WO2023185978A1 - 信道特征信息上报及恢复方法、终端和网络侧设备 - Google Patents

信道特征信息上报及恢复方法、终端和网络侧设备 Download PDF

Info

Publication number
WO2023185978A1
WO2023185978A1 PCT/CN2023/084962 CN2023084962W WO2023185978A1 WO 2023185978 A1 WO2023185978 A1 WO 2023185978A1 CN 2023084962 W CN2023084962 W CN 2023084962W WO 2023185978 A1 WO2023185978 A1 WO 2023185978A1
Authority
WO
WIPO (PCT)
Prior art keywords
coefficients
network
orthogonal basis
channel
information
Prior art date
Application number
PCT/CN2023/084962
Other languages
English (en)
French (fr)
Inventor
任千尧
Original Assignee
维沃移动通信有限公司
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 维沃移动通信有限公司 filed Critical 维沃移动通信有限公司
Publication of WO2023185978A1 publication Critical patent/WO2023185978A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/06—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station
    • H04B7/0613—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission
    • H04B7/0615—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal
    • H04B7/0619—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the transmitting station using simultaneous transmission of weighted versions of same signal using feedback from receiving side
    • H04B7/0621—Feedback content
    • H04B7/0626—Channel coefficients, e.g. channel state information [CSI]
    • 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
    • H04L25/0202—Channel estimation
    • H04L25/0224—Channel estimation using sounding signals
    • H04L25/0228—Channel estimation using sounding signals with direct estimation from sounding signals
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/06—Optimizing the usage of the radio link, e.g. header compression, information sizing, discarding information

Definitions

  • This application belongs to the field of communication technology, and specifically relates to a channel characteristic information reporting and recovery method, terminal and network side equipment.
  • CSI Channel State Information
  • the base station can precode the CSI Reference Signal (CSI-RS) in advance and send the coded CSI-RS to the terminal. What the terminal sees is the coded CSI-RS.
  • the terminal needs to input the channel information of the entire channel matrix into a large AI network model, so that the AI network model outputs the coefficients of all orthogonal basis vectors that need to be reported.
  • the encoding and decoding of CSI information is for the entire channel, which requires a relatively large AI network model.
  • the AI network model needs to be retrained or even a different AI network model needs to be used. Therefore, it is necessary to Larger transmission overhead configures all AI network models in advance.
  • Embodiments of the present application provide a channel characteristic information reporting and recovery method, terminal and network side equipment, which can use an AI network model to select a target orthogonal basis vector for which coefficients need to be reported, or use an AI network model to determine the specified target orthogonal basis vector.
  • the coefficient of the orthogonal basis vector enables the channel information to be reported with the granularity of the orthogonal basis vector, and the required AI network model is relatively small, thereby reducing the cost of transmitting the AI network model between the terminal and the network side device, and can even
  • the network side device uses the AI network model to determine the orthogonal basis vectors of coefficients that need to be reported, and instructs the terminal to report these coefficients, which can avoid transmitting the AI network model between the terminal and the network side device.
  • a method for reporting channel characteristic information includes:
  • the terminal obtains the first channel information of the target channel
  • the terminal obtains N coefficients according to the first channel information, wherein the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients are related to the N first AI network models.
  • the models correspond one to one, Alternatively, the N coefficients are coefficients of N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the terminal sends first channel characteristic information to the network side device, where the first channel characteristic information includes the N coefficients.
  • a device for reporting channel characteristic information which is applied to a terminal.
  • the device includes:
  • the first acquisition module is used to acquire the first channel information of the target channel
  • the second acquisition module is configured to acquire N coefficients according to the first channel information, wherein the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients are consistent with the N
  • the first AI network model has a one-to-one correspondence, or the N coefficients are coefficients of N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the first sending module is configured to send first channel characteristic information to the network side device, where the first channel characteristic information includes the N coefficients.
  • a channel characteristic information recovery method including:
  • the network side device receives the first channel characteristic information from the terminal, wherein the first channel characteristic information includes N coefficients, and the N coefficients are obtained by processing the first channel information using N first AI network models respectively.
  • the coefficients, or the N coefficients are the coefficients of the first channel information projected on the N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the network side device performs recovery processing on the first channel characteristic information to obtain the first channel information.
  • a device for recovering channel characteristic information which is applied to network side equipment.
  • the device includes:
  • the first receiving module is configured to receive the first channel characteristic information from the terminal, where the first channel characteristic information includes N coefficients, and the N coefficients are the first channel characteristics using N first AI network models respectively.
  • the coefficients obtained by processing the information, or the N coefficients are the coefficients of the first channel information projected on the N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the first processing module is used to restore the first channel characteristic information to obtain the first channel information.
  • a terminal in a fifth aspect, includes a processor and a memory.
  • the memory stores programs or instructions that can be run on the processor.
  • the program or instructions are executed by the processor, the following implementations are implemented: The steps of the method described in one aspect.
  • a terminal including a processor and a communication interface, wherein the communication interface is used to obtain first channel information of a target channel; the processor is used to obtain N coefficients according to the first channel information , wherein the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients correspond to the N first AI network models one-to-one, or the N coefficients are determined by using The coefficients of N target orthogonal basis vectors selected by the second AI network model, N is an integer greater than or equal to 1; the communication interface is also used to send first channel characteristic information to the network side device, the first channel characteristic The information includes the N coefficients.
  • a network side device in a seventh aspect, includes a processor and a memory.
  • the memory stores programs or instructions that can be run on the processor.
  • the program or instructions are executed by the processor.
  • the time is as follows: The steps of the method described in three aspects.
  • a network side device including a processor and a communication interface, wherein the communication interface is used to receive first channel characteristic information from a terminal, wherein the first channel characteristic information includes N coefficients , the N coefficients are coefficients obtained by respectively using N first AI network models to process the first channel information, or the N coefficients are coefficients obtained by using the second AI network model to project the first channel information.
  • the coefficients of N target orthogonal basis vectors, N is an integer greater than or equal to 1; the processor is used to restore the first channel characteristic information to obtain the first channel information.
  • a ninth aspect provides a communication system, including: a terminal and a network side device.
  • the terminal can be configured to perform the steps of the channel characteristic information reporting method described in the first aspect.
  • the network side device can be configured to perform the steps of the channel characteristic information reporting method as described in the first aspect. The steps of the channel characteristic information recovery method described in the three aspects.
  • a readable storage medium is provided. Programs or instructions are stored on the readable storage medium. When the programs or instructions are executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method are implemented as described in the first aspect. The steps of the method described in the third aspect.
  • a chip in an eleventh aspect, includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions to implement the method described in the first aspect. method, or implement a method as described in the third aspect.
  • a computer program/program product is provided, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement as described in the first aspect
  • the steps of the channel characteristic information reporting method, or the computer program/program product is executed by at least one processor to implement the steps of the channel characteristic information recovery method as described in the third aspect.
  • the terminal obtains the first channel information of the target channel; the terminal obtains N coefficients according to the first channel information, wherein the N coefficients are determined by using N first AI network models respectively.
  • the N coefficients correspond to the N first AI network models one-to-one, or the N coefficients are the coefficients of the N target orthogonal basis vectors selected using the second AI network model, and N is An integer greater than or equal to 1;
  • the terminal sends first channel characteristic information to the network side device, and the first channel characteristic information includes the N coefficients.
  • the terminal can use the AI network model to select the target orthogonal basis vector whose coefficients need to be reported, or use the AI network model to determine the coefficients of the specified target orthogonal basis vector, so that the channel information can be reported based on the orthogonal basis vector.
  • Granularity the required AI network model is relatively small, thus reducing the cost of transmitting the AI network model between the terminal and the network side device.
  • the AI network model can even be used on the network side device to determine the orthogonal basis vectors of coefficients that need to be reported. And instructing the terminal to report these coefficients can avoid transmitting the AI network model between the terminal and the network side device.
  • Figure 1 is a schematic structural diagram of a wireless communication system to which embodiments of the present application can be applied;
  • Figure 2 is a flow chart of a method for reporting channel characteristic information provided by an embodiment of the present application
  • Figure 3 is a schematic diagram of the architecture of the neural network model
  • Figure 4 is a schematic diagram of a neuron
  • Figure 5 is a flow chart of a method for recovering channel characteristic information provided by an embodiment of the present application.
  • Figure 6 is a schematic structural diagram of a device for reporting channel characteristic information provided by an embodiment of the present application.
  • Figure 7 is a schematic structural diagram of a device for recovering channel characteristic information provided by an embodiment of the present application.
  • Figure 8 is a schematic structural diagram of a communication device provided by an embodiment of the present application.
  • Figure 9 is a schematic structural diagram of a terminal provided by an embodiment of the present application.
  • Figure 10 is a schematic structural diagram of a network side device provided by an embodiment of the present application.
  • first, second, etc. in the description and claims of this application are used to distinguish similar objects and are not used to describe a specific order or sequence. It is to be understood that the terms so used are interchangeable under appropriate circumstances so that the embodiments of the present application can be practiced in sequences other than those illustrated or described herein, and that "first" and “second” are distinguished objects It is usually one type, and the number of objects is not limited.
  • the first object can be one or multiple.
  • “and/or” in the description and claims indicates at least one of the connected objects, and the character “/" generally indicates that the related objects are in an "or” relationship.
  • LTE Long Term Evolution
  • LTE-Advanced, LTE-A Long Term Evolution
  • LTE-A Long Term Evolution
  • CDMA Code Division Multiple Access
  • TDMA Time Division Multiple Access
  • FDMA Frequency Division Multiple Access
  • OFDMA Orthogonal Frequency Division Multiple Access
  • SC-FDMA Single-carrier Frequency Division Multiple Access
  • NR New Radio
  • FIG. 1 shows a block diagram of a wireless communication system to which embodiments of the present application are applicable.
  • the wireless communication system includes a terminal 11 and a network side device 12.
  • the terminal 11 may be a mobile phone, a tablet computer (Tablet Personal Computer), a laptop computer (Laptop Computer), or a notebook computer, a personal digital assistant (Personal Digital Assistant, PDA), a palmtop computer, a netbook, or a super mobile personal computer.
  • Tablet Personal Computer Tablet Personal Computer
  • laptop computer laptop computer
  • PDA Personal Digital Assistant
  • PDA Personal Digital Assistant
  • UMPC ultra-mobile personal computer
  • UMPC mobile Internet device
  • Mobile Internet Device MID
  • AR augmented reality
  • VR virtual reality
  • robots wearable devices
  • Wearable Device vehicle user equipment
  • VUE Vehicle User Equipment
  • pedestrian terminal Pedestrian User Equipment, PUE
  • smart home with Home equipment with wireless communication functions, such as refrigerators, TVs, washing machines or furniture, etc.
  • PCs personal computers
  • teller machines or self-service machines and other terminal-side devices such as refrigerators, TVs, washing machines or furniture, etc.
  • the network side device 12 may include an access network device or a core network device, where the access network device may also be called a radio access network device, a radio access network (Radio Access Network, RAN), a radio access network function or a wireless device.
  • Access network equipment may include base stations, Wireless Local Area Networks (WLAN) access points or WiFi nodes, etc.
