方法和装置
US20240072927A1 (en) Signal processing method, communication device, and communication system
WO2022151064A1 (zh) 信息发送方法、信息接收方法、装置、设备及介质
US20240048207A1 (en) Method and apparatus for transmitting and receiving feedback information based on artificial neural network
WO2023126007A1 (zh) 信道信息传输方法及装置
WO2023060503A1 (zh) 信息处理方法、装置、设备、介质、芯片、产品及程序
CN109889247B (zh) 适用于窄带物联网的低开销动态反馈安全传输方法及系统
WO2024007191A1 (zh) 模型训练方法及装置、样本数据生成方法及装置、电子设备
WO2023283785A1 (zh) 信号处理的方法及接收机
WO2022257121A1 (zh) 通信方法、设备及存储介质
WO2022236785A1 (zh) 信道信息的反馈方法、收端设备和发端设备
CN115694722A (zh) 一种通信方法及装置
WO2023097645A1 (zh) 数据获取方法、装置、设备、介质、芯片、产品及程序
WO2023115254A1 (zh) 处理数据的方法及装置
WO2024037321A1 (zh) Ai模型训练方法、装置、设备及可读存储介质
WO2022217502A1 (zh) 信息处理方法、装置、通信设备及存储介质
WO2023015499A1 (zh) 无线通信的方法和设备
WO2024027682A1 (zh) 特征信息传输方法、转换信息确定方法、装置和通信设备
WO2023004638A1 (zh) 信道信息反馈的方法、发端设备和收端设备
WO2024020793A1 (zh) 信道状态信息csi反馈的方法、终端设备和网络设备
WO2023070675A1 (zh) 数据处理的方法及装置
WO2023004563A1 (zh) 获取参考信号的方法及通信设备
WO2024108356A1 (zh) Csi反馈的方法、发端设备和收端设备
WO2024032775A1 (zh) 量化方法和装置

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21960240

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 202180101608.1

Country of ref document: CN

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21960240

Country of ref document: EP

Kind code of ref document: A1