  • the base stations may be called Node B, Evolved Node B (eNB), access point, base transceiver station ( Base Transceiver Station (BTS), radio base station, radio transceiver, Basic Service Set (BSS), Extended Service Set (ESS), home B-node, home evolved B-node, transmitting and receiving point ( Transmitting Receiving Point (TRP) or some other appropriate term in the field, as long as the same technical effect is achieved, the base station is not limited to specific technical terms. It should be noted that in the embodiment of this application, only in the NR system The base station is introduced as an example, and the specific type of base station is not limited.
  • the transmitter can optimize signal transmission based on CSI to better match the channel status.
  • CQI Channel Quality Indicator
  • MCS Modulation and Coding Scheme
  • PMI Precoding Matrix Indicator
  • eigen beamforming eigen beamforming
  • MIMO Multi-Input Multi-Output
  • the network side device sends CSI reference signals (CSI-Reference Signals, CSI-RS) on certain time-frequency resources of a certain time slot (slot).
  • CSI-RS CSI-Reference Signals
  • the terminal performs channel estimation based on the CSI-RS and calculates the channel on this slot.
  • Information, the PMI is fed back to the base station through the codebook.
  • the network side device combines the channel information based on the codebook information fed back by the terminal. Before the terminal reports the CSI next time, the network side device uses this channel information to perform data precoding and multi-user scheduling. .
  • the terminal can change the PMI reported on each subband to report PMI according to the delay (delay domain, that is, frequency domain). Since the channels in the delay domain are more concentrated, PMI with less delay can be approximated The PMI of all subbands can be regarded as reporting after compressing the delay field information.
  • the network side device can precode the CSI-RS in advance and send the coded CSI-RS to the terminal. What the terminal sees is the channel corresponding to the coded CSI-RS. The terminal only needs to Just select several stronger ports from the ports indicated by the network-side device and report the coefficients corresponding to these ports.
  • AI network models have many implementation methods, such as: neural networks, decision trees, support vector machines, Bayesian classifier etc.
  • the AI network model is a neural network as an example, but the specific type of the AI network model is not limited.
  • the terminal can estimate the CSI Reference Signal (CSI-RS) or the Tracking Reference Signal (TRS), perform calculations based on the estimated channel information, and obtain the calculated channel information. Then, the calculated channel information is calculated.
  • the channel information or the original estimated channel information is encoded through the encoding network model to obtain the encoding result, and finally the encoding result is sent to the base station.
  • the base station can input it into the decoding network model and use the decoding network model to restore the channel information.
  • the network side sends precoded CSI-RS, the terminal receives the channel matrix, and selects 2L air domain orthogonal basis vectors, Mv delay domain orthogonal basis vectors, and then The selected orthogonal basis vectors and corresponding coefficients are reported, and the network side device can restore the channel information based on the orthogonal basis vectors and corresponding coefficients, where the delay domain corresponds to the frequency domain.
  • the process of encoding and decoding channel information using AI network models is for the entire channel. Therefore, the amount of data in the AI network model will be very large, and when transmitting the AI network model, it will generate Large overhead.
  • channel information of different lengths corresponds to different AI network models.
  • the network side device needs to train the AI network model separately for each length of channel information, or even configure different AI network models separately. This increases the amount of calculation, occupied resources, and delay caused by training and configuring the AI network model between the terminal and the network-side device.
  • the terminal can determine N coefficients by using N first AI network models, or use N target orthogonal basis vectors selected by the second AI network model, and determine the respective values of the N target orthogonal basis vectors.
  • Coefficients which use coefficients of orthogonal basis vectors as granularity, make the data volume of the first AI network model or the second AI network model small, thereby reducing the need for transmission of the first AI network model or the second AI network model between the network side device and the terminal. 2.
  • Resource consumption caused by the AI network model by using the coefficients of orthogonal basis vectors as the granularity, it is also possible to reduce the number of reported coefficients on the basis of meeting the requirement for channel information reporting, thereby achieving the beneficial effect of reducing the channel information reporting overhead.
  • the encoding process of the channel information in the embodiment of the present application may include the following steps:
  • Step 1 The terminal detects CSI-RS or TRS at the time-frequency domain location specified by the network, and performs channel estimation to obtain the first channel information;
  • Step 2 The terminal encodes the K groups of first channel information into first channel characteristic information through the first AI network model (i.e., the encoding AI network model) respectively;
  • Step 3 The terminal combines part or all of the first channel characteristic information and other control information into uplink control information (Uplink Control Information, UCI), or uses part or all of the first channel characteristic information as UCI;
  • UCI Uplink Control Information
  • Step 4 The terminal divides the UCI according to the length of the UCI and adds cyclic redundancy check (CRC) bits;
  • CRC cyclic redundancy check
  • Step 5 The terminal performs channel coding on the UCI with CRC bits added
  • Step 6 The terminal performs rate matching on UCI
  • Step 7 The terminal performs code block association on UCI
  • Step 8 The terminal maps the UCI to the Physical Uplink Control Channel (PUCCH) or the Physical Uplink Shared Channel (PUSCH) for reporting.
  • PUCCH Physical Uplink Control Channel
  • PUSCH Physical Uplink Shared Channel
  • channel characteristic information reporting method channel characteristic information recovery method, channel characteristic information reporting device, channel characteristic information recovery device and communication equipment provided by the embodiments of the present application will be described in detail through some embodiments and application scenarios. .
  • an embodiment of the present application provides a method for reporting channel characteristic information.
  • the execution subject may be a terminal.
  • the terminal may be various types of terminals 11 listed in Figure 1, or other than those shown in Figure 1. Terminals other than the terminal types listed in the embodiment are not specifically limited here.
  • the channel characteristic information reporting method may include the following steps:
  • Step 201 The terminal obtains the first channel information of the target channel.
  • the terminal can obtain the above-mentioned first channel information by performing channel estimation on reference signals such as CSR-RS or TRS, or the above-mentioned first channel information is the terminal performing certain calculations or preprocessing on the estimated original channel information.
  • the channel information obtained later for example, the first channel information may be a precoding matrix determined based on the channel information obtained by channel estimation or a precoding matrix of a specific layer, which is not specifically limited here.
  • Step 202 The terminal obtains N coefficients according to the first channel information, wherein the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients are the same as the Nth coefficients.
  • One AI network model has a one-to-one correspondence, or the N coefficients are coefficients of N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1.
  • the above-mentioned target orthogonal basis vectors may include at least one of a spatial domain orthogonal basis vector, a frequency domain orthogonal basis vector, and a Doppler domain orthogonal basis vector, and the coefficients of the above-mentioned target orthogonal basis vectors are It may be the projection of the first channel information onto the corresponding target orthogonal basis vector.
  • the coefficient may be a numerical value or a set of numerical values.
  • the coefficient may be a complex number including a real part and an imaginary part, or the The coefficients may be real numbers including at least one amplitude and/or phase.
  • the N coefficients are coefficients determined using N first AI network models respectively, and may be one-to-one correspondence between N first AI network models and N orthogonal basis vectors.
  • the terminal obtains N coefficients according to the first channel information, which may be: the terminal uses N first AI network models to process the first channel information to obtain N coefficients.
  • the above-mentioned first AI network model can represent the corresponding orthogonal basis vector, for example: represent a spatial domain orthogonal basis vector, represent a frequency domain orthogonal basis vector, represent a Doppler domain orthogonal basis vector, or Orthogonal basis vectors that represent the union of at least two orthogonal basis vectors.
  • the first AI network model can be used to represent the information of the corresponding orthogonal basis vectors, without using specific orthogonal basis vectors to calculate the corresponding coefficients.
  • N first The AI network model can be represented as an encoding AI network model.
  • the encoding AI network model and its corresponding decoding AI network model can be jointly trained, and the corresponding orthogonal basis vector information can be reflected in the encoding AI network model and the decoding AI network model.
  • the encoding AI network model can be used to determine the coefficients of the corresponding orthogonal basis vectors, and on the network side, the decoding AI network model can be used to restore the coefficients derived from the corresponding encoding AI network model.
  • N first AI network models can be used to form N orthogonal basis vectors.
  • the first AI network model can be used to calculate the first The coefficient of channel information projected onto the orthogonal basis vector corresponding to the first AI network model.
  • the above-mentioned N first AI network models can be configured by the network side device to the terminal. Since the first AI network model only needs to calculate the coefficient of a single orthogonal basis vector, its size is very small. Therefore, on the network side device When N first AI network models are configured to the terminal, the resource overhead caused is also very small.
  • using the first AI network model to form orthogonal basis vectors can use fewer orthogonal basis vectors to represent the actual orthogonal basis vectors of the target channel, which is compared to projecting on a fixed DFT vector.
  • the number of orthogonal basis vectors of coefficients reported by the terminal can be reduced, thereby reducing the overhead of reporting channel characteristic information.
  • the N coefficients are coefficients of N target orthogonal basis vectors selected using the second AI network model.
  • the terminal obtains N coefficients based on the first channel information, which may include:
  • the terminal projects the first channel information onto N target orthogonal basis vectors to obtain N coefficients.
  • the above-mentioned second AI network model can be the AI network model used by the network side device, that is, different network side devices can configure their own second AI network models.
  • the The network side device configures its own second AI network model to the terminal, so that the terminal also uses the same second AI network model to determine N target orthogonal basis vectors, and reports the coefficients of these target orthogonal basis vectors, or the network side device
  • the terminal can be directly instructed to use the second AI network model to determine the target orthogonal basis vectors, so that the terminal does not need to use the second AI network model to determine the target orthogonal basis vectors, and can directly report these target orthogonal basis vectors. coefficient.
  • the above-mentioned second AI network model may also be an offline second AI network model agreed in the communication protocol, or an AI network model obtained by terminal training.
  • the terminal uses the second AI network model to determine the target orthogonal basis vector, or the network side device uses the target orthogonal basis vector selected by the second AI network model, and can select a larger projection. Partial orthogonal basis vectors, in this way, the terminal only reports the coefficients of these orthogonal basis vectors.
  • the channel characteristic information reporting method further includes:
  • the terminal receives first indication information from the network side device, the first indication information is used to indicate the target orthogonal basis vector, wherein the target orthogonal basis vector is adopted by the network side device.
  • the second AI network model is trained;
  • the terminal adopts the second AI network model and trains based on the first channel information to obtain the target orthogonal basis direction quantity;
  • the terminal uses the target orthogonal basis vector agreed in the communication protocol.
  • the second AI network model may be a network side device configuration.
  • different network-side devices can configure different second AI network models respectively, and determine which orthogonal basis vector coefficients need to be reported by terminals accessing the network-side device based on the respectively configured second AI network models. Therefore, the terminal is instructed to report the coefficients of these orthogonal basis vectors through the first indication information.
  • the terminal may obtain the above-mentioned first indication information from the network side device when accessing the network side device, and during the period of accessing the network side device, report the specified target correctness according to the instructions of the first indication information. Coefficients of intersection basis vectors.
  • the network side device uses the second AI network model training to obtain N target orthogonal basis vectors, when receiving the N coefficients reported by the terminal, it can determine which target each of these N coefficients corresponds to. Orthogonal basis vectors, thereby restoring the first channel information based on the N coefficients and their respective corresponding orthogonal basis vectors.
  • the second AI network model may be stipulated in the protocol or indicated by the network side device.
  • the network side device indicates the structure of the second AI network model
  • the terminal adopts the second AI network model, and obtains the target orthogonal basis vector based on the first channel information training.
  • different terminals can use their own
  • the second AI network model is used to train the respective target orthogonal basis vectors, and the coefficients of the trained target orthogonal basis vectors are reported.
  • the method further includes:
  • the terminal sends the target orthogonal basis vector to the network side device.
  • the terminal can also report the trained target orthogonal basis vector to the network side device, for example, periodically report the target orthogonal basis vector to the network side device.
  • the target orthogonal basis vector changes more frequently. Low, the terminal can report the target orthogonal basis vector at longer intervals. In this way, the network side device can learn which orthogonal basis vector coefficients the coefficients reported by the terminal are, and thereby restore the first channel information based on the N coefficients and their corresponding orthogonal basis vectors.
  • the target orthogonal basis vector may be an orthogonal basis vector obtained by training using an offline second AI network model.
  • the differences between this method three and the above-mentioned methods one and two include:
  • the target orthogonal basis vector is an orthogonal basis vector obtained by pre-training with the offline second AI network model, and will not follow changes in the network state. Therefore, there is no need to train in real time, and by agreeing on the target orthogonal basis vector in the communication protocol, both the terminal and the network side device can learn the target orthogonal basis vector. Therefore, the terminal and the network side device There is no need to interact or indicate the target orthogonal basis vector.
  • Step 203 The terminal sends first channel characteristic information to the network side device, where the first channel characteristic information includes the N coefficients.
  • the above-mentioned terminal sends the first channel characteristic information to the network side device, and may use a CSI reporting method to carry the first channel characteristic information in the CSI report to report to the network side device, where the channel Specifically, the characteristic information may be PMI information.
  • the above-mentioned first channel characteristic information can also be reported to the network side device in any other manner.
  • the first channel characteristic information is reported using CSI reporting as an example.
  • CSI reporting does not constitute a specific limitation.
  • the above N first AI network models can be AI network models pre-configured for the terminal by the network side device, and the AI network model can It includes multiple types of AI algorithm modules, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc., which are not specifically limited here, and for ease of explanation, the AI algorithm model is used in the following embodiments.
  • the neural network model is taken as an example for illustration and does not constitute a specific limitation here.
  • the neural network model includes an input layer, a hidden layer and an output layer, which can predict possible output results (Y) based on the entry and exit information (X 1 ⁇ X n ) obtained by the input layer.
  • the neural network model consists of a large number of neurons, as shown in Figure 4.
  • K represents the total number of input parameters.
  • the parameters of the neural network are optimized through optimization algorithms.
  • An optimization algorithm is a type of algorithm that can help us minimize or maximize an objective function (sometimes also called a loss function).
  • the objective function is often a mathematical combination of model parameters and data. For example, given the data The difference (f(x)-Y) between it and the true value is the loss function. Our purpose is to find appropriate W and b to minimize the value of the above loss function. The smaller the loss value, the closer our model is to the real situation.
  • error back propagation is basically based on error back propagation algorithm.
  • the basic idea of the error back propagation algorithm is that the learning process consists of two processes: forward propagation of signals and back propagation of errors.
  • the input sample is passed in from the input layer, processed layer by layer by each hidden layer, and then transmitted to the output layer. If the actual output of the output layer does not match the expected output, it will enter the error backpropagation stage.
  • Error backpropagation is to backpropagate the output error in some form to the input layer layer by layer through the hidden layer, and allocate the error to all units in each layer, thereby obtaining the error signal of each layer unit. This error signal is used as a correction for each unit. The basis for the weight.
  • This process of adjusting the weights of each layer in forward signal propagation and error back propagation is carried out over and over again.
  • the process of continuous adjustment of weights is the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level, or until a preset number of learning times.
  • the terminal uses a first AI network model to process the first channel information into N coefficients, including:
  • the terminal inputs the first information to the M first AI network models to obtain coefficients corresponding to the M target orthogonal basis vectors, wherein the M first AI network models and the M target orthogonal basis vectors are Vectors correspond one to one, M is an integer greater than or equal to N;
  • the terminal determines N coefficients from the coefficients corresponding to each of the M target orthogonal basis vectors
  • the first information includes:
  • the first coefficient of the first channel information projected in the complete orthogonal space or,
  • the precoding matrix or the precoding matrix of a specific layer obtained by precoding calculation of the first channel information; or,
  • the second coefficient, and part or all of the weighted orthogonal basis vector corresponding to the second coefficient are the weighted orthogonal basis vector corresponding to the second coefficient.
  • the above M first AI network models may be all first AI network models configured by the network side device to the terminal. According to the actual situation of the channel state, the terminal may only need to report some of the coefficients output by the first AI network model to reflect the actual orthogonal basis vectors of the target channel. At this time, the above-mentioned terminal obtains the information from the M target orthogonal basis vectors. Determining N coefficients from the corresponding coefficients of each vector may be that the terminal selects N coefficients with larger amplitudes from the coefficients of the M target orthogonal basis vectors.
  • the input to the first AI network model may be the first channel information, or the result of projecting the first channel information in the complete orthogonal space (the first coefficient), or the first coefficient and Part or all of its corresponding orthogonal basis vector weighted, for example: Discrete Fourier Transform (DFT) orthogonal space can be over-sampled or not, and the result projected on the DFT orthogonal space is input to the first AI network model.
  • DFT Discrete Fourier Transform
  • the input to the first AI network model may also be a precoding matrix calculated by precoding the first channel information or a precoding matrix of a specific layer, or the precoding matrix may be projected The second coefficient in the complete orthogonal space, or the second coefficient, and part or all of the weighted orthogonal basis vector corresponding to the second coefficient.
  • the coefficients of the frequency domain-spatial domain orthogonal basis vector can be determined through the following coding preprocessing:
  • the PMI in the first column is a 32-length vector.
  • the 32-length The vector of is projected on the 32-point DFT orthogonal basis vector group.
  • the 32-point DFT requires 32 DFT orthogonal basis vectors to form a complete orthogonal space.
  • a 1 to a 32 are the coefficients corresponding to the 32 DFT orthogonal basis vectors DFT 1 to DFT 32 respectively.
  • One implementation method is that there are 32 corresponding coding AI network models, and the input of each network model is the corresponding two DFT vectors and the coefficients of the DFT vector, for example: [DFT 1 ⁇ a 1 ,DFT 2 ⁇ a 2 ] input into the first AI network model, input [DFT 2 ⁇ a 2 , DFT 3 ⁇ a 3 ] into the second AI network model, and so on, and finally input [DFT 32 ⁇ a 32 + DFT 1 ⁇ a 1 ] Enter the 32nd AI network model.
  • each network model is the corresponding two DFT vectors and the coefficients of the DFT vector, for example: [DFT 1 ⁇ a 1 ,DFT 2 ⁇ a 2 ] input the first AI network model, input [DFT 3 ⁇ a 3 ,DFT 4 ⁇ a 4 ] into the second AI network model, and so on, finally input [DFT 31 ⁇ a 31 +DFT 32 ⁇ a 32 ] Enter the 16th AI network model.
  • the above encoding preprocessing method may correspond to the first AI network model used, or may be agreed upon by the communication protocol, or may be instructed by the network side device, which is not specifically limited here.
  • the information input to the first AI network model may be all or part of the above-mentioned second coefficients, for example: directly input the above-mentioned a 1 to a 32 into the first AI network model, or input a 1 to a 32 into Several coefficients with larger amplitudes are input into the first AI network model.
  • the first AI network model when the first channel information is input to the first AI network model, can be used to form the corresponding one or at least two orthogonal basis vectors, such that , input the first channel information to the first AI network model, and the first AI network model can be used to determine the coefficients of the first channel information projected on the orthogonal basis vector formed by the first AI network model.
  • N first AI network models can correspond one-to-one with N target orthogonal basis vectors.
  • a certain first AI network model can constitute its corresponding target orthogonal basis vector.
  • the first AI network model can determine the result of projecting the first channel information onto its corresponding target orthogonal basis vector, thereby obtaining the coefficient corresponding to the target orthogonal basis vector.
  • the terminal inputs the first channel information to the N first AI network models respectively, and can obtain the coefficients of the first channel information projected onto the N target orthogonal basis vectors respectively, thereby realizing encoding of the channel information, that is, compressing the first channel information.
  • the network side device can restore the above-mentioned first channel information based on the N target orthogonal basis vectors and their corresponding coefficients, thereby enabling the terminal to report the first channel information while reducing the time required to report the first channel.
  • Information overhead is a coded overhead.
  • the first AI network model may also constitute at least two orthogonal basis vectors.
  • the one-to-one correspondence between the first AI network model and the orthogonal basis vectors is used as an example. In this Does not constitute a specific limitation.
  • the above-mentioned input of the result of projection in the complete orthogonal space to the first AI network model specifically refers to inputting the first coefficient of the first channel information projected into the complete orthogonal space into the first AI. network model, so that the first AI network model outputs coefficients of the target orthogonal basis vector.
  • the 32 channel coefficients in each row can be projected to 32 DFT orthogonal bases, with a total of 32 coefficients. In this way, the projection coefficients obtained after all 4 rows of channel coefficients are projected are still the same. is a 4 ⁇ 32 matrix.
  • the above-mentioned input of the result of projection in the complete orthogonal space to the first AI network model specifically refers to inputting part or all of the weighted first coefficient and the corresponding orthogonal basis vector into The first AI network model is so that the first AI network model outputs coefficients of the target orthogonal basis vector.
  • a 4 ⁇ 32 channel matrix 32 channel coefficients in each row are projected onto 32 DFT orthogonal basis, which can be expressed as 32 coefficients and 32 orthogonal basis vectors, and the length of each orthogonal basis vector is 32.
  • Weight the 32 orthogonal basis vectors and merge them into a vector of 32 ⁇ 32 1024 length, and then input the 1024-length vector into the first AI network model to calculate the corresponding coefficients, or convert a part of it (For example: a 64-length vector obtained by weighting the sum of the two strongest orthogonal basis vectors among the 32 orthogonal basis vectors) is input into the first AI network model to calculate the coefficient corresponding to this part.
  • the first coefficient includes:
  • At least part of the first channel information is projected in the coefficients of the complete orthogonal space; or,
  • the amplitude of the first channel information projected in the coefficients of the complete orthogonal space is greater than or equal to the preset amplitude; or,
  • Parts of the coefficients of the first channel information projected in the complete orthogonal space that are greater than or equal to the preset coefficients are sorted in order of amplitude value.
  • the first coefficient may be a part of the coefficients of the first channel information projected in the complete orthogonal space, for example: a part with a larger coefficient, or a part with a corresponding larger amplitude, or a part with a larger coefficient according to Coefficient sequence arranged by amplitude values. That is, the first coefficient may be a coefficient value or a set of coefficient values.
  • the first AI network model may also output a weighted orthogonal vector, so the terminal also needs to use a known vector (i.e., a preset vector) and this weighted orthogonal vector to Calculate the coefficients of the target orthogonal basis vectors.
  • a known vector i.e., a preset vector
  • the terminal inputs the first channel information to M first AI network models to obtain coefficients corresponding to the M target orthogonal basis vectors, including:
  • the terminal inputs the first channel information to M first AI network models to obtain M first orthogonal basis vectors, where the first orthogonal basis vectors are weighted orthogonal vectors;
  • the terminal determines M coefficients based on the preset vector and the M first orthogonal basis vectors.
  • the preset vector may be a preconfigured vector, or may be the result of inputting a default vector (for example, a unit orthogonal basis vector, an all-0 vector, or an all-1 vector, etc.) into the first AI network model.
  • a default vector for example, a unit orthogonal basis vector, an all-0 vector, or an all-1 vector, etc.
  • the channel characteristic information reporting method further includes:
  • the terminal sends identification information of the first AI network model corresponding to each of the N coefficients to the network side device.
  • the terminal uses the first AI network model to form the target orthogonal basis vector and determines the coefficients of the target orthogonal basis vector, it can report the identification information of the first AI network model to the network side device.
  • the network side device determines the target orthogonal basis vector corresponding to each of the N coefficients in the received first channel characteristic information based on the identification information of the first AI network model, so that based on the N coefficients and the corresponding corresponding N coefficients Target orthogonal basis vectors to restore the first channel information.
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the terminal when the terminal is configured with only the first AI network model that constitutes the spatial domain orthogonal basis vectors, the terminal can select several preset frequency domain orthogonal basis vectors (ie delays) according to conventional technical means in the existing technology, and then The above-mentioned first AI network model is used to determine the coefficients of the airspace orthogonal basis vectors for each delay, so that non-zero values or larger ones of each airspace coefficient of each delay can be reported to the network side device.
  • the terminal when the terminal is configured with only the first AI network model that constitutes the spatial domain orthogonal basis vectors, the terminal can select several preset frequency domain orthogonal basis vectors (ie delays) according to conventional technical means in the existing technology, and then The above-mentioned first AI network model is used to determine the coefficients of the airspace orthogonal basis vectors for each delay, so that non-zero values or larger ones of each airspace coefficient of each delay can be reported to the network side device.
  • the coefficients of the frequency domain-spatial domain orthogonal basis vectors can be determined by selecting a preset frequency domain orthogonal basis through coding preprocessing.
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the coefficients of the preset spatial domain orthogonal basis vectors are determined according to the target delay domain channel information, wherein the target delay domain channel information is obtained by processing the first channel information using the first AI network model.
  • the above-mentioned preset spatial domain orthogonal basis vectors may be selected by conventional technical means in the prior art to select several spatial domain orthogonal basis vectors (i.e., beams).
  • the preset spatial domain may be selected first. Orthogonal basis vectors, and then calculate the coefficients of the frequency domain orthogonal basis vectors of each preset spatial domain orthogonal basis vector. Alternatively, you can first calculate the delay domain information of the target channel, and then calculate the preset corresponding to the delay domain information. Coefficients of spatial orthogonal basis vectors.
  • the terminal when the terminal is configured with only the first AI network model that constitutes the frequency domain orthogonal basis vector, the terminal can select several preset air domain orthogonal basis vectors according to conventional technical means in the existing technology. vector (i.e., beam), and then for each polarization of each beam corresponding to the spatial domain orthogonal basis vector, the above-mentioned first AI network model is used to determine the coefficient of the frequency domain orthogonal basis vector, so that each beam can be reported to the network side device. Non-zero values or larger ones in each frequency domain coefficient of the preset spatial domain orthogonal basis vectors.
  • vector i.e., beam
  • the terminal is configured with only the first AI network module that constitutes the frequency domain orthogonal basis vector.
  • the terminal can input all channel information into the first AI network model to obtain the delay domain channel information of the target channel, then select several beams, and calculate the orthogonal airspace corresponding to each polarization of the selected beams. Coefficients of basis vectors.
  • each subband corresponds to a 4 ⁇ 32 channel matrix, where 4 is the number of receiving antennas and 32 is the number of CSI-RS ports.
  • the input dimensions of the first AI network model can be 13 The length corresponding to the complex number, and the output can be the length corresponding to N complex numbers.
  • the terminal inputs the coefficients of the 13 subbands into the first AI network model for each CSI-RS port of each receiving antenna, and obtains N coefficients through the first AI network model, which is the CSI-RS of the receiving antenna.
  • N delay domain channel matrices For the delay domain information corresponding to the port, after traversing all CSI-RS ports of all receiving antennas, N delay domain channel matrices can be obtained, and then several beams are selected according to the conventional scheme in the existing technology, and each beam is calculated. The coefficients of a delay domain channel matrix on each selected beam are reported to the network side device, and the coefficients of the preset spatial domain orthogonal basis vector corresponding to the delay domain channel matrix are reported to the network side device.
  • the spatial domain-frequency domain orthogonal basis vector or the frequency domain-spatial domain orthogonal basis vector can be determined by selecting a preset spatial domain orthogonal basis. coefficient.
  • the method before the terminal inputs the first information to the M first AI network models, the method further includes:
  • the terminal receives the M first AI network models from the network side device.
  • the terminal can first obtain multiple first AI network models from the network side device, and each first AI network model can constitute one or at least two orthogonal basis vectors. In this way, when the terminal reports channel information, The configured first AI network model can be used to calculate the coefficients of the corresponding orthogonal basis vectors.
  • the method further includes:
  • the terminal sends target capability information to the network side device, where the target capability information is used to indicate a maximum number of first AI network models supported by the terminal, where M is less than or equal to the first AI supported by the terminal.
  • the maximum number of network models is used to indicate a maximum number of first AI network models supported by the terminal, where M is less than or equal to the first AI supported by the terminal. The maximum number of network models.
  • the above target capability information may be the maximum number of first AI network models that the terminal can configure.
  • the network side device can determine how many first AI network models to configure for the terminal according to the capability information of the terminal, or which first AI network model to configure for the terminal. Some of the first AI network models.
  • the terminal obtains the first channel information of the target channel; the terminal obtains N coefficients according to the first channel information, wherein the N coefficients are determined by using N first AI network models respectively.
  • the N coefficients correspond to the N first AI network models one-to-one, or the N coefficients are the coefficients of the N target orthogonal basis vectors selected using the second AI network model, and N is An integer greater than or equal to 1;
  • the terminal sends first channel characteristic information to the network side device, and the first channel characteristic information includes the N coefficients.
  • the terminal can use the AI network model to select the target orthogonal basis vector whose coefficients need to be reported, or use the AI network model to determine the coefficients of the specified target orthogonal basis vector, so that the channel information can be reported based on the orthogonal basis vector.
  • Granularity the required AI network model is relatively small, thus reducing the cost of transmitting the AI network model between the terminal and the network side device,
  • the AI network model can even be used on the network side device to determine the orthogonal basis vectors of coefficients that need to be reported, and instruct the terminal to report these coefficients, which can avoid transmitting the AI network model between the terminal and the network side device.
  • an embodiment of the present application provides a channel characteristic information recovery method.
  • the execution subject may be a network side device.
  • the terminal may be various types of network side devices 12 listed in Figure 1, or other than Network-side devices other than the network-side device types listed in the embodiment shown in FIG. 1 are not specifically limited here.
  • the channel characteristic information recovery method may include the following steps:
  • Step 501 The network side device receives the first channel characteristic information from the terminal, where the first channel characteristic information includes N coefficients, and the N coefficients are the first channel information using N first AI network models respectively.
  • the coefficients obtained by the processing, or the N coefficients are the coefficients of the first channel information projected on the N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1.
  • first channel characteristic information, coefficients, first AI network model, first channel information, second AI network model and target orthogonal basis vector are respectively the same as the first channel in the method embodiment as shown in Figure 2
  • Feature information, coefficients, first AI network model, first channel information, second AI network model and target orthogonal basis vector have the same meaning and will not be described again here.
  • Step 502 The network side device performs recovery processing on the first channel characteristic information to obtain the first channel information.
  • the network side device can also obtain the target orthogonal basis vector, thereby based on the target orthogonal basis vector.
  • the first channel information is restored by intersecting basis vectors and respective corresponding coefficients.
  • the network side device has already learned the target orthogonal basis vector, and thus can also use the target orthogonal basis vector based on the target orthogonal basis vector. basis vectors and respective corresponding coefficients to restore the first channel information.
  • the terminal can also report the target orthogonal basis vector to the network side device, so that the network side device is based on Target orthogonal basis vectors and respective corresponding coefficients are used to restore the first channel information.
  • the terminal can also report the identification information of the N first AI network models to the network side device, so that the network side device adopts the same Decode the AI network model corresponding to the N first AI network models to restore the first channel information based on the N coefficients, or enable the network side device to use the target orthogonal basis corresponding to the N first AI network models. vector to determine the orthogonal basis vector corresponding to each of the N coefficients, thereby recovering the first channel information.
  • the channel characteristic information recovery method further includes:
  • the network side device receives identification information of the first AI network model corresponding to each of the N coefficients from the terminal.
  • the network side device can determine the corresponding decoding according to the identification information of the first AI network model. AI network model, and/or determine the corresponding target orthogonal basis vector.
  • the network side device performs recovery processing on the first channel characteristic information to obtain the first channel information, including:
  • the network side device determines the target orthogonal basis vectors corresponding to the N coefficients based on the identification information of the N first AI network models;
  • the network side device restores the first channel information according to the target orthogonal basis vector and the N coefficients.
  • the N first AI network models can correspond one-to-one to the N target orthogonal basis vectors.
  • the network side device can determine the N first AI network models based on the identification information reported by the terminal.
  • the coefficients respectively refer to the coefficients on which target orthogonal basis vector the first channel information is projected, thereby restoring the first channel information.
  • the first AI network model corresponds to the target orthogonal basis vector one-to-one
  • the network side device performs recovery processing on the first channel characteristic information, including:
  • the network side device uses a third AI network model to restore the first channel characteristic information, wherein the input length of the third AI network model is M, and M is equal to the maximum number of the target orthogonal basis vectors. , M is an integer greater than or equal to N.
  • the above M may be equal to the maximum number of the target orthogonal basis vectors, for example: complete orthogonal basis vectors.
  • the N coefficients reported by the terminal may only include the first channel projected on part of the target orthogonal basis. coefficients on the vector.
  • a third AI network model with a fixed input length of M is used to calculate N Coefficients are restored.
  • the network side device uses a third AI network model to restore the first channel characteristic information, including:
  • the network side device performs first processing on the first channel characteristic information, so that the length of the first channel characteristic information is adjusted from a length of N coefficients to a length of M coefficients;
  • the network side device uses a third AI network model to restore the first processed first channel characteristic information.
  • the above-mentioned first process may be a process of supplementing the default value.
  • the coefficients other than N coefficients among the M coefficients they are all equal to 0 by default. In this way, the length of the first channel characteristic information reported by the terminal can be made consistent with the input length of the third AI network model.
  • the network side device performs recovery processing on the first channel characteristic information to obtain the first channel information, including:
  • the network-side device uses N fourth AI network models to perform recovery processing on respective corresponding coefficients, and the N coefficients correspond to the N fourth AI network models one-to-one;
  • the network side device determines the first channel information based on the channel information recovered from each of the N fourth AI network models.
  • the above-mentioned fourth AI network model can correspond to the first AI network model one-to-one.
  • the above fourth AI network model and the corresponding first AI network model can be jointly trained by the network side device.
  • independent N fourth AI network models are used to recover a corresponding set of channel information, and then the channel information recovered by the N fourth AI network models are combined, for example: directly added, or using Another AI network model is used to combine the channel information recovered by each of the N fourth AI network models to obtain the first channel information.
  • the network side device performs recovery processing on the first channel characteristic information to obtain the first channel information, including:
  • the network side device restores the second channel information according to the preset orthogonal basis vector and the first channel characteristic information
  • the network side device uses the fifth AI network model to correct the second channel information to obtain the first channel information.
  • the network side device first uses a preset orthogonal basis vector (for example, an orthogonal vector of all 0s or all 1s, or a unit orthogonal basis vector) and the N coefficients reported by the terminal to combine the second channel information.
  • the fifth AI network model is then used to correct the second channel information to obtain the finally restored first channel information.
  • the network side device performs recovery processing on the first channel characteristic information to obtain the first channel information, including:
  • the network side device performs weighting processing on the N coefficients according to N second orthogonal basis vectors to obtain the first channel information, wherein the N second orthogonal basis vectors and the Nth An AI network model has a one-to-one correspondence, and each of the second orthogonal basis vectors is obtained by joint training with its corresponding first AI network model.
  • the network side device can use the second orthogonal basis vectors corresponding to the N first AI network models, and then use the N second orthogonal basis vectors and their corresponding coefficients to perform weighting processing, obtaining
  • the process of weighting the first channel information using the second orthogonal basis vector and coefficient is similar to the weighting process in the prior art, and will not be described again here.
  • the network side device performs recovery processing on the first channel characteristic information to obtain
  • the first channel information includes:
  • the network side device uses the second AI network model to determine the N target orthogonal basis vectors, or the network side device receives the target orthogonal basis vectors from the terminal;
  • the network side device restores the first channel information according to the N target orthogonal basis vectors and the coefficients corresponding to the N target orthogonal basis vectors.
  • the network side device when the terminal uses the second AI network model to select N target orthogonal basis vectors, the network side device also uses the same second AI network model to determine the N target orthogonal basis vectors, or from the terminal By receiving the N target orthogonal basis vectors, the network side device can determine the target orthogonal basis vectors corresponding to each of the N coefficients, and restore the first channel information accordingly.
  • the method further includes:
  • the network side device sends first indication information to the terminal, where the first indication information is used to indicate the N target orthogonal basis vectors.
  • the network side device uses the second AI network model to select N target orthogonal basis vectors, it directly sends the selected N target orthogonal basis vectors to the terminal, so that the terminal can directly calculate and report The coefficients of the orthogonal basis vectors indicated by the network side device are sufficient. In this way, there is no need to transmit the second AI network model between the network side device and the terminal.
  • the method before the network side device receives the first channel characteristic information from the terminal, the method further includes:
  • the network side device sends M first AI network models to the terminal, where the M first AI network models include the N first AI network models.
  • the network side device configures M first AI network models for the terminal, so that the terminal selects at least part of the M first AI network models that have been configured to determine the coefficients.
  • the method further includes:
  • the network side device receives target capability information from the terminal, where the target capability information is used to indicate a maximum number of first AI network models supported by the terminal, where M is less than or equal to the first AI network model supported by the terminal.
  • the maximum number of AI network models are used to indicate a maximum number of first AI network models supported by the terminal, where M is less than or equal to the first AI network model supported by the terminal. The maximum number of AI network models.
  • the network side device determines how many first AI network models to configure for the terminal according to the capabilities of the terminal, and/or which first AI network models to configure for the terminal, so that the first AI network model configured for the terminal Match its capabilities and reduce the problem of resource waste caused by configuring the first AI network model that does not match the capabilities of the terminal.
  • the network side device can recover the channel characteristic information reported by the terminal with the orthogonal basis vector coefficient as the granularity, thus realizing the reporting of the channel characteristic information with the orthogonal basis vector as the granularity.
  • the required encoding AI network model and/or decoding AI network model is relatively small, thus reducing the cost of transmitting the AI network model between the terminal and the network side device.
  • the AI network model can even be used in the network side device to determine the coefficients that need to be reported. Orthogonal basis vectors and instructing the terminal to report these coefficients can avoid transmitting the AI network model between the terminal and the network side device.
  • each subband has a 4 ⁇ 32 channel matrix, that is, the terminal has 4 receiving antennas, and each receiving antenna has 32 CSI-RS ports.
  • the terminal can input 13 4 ⁇ 32 channel matrices into the first AI network model 1 to obtain the coefficient 1, where the first AI network model 1 can constitute the target orthogonal basis vector 1, then the coefficient 1 can be Coefficients that project 13 4 ⁇ 32 channel matrices onto the target orthogonal basis vector 1. Then 13 4 ⁇ 32 channel matrices are input into the first AI network model 2 to obtain coefficient 2, and M first AI network models are traversed in sequence to obtain M coefficients.
  • the terminal can select 12 coefficients from 32 coefficients and quantize these 12 coefficients Then, report it to the network side device. At this time, the terminal can also report to the network side device to obtain the identification information of the first AI network model of these 12 coefficients.
  • the identification information of the 12 first AI network models are: 1 to 9 respectively. , 11, 13 and 14.
  • the network side device After receiving the above 12 coefficients and the identification information of the 12 first AI network models, the network side device selects the orthogonal basis vectors corresponding to the identifiers 1-9, 11, 13, and 14 from the 32 basic orthogonal basis vectors. and the corresponding reported coefficients to obtain complete channel information, and then the network side device can correct the complete channel information through another AI network model to obtain the final result; or,
  • the network side device can correspond the coefficients at positions 1-9, 11, 13, and 14 among the 32 coefficients as reported coefficients, and set the coefficients at other positions to zero, and then decode the AI network model (and/or orthogonal basis vectors Coding AI network model) to restore complete channel information based on these 32 coefficients; or,
  • the network side device can weight the received 12 coefficients to the corresponding basic orthogonal basis vectors (for example: DFT orthogonal basis vectors), and then input the 12 32-dimensional DFT orthogonal basis vectors into the decoding network to obtain the restored channel information.
  • the corresponding basic orthogonal basis vectors for example: DFT orthogonal basis vectors
  • the execution subject may be a channel characteristic information reporting device.
  • the method for reporting channel characteristic information performed by the channel characteristic information reporting device is used as an example to illustrate the channel characteristic information reporting device provided by the embodiment of the present application.
  • a device for reporting channel characteristic information provided by an embodiment of the present application may be a device within a terminal. As shown in Figure 6, the device 600 for reporting channel characteristic information may include the following modules:
  • the first acquisition module 601 is used to acquire the first channel information of the target channel
  • the second acquisition module 602 is configured to acquire N coefficients according to the first channel information, where the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients are consistent with the N There is a one-to-one correspondence between the first AI network models, or the N coefficients are the coefficients of the N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the first sending module 603 is configured to send first channel characteristic information to the network side device, where the first channel characteristic information includes the N coefficients.
  • the second acquisition module 602 is specifically used for:
  • the first channel information is respectively projected onto N target orthogonal basis vectors to obtain N coefficients, wherein the target orthogonal basis vector is determined by the second AI network model.
  • the target orthogonal basis vector includes at least one of the following:
  • the channel characteristic information reporting device 600 also includes:
  • the second receiving module is configured to receive first indication information from the network side device, where the first indication information is Instructing the target orthogonal basis vector, wherein the target orthogonal basis vector is trained by the network side device using the second AI network model; or,
  • a training module configured to use the second AI network model to train based on the first channel information to obtain the target orthogonal basis vector
  • the second acquisition module is used to acquire the target orthogonal basis vector agreed in the communication protocol.
  • the channel characteristic information reporting device 600 also includes:
  • the second sending module is configured to send the target orthogonal basis vector to the network side device.
  • the second acquisition module 602 includes:
  • a first processing unit configured to input first information to M first AI network models and obtain coefficients corresponding to M target orthogonal basis vectors, wherein the M first AI network models and the M The target orthogonal basis vectors correspond one to one, and M is an integer greater than or equal to N;
  • a first determination unit configured to determine N coefficients from the coefficients corresponding to each of the M target orthogonal basis vectors
  • the first information includes:
  • the first coefficient of the first channel information projected in the complete orthogonal space or,
  • the precoding matrix or the precoding matrix of a specific layer obtained by precoding calculation of the first channel information; or,
  • the second coefficient, and part or all of the weighted orthogonal basis vector corresponding to the second coefficient are the weighted orthogonal basis vector corresponding to the second coefficient.
  • the first coefficient includes:
  • At least part of the first channel information is projected in the coefficients of the complete orthogonal space; or,
  • the amplitude of the first channel information projected in the coefficients of the complete orthogonal space is greater than or equal to the preset amplitude; or,
  • Parts of the coefficients of the first channel information projected in the complete orthogonal space that are greater than or equal to the preset coefficients are sorted in order of amplitude value.
  • the channel characteristic information reporting device 600 also includes:
  • the third sending module is configured to send the identification information of the first AI network model corresponding to each of the N coefficients to the network side device.
  • the first processing unit includes:
  • a first processing subunit configured to input the first channel information to M first AI network models to obtain M first orthogonal basis vectors, where the first orthogonal basis vectors are weighted positive intersection vector;
  • the first determination subunit is used to determine M coefficients based on the preset vector and the M first orthogonal basis vectors.
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the coefficients of the preset spatial domain orthogonal basis vectors are determined according to the target delay domain channel information, wherein the target delay domain channel information is obtained by processing the first channel information using the first AI network model.
  • the channel characteristic information reporting device 600 also includes:
  • the third receiving module is configured to receive the M first AI network models from the network side device.
  • the channel characteristic information reporting device 600 also includes:
  • the fourth sending module is configured to send target capability information to the network side device, where the target capability information is used to indicate the maximum number of first AI network models supported by the terminal, where M is less than or equal to the number supported by the terminal. The maximum number of first AI network models.
  • the coefficient is a complex number including a real part and an imaginary part, or the coefficient is a real number including at least one amplitude and/or phase.
  • the channel characteristic information reporting device 600 in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or may be a component in the electronic device, such as an integrated circuit or chip.
  • the electronic device may be a terminal or other devices other than the terminal.
  • terminals may include but are not limited to the types of terminals 11 listed above, and other devices may be servers, network attached storage (Network Attached Storage, NAS), etc., which are not specifically limited in the embodiment of this application.
  • NAS Network Attached Storage
  • the channel characteristic information reporting device 600 provided by the embodiment of this application can implement each process implemented by the method embodiment shown in Figure 2 and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • the execution subject may be a channel characteristic information recovery device.
  • the channel characteristic information restoration method performed by the channel characteristic information restoration apparatus is used as an example to illustrate the channel characteristic information restoration apparatus provided by the embodiments of the present application.
  • a device for recovering channel characteristic information provided by an embodiment of the present application can be a device in a network-side device. As shown in Figure 7, the device for restoring channel characteristic information 700 can include the following modules:
  • the first receiving module 701 is configured to receive first channel characteristic information from the terminal, where the first channel characteristic information includes N coefficients, and the N coefficients are the first The coefficients obtained by processing the channel information, or the N coefficients are the coefficients of the first channel information projected on the N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1. ;
  • the first processing module 702 is used to restore the first channel characteristic information to obtain the first channel information.
  • the channel characteristic information recovery device 700 further includes:
  • the fourth receiving module is configured to receive identification information of the first AI network model corresponding to each of the N coefficients from the terminal.
  • the first processing module 702 includes:
  • a second determination unit configured to determine the target orthogonal basis vectors corresponding to the N coefficients according to the identification information of the N first AI network models
  • a first restoration unit configured to restore the first channel information according to the target orthogonal basis vector and the N coefficients.
  • the first AI network model has a one-to-one correspondence with the target orthogonal basis vector.
  • the first processing module 702 is specifically used for:
  • a third AI network model is used to restore the first channel characteristic information, wherein the input length of the third AI network model is M, M is equal to the maximum number of the target orthogonal basis vectors, and M is greater than or An integer equal to N.
  • the first processing module 702 includes:
  • a second processing unit configured to perform first processing on the first channel characteristic information, so that the length of the first channel characteristic information is adjusted from the length of N coefficients to the length of M coefficients;
  • the second restoration unit is configured to restore the first processed first channel characteristic information using a third AI network model.
  • the first processing module 702 includes:
  • a third recovery unit configured to use N fourth AI network models to perform recovery processing on respective corresponding coefficients, where the N coefficients correspond to the N fourth AI network models in one-to-one correspondence;
  • a third determination unit is configured to determine the first channel information based on the channel information recovered from each of the N fourth AI network models.
  • the first processing module 702 includes:
  • a fourth restoration unit configured to restore the second channel information according to the preset orthogonal basis vector and the first channel characteristic information
  • a correction unit configured to use a fifth AI network model to correct the second channel information to obtain the first channel information.
  • the first processing module 702 is specifically used to:
  • the N coefficients are weighted according to N second orthogonal basis vectors to obtain the first channel information, wherein the N second orthogonal basis vectors are the same as the N first AI network models.
  • N second orthogonal basis vectors are the same as the N first AI network models.
  • each of the mentioned The two orthogonal basis vectors are obtained by joint training with their corresponding first AI network models.
  • the first processing module 702 includes:
  • a fourth determination unit configured to determine the N target orthogonal basis vectors using the second AI network model, or the network side device receives the target orthogonal basis vectors from the terminal;
  • the fifth restoration unit is configured to restore the first channel information according to the N target orthogonal basis vectors and the corresponding coefficients of the N target orthogonal basis vectors.
  • the channel characteristic information recovery device 700 also includes:
  • the fifth sending module is configured to send first indication information to the terminal, where the first indication information is used to indicate the N target orthogonal basis vectors.
  • the channel characteristic information recovery device 700 also includes:
  • a sixth sending module configured to send M first AI network models to the terminal, where the M first AI network models include the N first AI network models.
  • the channel characteristic information recovery device 700 also includes:
  • the fifth receiving module is configured to receive target capability information from the terminal, where the target capability information is used to indicate the maximum number of first AI network models supported by the terminal, where M is less than or equal to the number of first AI network models supported by the terminal. The maximum number of first AI network models.
  • the channel characteristic information recovery device 700 in the embodiment of the present application may be an electronic device, such as an electronic device with an operating system, or may be a component in the electronic device, such as an integrated circuit or chip.
  • the electronic device may be a network-side device, or may be other devices besides the network-side device.
  • the terminal may include but is not limited to the types of network side devices 12 listed above.
  • Other devices may be servers, network attached storage (Network Attached Storage, NAS), etc., which are not specifically limited in the embodiment of this application.
  • the channel characteristic information recovery device 700 provided by the embodiment of the present application can implement each process implemented by the method embodiment shown in Figure 5 and achieve the same technical effect. To avoid duplication, the details will not be described here.
  • this embodiment of the present application also provides a communication device 800, which includes a processor 801 and a memory 802.
  • the memory 802 stores programs or instructions that can be run on the processor 801, such as , when the communication device 800 is a terminal, when the program or instruction is executed by the processor 801, each step of the above channel characteristic information reporting method embodiment is implemented, and the same technical effect can be achieved.
  • the communication device 800 is a network-side device, when the program or instruction is executed by the processor 801, the steps of the above channel characteristic information recovery method embodiment are implemented, and the same technical effect can be achieved. To avoid duplication, they will not be described again here.
  • An embodiment of the present application also provides a terminal, including a processor and a communication interface.
  • the communication interface is used to obtain first channel information of a target channel; the processor is used to obtain N coefficients according to the first channel information, wherein, The N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients correspond to the N first AI network models one-to-one, or the N coefficients are determined by using the second AI network.
  • the coefficients of the N target orthogonal basis vectors selected by the model, N is an integer greater than or equal to 1; the communication interface is also used to send the first message to the network side device.
  • Channel characteristic information, the first channel characteristic information includes the N coefficients.
  • FIG. 9 is a schematic diagram of the hardware structure of a terminal that implements an embodiment of the present application.
  • the terminal 900 includes but is not limited to: a radio frequency unit 901, a network module 902, an audio output unit 903, an input unit 904, a sensor 905, a display unit 906, a user input unit 907, an interface unit 908, a memory 909, a processor 910, etc. At least some parts.
  • the terminal 900 may also include a power supply (such as a battery) that supplies power to various components.
  • the power supply may be logically connected to the processor 910 through a power management system, thereby managing charging, discharging, and power consumption through the power management system. Management and other functions.
  • the terminal structure shown in FIG. 9 does not constitute a limitation on the terminal.
  • the terminal may include more or fewer components than shown in the figure, or may combine certain components, or arrange different components, which will not be described again here.
  • the input unit 904 may include a graphics processing unit (Graphics Processing Unit, GPU) 9041 and a microphone 9042.
  • the graphics processor 9041 is responsible for the image capture device (GPU) in the video capture mode or the image capture mode. Process the image data of still pictures or videos obtained by cameras (such as cameras).
  • the display unit 906 may include a display panel 9061, which may be configured in the form of a liquid crystal display, an organic light emitting diode, or the like.
  • the user input unit 907 includes a touch panel 9071 and at least one of other input devices 9072 .
  • Touch panel 9071 also known as touch screen.
  • the touch panel 9071 may include two parts: a touch detection device and a touch controller.
  • Other input devices 9072 may include but are not limited to physical keyboards, function keys (such as volume control keys, switch keys, etc.), trackballs, mice, and joysticks, which will not be described again here.
  • the radio frequency unit 901 after receiving downlink data from the network side device, can transmit it to the processor 910 for processing; in addition, the radio frequency unit 901 can send uplink data to the network side device.
  • the radio frequency unit 901 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low noise amplifier, a duplexer, etc.
  • Memory 909 may be used to store software programs or instructions as well as various data.
  • the memory 909 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, Image playback function, etc.) etc.
  • memory 909 may include volatile memory or nonvolatile memory, or memory 909 may include both volatile and nonvolatile memory.
  • 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), electrically removable memory. Erase programmable read-only memory (Electrically EPROM, EEPROM) or flash memory.
  • Volatile memory can be random access memory (Random Access Memory, RAM), static random access memory (Static RAM, SRAM), 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, DDRSDRAM), enhanced synchronous dynamic random access memory (Enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (Synch link DRAM) , SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DRRAM).
  • RAM Random Access Memory
  • SRAM static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM Double Data Rate SDRAM
  • DDRSDRAM double data rate synchronous dynamic random access memory
  • Enhanced SDRAM, ESDRAM enhanced synchronous dynamic random access memory
  • Synch link DRAM synchronous link dynamic random access memory
  • SLDRAM direct memory bus
  • the processor 910 may include one or more processing units; optionally, the processor 910 integrates an application processor and a modem processor, where the application processor mainly handles operations related to the operating system, user interface, application programs, etc., Modem processors mainly process wireless communication signals, such as baseband processors. It can be understood that the above modem processor may not be integrated into the processor 910.
  • the radio frequency unit 901 is used to obtain the first channel information of the target channel
  • the processor 910 is configured to obtain N coefficients according to the first channel information, wherein the N coefficients are coefficients determined by using N first AI network models respectively, and the N coefficients are consistent with the Nth One AI network model has a one-to-one correspondence, or the N coefficients are coefficients of N target orthogonal basis vectors selected using the second AI network model, and N is an integer greater than or equal to 1;
  • the radio frequency unit 901 is also configured to send first channel characteristic information to the network side device, where the first channel characteristic information includes the N coefficients.
  • the acquisition of N coefficients according to the first channel information performed by the processor 910 includes:
  • the first channel information is respectively projected onto N target orthogonal basis vectors to obtain N coefficients, wherein the target orthogonal basis vector is determined by the second AI network model.
  • the target orthogonal basis vector includes at least one of the following:
  • the radio frequency unit 901 is also configured to receive first indication information from the network side device, where the first indication information is used to indicate the target orthogonal basis vector, wherein the target orthogonal basis vector Obtained by training by the network side device using the second AI network model;
  • the processor 910 is also configured to use a second AI network model to train based on the first channel information to obtain the target orthogonal basis vector;
  • the processor 910 is also configured to obtain the target orthogonal basis vector agreed in the communication protocol.
  • the radio frequency unit 901 is also configured to provide the network side device with Send the target orthogonal basis vectors.
  • the processor 910 performs processing of the first channel information using N first AI network models to obtain N coefficients, including:
  • the first information includes:
  • the first coefficient of the first channel information projected in the complete orthogonal space or,
  • the precoding matrix or the precoding matrix of a specific layer obtained by precoding calculation of the first channel information; or,
  • the second coefficient, and part or all of the weighted orthogonal basis vector corresponding to the second coefficient are the weighted orthogonal basis vector corresponding to the second coefficient.
  • the first coefficient includes:
  • At least part of the first channel information is projected in the coefficients of the complete orthogonal space; or,
  • the amplitude of the first channel information projected in the coefficients of the complete orthogonal space is greater than or equal to the preset amplitude; or,
  • Parts of the coefficients of the first channel information projected in the complete orthogonal space that are greater than or equal to the preset coefficients are sorted in order of amplitude value.
  • the radio frequency unit 901 is also configured to send identification information of the first AI network model corresponding to each of the N coefficients to the network side device.
  • the processor 910 performs the input of the first information to the M first AI network models to obtain coefficients corresponding to the M target orthogonal basis vectors, including:
  • M coefficients are determined according to the preset vector and the M first orthogonal basis vectors.
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the coefficients corresponding to each of the M target orthogonal basis vectors include:
  • the coefficients of the preset spatial domain orthogonal basis vectors are determined according to the target delay domain channel information, wherein the target delay domain channel information is obtained by processing the first channel information using the first AI network model.
  • the radio frequency Unit 901 is also configured to receive the M first AI network models from the network side device.
  • the radio frequency unit 901 before performing the receiving of the M first AI network models from the network side device, the radio frequency unit 901 is also configured to:
  • Target capability information is used to indicate a maximum number of first AI network models supported by the terminal, where M is less than or equal to the number of first AI network models supported by the terminal. greatest amount.
  • the coefficient is a complex number including a real part and an imaginary part, or the coefficient is a real number including at least one amplitude and/or phase.
  • the terminal 900 provided by the embodiment of the present application can perform each process performed by each module in the channel characteristic information reporting device 600 as shown in Figure 6, and can achieve the same beneficial effects. To avoid duplication, details will not be described here.
  • An embodiment of the present application also provides a network side device, including a processor and a communication interface.
  • the communication interface is used to receive first channel characteristic information from a terminal, wherein the first channel characteristic information includes N coefficients, and the N
  • the N coefficients are coefficients obtained by respectively using N first AI network models to process the first channel information, or the N coefficients are the projections of the first channel information on N targets selected by using the second AI network model.
  • the coefficient of the orthogonal basis vector, N is an integer greater than or equal to 1; the processor is used to restore the first channel characteristic information to obtain the first channel information.
  • This network-side device embodiment corresponds to the above-mentioned network-side device method embodiment.
  • Each implementation process and implementation manner of the above-mentioned method embodiment can be applied to this network-side device embodiment, and can achieve the same technical effect.
  • the embodiment of the present application also provides a network side device.
  • the network side device 1000 includes: an antenna 1001, a radio frequency device 1002, a baseband device 1003, a processor 1004 and a memory 1005.
  • Antenna 1001 is connected to radio frequency device 1002.
  • the radio frequency device 1002 receives information through the antenna 1001 and sends the received information to the baseband device 1003 for processing.
  • the baseband device 1003 processes the information to be sent and sends it to the radio frequency device 1002.
  • the radio frequency device 1002 processes the received information and sends it out through the antenna 1001.
  • the method performed by the network side device in the above embodiment can be implemented in the baseband device 1003, which includes a baseband processor.
  • the baseband device 1003 may include, for example, at least one baseband board on which multiple chips are disposed, as shown in FIG. Program to perform the network device operations shown in the above method embodiments.
  • the network side device may also include a network interface 1006, which is, for example, a Common Public Radio Interface (CPRI).
  • CPRI Common Public Radio Interface
  • the network side device 1000 in the embodiment of the present application also includes: instructions or programs stored in the memory 1005 and executable on the processor 1004.
  • the processor 1004 calls the instructions or programs in the memory 1005 to execute each of the steps shown in Figure 7
  • the method of module execution and achieving the same technical effect will not be described in detail here to avoid duplication.
  • Embodiments of the present application also provide a readable storage medium, where programs or instructions are stored on the readable storage medium.
  • programs or instructions are stored on the readable storage medium.
  • the processor is the processor in the terminal described in the above embodiment.
  • the readable storage medium includes computer readable storage media, such as computer read-only memory ROM, random access memory RAM, magnetic disk or optical disk, etc.
  • An embodiment of the present application further provides a chip.
  • the chip includes a processor and a communication interface.
  • the communication interface is coupled to the processor.
  • the processor is used to run programs or instructions.
  • the implementation is as shown in Figure 2 or Figure 5. Each process of the method embodiment is shown, and the same technical effect can be achieved. To avoid repetition, the details will not be described here.
  • chips mentioned in the embodiments of this application may also be called system-on-chip, system-on-a-chip, system-on-chip or system-on-chip, etc.
  • Embodiments of the present application further provide a computer program/program product, the computer program/program product is stored in a storage medium, and the computer program/program product is executed by at least one processor to implement Figure 2 or Figure 5
  • the computer program/program product is executed by at least one processor to implement Figure 2 or Figure 5
  • An embodiment of the present application also provides a communication system, including: a terminal and a network side device.
  • the terminal can be used to perform the steps of the channel characteristic information reporting method described in the first aspect
  • the network side device can be used to perform the steps of the channel characteristic information reporting method as described in the first aspect.
  • the methods of the above embodiments can be implemented by means of software plus the necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is better. implementation.
  • the technical solution of the present application can be embodied in the form of a computer software product that is essentially or contributes to the existing technology.
  • the computer software product is stored in a storage medium (such as ROM/RAM, disk , CD), including several instructions to cause a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of this application.

Landscapes

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

Abstract

本申请公开了一种信道特征信息上报及恢复方法、终端和网络侧设备，属于通信技术领域，本申请实施例的信道特征信息上报方法包括：终端获取目标信道的第一信道信息；所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。

Description

信道特征信息上报及恢复方法、终端和网络侧设备
相关申请的交叉引用
本申请主张在2022年04月01日在中国提交的中国专利申请No.202210349421.X的优先权，其全部内容通过引用包含于此。
技术领域
本申请属于通信技术领域，具体涉及一种信道特征信息上报及恢复方法、终端和网络侧设备。
背景技术
准确的信道状态信息(Channel State Information，CSI)对信道容量的至关重要，随着人工智能(Artificial Intelligence，AI)在通信领域的应用，可以使用AI网络模型对信道状态信息(Channel State Information，CSI)信息进行编码和解码。
在相关技术中，为了减少CSI反馈的开销，基站可以事先对CSI参考信号(CSI Reference Signal，CSI-RS)进行预编码，将编码后的CSI-RS发送给终端，终端看到的是经过编码之后的CSI-RS对应的信道矩阵，终端需要将整个信道矩阵的信道信息输入至一个很大的AI网络模型中，以使该AI网络模型输出需要上报的全部正交基向量的系数。也就是说，CSI信息的编解码是针对整个信道而言的，需要的AI网络模型比较大，此外，针对不同长度的CSI，需要重新训练AI网络模型，甚至使用不同的AI网络模型，因此需要较大的传输开销提前配置所有的AI网络模型。
发明内容
本申请实施例提供一种信道特征信息上报及恢复方法、终端和网络侧设备，能够采用AI网络模型来选择需要上报系数的目标正交基向量，或者，采用AI网络模型来确定指定的目标正交基向量的系数，使得信道信息的上报能够以正交基向量为粒度，所需的AI网络模型比较小，从而减少了终端与网络侧设备之间传输AI网络模型的开销，甚至还可以在网络侧设备采用AI网络模型来确定需要上报系数的正交基向量，并指示终端上报这些系数，能够避免在终端与网络侧设备之间传输AI网络模型。
第一方面，提供了一种信道特征信息上报方法，该方法包括：
终端获取目标信道的第一信道信息；
所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应， 或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
第二方面，提供了一种信道特征信息上报装置，应用于终端，该装置包括：
第一获取模块，用于获取目标信道的第一信道信息；
第二获取模块，用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
第一发送模块，用于向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
第三方面，提供了一种信道特征信息恢复方法，包括：
网络侧设备接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
第四方面，提供了一种信道特征信息恢复装置，应用于网络侧设备，该装置包括：
第一接收模块，用于接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
第一处理模块，用于对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
第五方面，提供了一种终端，该终端包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第一方面所述的方法的步骤。
第六方面，提供了一种终端，包括处理器及通信接口，其中，所述通信接口用于获取目标信道的第一信道信息；所述处理器用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述通信接口还用于向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
第七方面，提供了一种网络侧设备，该网络侧设备包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如第 三方面所述的方法的步骤。
第八方面，提供了一种网络侧设备，包括处理器及通信接口，其中，所述通信接口用于接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述处理器用于对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
第九方面，提供了一种通信系统，包括：终端及网络侧设备，所述终端可用于执行如第一方面所述的信道特征信息上报方法的步骤，所述网络侧设备可用于执行如第三方面所述的信道特征信息恢复方法的步骤。
第十方面，提供了一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如第一方面所述的方法的步骤，或者实现如第三方面所述的方法的步骤。
第十一方面，提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如第一方面所述的方法，或实现如第三方面所述的方法。
第十二方面，提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现如第一方面所述的信道特征信息上报方法的步骤，或者所述计算机程序/程序产品被至少一个处理器执行以实现如第三方面所述的信道特征信息恢复方法的步骤。
在本申请实施例中，终端获取目标信道的第一信道信息；所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。这样，终端能够采用AI网络模型来选择需要上报系数的目标正交基向量，或者，采用AI网络模型来确定指定的目标正交基向量的系数，使得信道信息的上报能够以正交基向量为粒度，所需的AI网络模型比较小，从而减少了终端与网络侧设备之间传输AI网络模型的开销，甚至还可以在网络侧设备采用AI网络模型来确定需要上报系数的正交基向量，并指示终端上报这些系数，能够避免在终端与网络侧设备之间传输AI网络模型。
附图说明
图1是本申请实施例能够应用的一种无线通信系统的结构示意图；
图2是本申请实施例提供的一种信道特征信息上报方法的流程图；
图3是神经网络模型的架构示意图；
图4是神经元的示意图；
图5是本申请实施例提供的一种信道特征信息恢复方法的流程图；
图6是本申请实施例提供的一种信道特征信息上报装置的结构示意图；
图7是本申请实施例提供的一种信道特征信息恢复装置的结构示意图；
图8是本申请实施例提供的一种通信设备的结构示意图；
图9是本申请实施例提供的一种终端的结构示意图；
图10是本申请实施例提供的一种网络侧设备的结构示意图。
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清楚描述，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。
本申请的说明书和权利要求书中的术语“第一”、“第二”等是用于区别类似的对象，而不用于描述特定的顺序或先后次序。应该理解这样使用的术语在适当情况下可以互换，以便本申请的实施例能够以除了在这里图示或描述的那些以外的顺序实施，且“第一”、“第二”所区别的对象通常为一类，并不限定对象的个数，例如第一对象可以是一个，也可以是多个。此外，说明书以及权利要求中“和/或”表示所连接对象的至少其中之一，字符“/”一般表示前后关联对象是一种“或”的关系。
值得指出的是，本申请实施例所描述的技术不限于长期演进型(Long Term Evolution，LTE)/LTE的演进(LTE-Advanced，LTE-A)系统，还可用于其他无线通信系统，诸如码分多址(Code Division Multiple Access，CDMA)、时分多址(Time Division Multiple Access，TDMA)、频分多址(Frequency Division Multiple Access，FDMA)、正交频分多址(Orthogonal Frequency Division Multiple Access，OFDMA)、单载波频分多址(Single-carrier Frequency Division Multiple Access，SC-FDMA)和其他系统。本申请实施例中的术语“系统”和“网络”常被可互换地使用，所描述的技术既可用于以上提及的系统和无线电技术，也可用于其他系统和无线电技术。以下描述出于示例目的描述了新空口(New Radio，NR)系统，并且在以下大部分描述中使用NR术语，但是这些技术也可应用于NR系统应用以外的应用，如第6代(6th Generation，6G)通信系统。
图1示出本申请实施例可应用的一种无线通信系统的框图。无线通信系统包括终端11和网络侧设备12。其中，终端11可以是手机、平板电脑(Tablet Personal Computer)、膝上型电脑(Laptop Computer)或称为笔记本电脑、个人数字助理(Personal Digital Assistant，PDA)、掌上电脑、上网本、超级移动个人计算机(ultra-mobile personal computer，UMPC)、移动上网装置(Mobile Internet Device，MID)、增强现实(augmented reality，AR)/虚拟现实(virtual reality，VR)设备、机器人、可穿戴式设备(Wearable Device)、车载设备(Vehicle User Equipment，VUE)、行人终端(Pedestrian User Equipment，PUE)、智能家居(具有 无线通信功能的家居设备，如冰箱、电视、洗衣机或者家具等)、游戏机、个人计算机(personal computer，PC)、柜员机或者自助机等终端侧设备，可穿戴式设备包括：智能手表、智能手环、智能耳机、智能眼镜、智能首饰(智能手镯、智能手链、智能戒指、智能项链、智能脚镯、智能脚链等)、智能腕带、智能服装等。需要说明的是，在本申请实施例并不限定终端11的具体类型。网络侧设备12可以包括接入网设备或核心网设备，其中，接入网设备也可以称为无线接入网设备、无线接入网(Radio Access Network,RAN)、无线接入网功能或无线接入网单元。接入网设备可以包括基站、无线局域网(Wireless Local Area Networks，WLAN)接入点或WiFi节点等，基站可被称为节点B、演进节点B(eNB)、接入点、基收发机站(Base Transceiver Station，BTS)、无线电基站、无线电收发机、基本服务集(Basic Service Set，BSS)、扩展服务集(Extended Service Set，ESS)、家用B节点、家用演进型B节点、发送接收点(Transmitting Receiving Point，TRP)或所述领域中其他某个合适的术语，只要达到相同的技术效果，所述基站不限于特定技术词汇，需要说明的是，在本申请实施例中仅以NR系统中的基站为例进行介绍，并不限定基站的具体类型。
在无线通信技术中，准确的CSI反馈对信道容量至关重要。尤其是对于多天线系统来讲，发送端可以根据CSI优化信号的发送，使其更加匹配信道的状态。如：信道质量指示(Channel Quality Indicator，CQI)可以用来选择合适的调制编码方案(Modulation and Coding Scheme，MCS)，以实现链路自适应；预编码矩阵指示(Precoding Matrix Indicator，PMI)可以用来实现特征波束成形(eigen beamforming)，从而最大化接收信号的强度，或者用来抑制干扰(如小区间干扰、多用户之间干扰等)。因此，自从多天线技术(如：多输入多输出(Multi-Input Multi-Output，MIMO))被提出以来，CSI的获取一直都是研究热点。
通常，网络侧设备在某个时隙(slot)的某些时频资源上发送CSI参考信号(CSI-Reference Signals，CSI-RS)，终端根据CSI-RS进行信道估计，计算这个slot上的信道信息，通过码本将PMI反馈给基站，网络侧设备根据终端反馈的码本信息组合出信道信息，并在终端下一次上报CSI之前，网络侧设备以此信道信息进行数据预编码及多用户调度。
为了进一步减少CSI反馈开销，终端可以将每个子带上报PMI改成按照时延(delay域，即频域)上报PMI，由于delay域的信道更集中，用更少的delay的PMI就可以近似表示全部子带的PMI，其可以视作是将delay域信息压缩之后再上报。
同样，为了减少开销，网络侧设备可以事先对CSI-RS进行预编码，将编码后的CSI-RS发送给终端，终端看到的是经过编码之后的CSI-RS对应的信道，终端只需要在网络侧设备指示的端口中选择若干个强度较大的端口，并上报这些端口对应的系数即可。
在相关技术中，利用AI网络模型对信道信息进行压缩，能够提升信道特征信息的压缩效果，其中，AI网络模型有多种实现方式，例如：神经网络、决策树、支持向量机、 贝叶斯分类器等。为了便于说明，本申请实施例中以AI网络模型为神经网络为例进行说明，但是并不限定AI网络模型的具体类型。具体的，终端可以估计CSI参考信号(CSI Reference Signal，CSI-RS)或跟踪参考信号(Tracking Reference Signal，TRS)，根据该估计到的信道信息进行计算，得到计算的信道信息，然后，将计算的信道信息或者原始的估计到的信道信息通过编码网络模型进行编码，得到编码结果，最后将编码结果发送给基站。在基站侧，基站可以在接收编码后的结果后，将其输入到解码网络模型中，利用该解码网络模型恢复信道信息。
例如：在R16码本结构下，网络侧发送预编码后的CSI-RS，终端接收信道矩阵，并从中选择2L个空域正交基向量，Mv个时延(delay)域正交基向量，然后上报选择的正交基向量以及对应的系数，网络侧设备则可以根据正交基向量和对应的系数恢复信道信息，其中，delay域对应频域。
但是，相关技术中，采用AI网络模型对信道信息进行编码和解码的过程，都是针对整个信道而言的，因此，AI网络模型的数据量会很大，在传递AI网络模型时，会产生较大的开销，此外，不同长度(overhead)的信道信息，对应不同的AI网络模型，网络侧设备需要针对每一种长度的信道信息分别训练AI网络模型，甚至分别配置不同的AI网络模型，这就增加了终端与网络侧设备之间由于训练和配置AI网络模型所造成的计算量、占用资源和时延等。
本申请实施例中，终端可以根据采用N个第一AI网络模型确定N个系数，或者采用第二AI网络模型选择的N个目标正交基向量，并确定N个目标正交基向量各自的系数，其以正交基向量的系数为粒度，使得第一AI网络模型或第二AI网络模型的数据量小，从而能够减少网络侧设备与终端之间因传输该第一AI网络模型或第二AI网络模型所造成的资源消耗。此外，以正交基向量的系数为粒度，还可以在满足信道信息上报的基础上，实现减少上报的系数的数量的目的，达到减少信道信息上报开销的有益效果。
值得注意的是，本申请实施例中的信道信息的编码不同于相关技术中的信道编码，本申请实施例中的信道信息的编码过程可以包括以下步骤：
步骤1、终端在网络指定的时频域位置检测CSI-RS或TRS，并进行信道估计，得到第一信道信息；
步骤2、终端通过第一AI网络模型(即编码AI网络模型)分别将K组第一信道信息编码为第一信道特征信息；
步骤3、终端将第一信道特征信息的部分或全部内容以及其他控制信息组合为上行控制信息(Uplink Control Information，UCI)，或者将第一信道特征信息的部分或全部内容作为UCI；
步骤4、终端根据UCI的长度对UCI进行分割，并添加循环冗余校验(Cyclic redundancy check，CRC)比特；
步骤5、终端对添加CRC比特的UCI进行信道编码；
步骤6、终端对UCI进行速率匹配；
步骤7、终端对UCI进行码块关联；
步骤8、终端将UCI映射到物理上行控制信道(Physical Uplink Control Channel，PUCCH)或物理上行共享信道(Physical Uplink Shared Channel，PUSCH)进行上报。
需要说明的是，上述信道信息的编码流程中，部分步骤的顺序可以调整或者省略，在此不构成具体限定。
下面结合附图，通过一些实施例及其应用场景对本申请实施例提供的信道特征信息上报方法、信道特征信息恢复方法、信道特征信息上报装置、信道特征信息恢复装置及通信设备等进行详细地说明。
请参阅图2，本申请实施例提供的一种信道特征信息上报方法，其执行主体可以是终端，该终端可以是如图1中列举的各种类型的终端11，或者是除了如图1所示实施例中列举的终端类型之外的其他终端，在此不作具体限定。如图2所示，该信道特征信息上报方法可以包括以下步骤：
步骤201、终端获取目标信道的第一信道信息。
在实施中，终端可以通过对CSR-RS或TRS等参考信号进行信道估计，得到上述第一信道信息，或者，上述第一信道信息是终端对估计到的原始信道信息进行一定的计算或预处理后得到的信道信息，例如：第一信道信息可能是根据信道估计得到的信道信息确定的预编码矩阵或特定层的预编码矩阵，在此不作具体限定。
步骤202、所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数。
在实施中，上述目标正交基向量可以包括空域正交基向量、频域正交基向量、多普勒域正交基向量中的至少一项，且上述目标正交基向量的系数，其可以是第一信道信息投影在对应的目标正交基向量上的投影，该系数可以是一个数值，也可能是一组数值，例如：系数可以是包括实部和虚部的复数，或者所述系数可以是包括至少一个幅度和/或相位的实数。
在一种可能的实现方式中，所述N个系数为分别采用N个第一AI网络模型确定的系数，可以是N个第一AI网络模型与N个正交基向量一一对应。
此时，所述终端根据所述第一信道信息获取N个系数，可以是：所述终端分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数。
需要说明的是，上述第一AI网络模型可以表征对应的正交基向量，例如：表征一个空域正交基向量、表征一个频域正交基向量、表征一个多普勒域正交基向量或者表征至少两个正交基向量联合的正交基向量，此时，可以采用第一AI网络模型来表示对应的正交基向量的信息，而无需采用具体的正交基向量来计算对应的系数。在实施中，N个第一 AI网络模型可以表示为编码AI网络模型，该编码AI网络模型与其对应的解码AI网络模型可以是联合训练的，且对应的正交基向量的信息能够体现在编码AI网络模型和解码AI网络模型中，这样，能够采用编码AI网络模型来确定对应的正交基向量的系数，在网络侧则可以采用解码AI网络模型对对应的编码AI网络模型得出的系数进行恢复。
这样，可以使用N个第一AI网络模型来构成N个正交基向量，在采用第一AI网络模型对所述第一信道信息进行处理时，可以采用第一AI网络模型来计算将第一信道信息投影在该第一AI网络模型对应的正交基向量上的系数。
在实施中，上述N个第一AI网络模型可以是网络侧设备配置给终端的，鉴于第一AI网络模型只需要计算单个正交基向量的系数，其尺寸非常小，因此，在网络侧设备将N个第一AI网络模型配置给终端时，所造成的资源开销也很小。
值得提出的是，使用第一AI网络模型来构成正交基向量，能够使用更少的正交基向量来表征目标信道实际的正交基向量，其相较于在固定的DFT向量上投影而言，通过采用信道信息训练得到的第一AI网络模型，能够减少终端上报系数的正交基向量的数量，从而减少信道特征信息上报的开销。
在另一种可能的实现方式中，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，这样，在采用第二AI网络模型确定N个目标正交基向量的情况下，上述终端根据所述第一信道信息获取N个系数，可以包括：
所述终端将所述第一信道信息分别投影在N个目标正交基向量上，得到N个系数。
需要说明的是，上述第二AI网络模型可以是所述网络侧设备使用的AI网络模型，即不同的网络侧设备可以配置各自的第二AI网络模型，在终端接入网络侧设备时，该网络侧设备向终端配置自身的第二AI网络模型，以使终端也采用相同的第二AI网络模型确定N个目标正交基向量，并上报这些目标正交基向量的系数，或者网络侧设备可以直接向终端指示采用第二AI网络模型所确定的目标正交基向量，以使终端不需要采用第二AI网络模型来确定目标正交基向量，就可以直接上报这些目标正交基向量的系数。
当然，上述第二AI网络模型也可能是通信协议中约定的离线的第二AI网络模型，或者是终端训练得到的AI网络模型。
需要说明的是，本申请实施例中，终端采用第二AI网络模型确定目标正交基向量，或者按照网络侧设备采用第二AI网络模型选择的目标正交基向量，能够选择投影较大的部分正交基向量，这样，终端只上报这些正交基向量的系数即可。
可选地，所述信道特征信息上报方法还包括：
所述终端接收来自所述网络侧设备的第一指示信息，所述第一指示信息用于指示所述目标正交基向量，其中，所述目标正交基向量由所述网络侧设备采用所述第二AI网络模型训练得到；
或者，
所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向 量；
或者，
所述终端根据通信协议中约定的所述目标正交基向量。
方式一，对于终端接收来自所述网络侧设备的第一指示信息，所述第一指示信息用于指示所述目标正交基向量的情况，所述第二AI网络模型可以是网络侧设备配置的，在实施中，不同的网络侧设备可以各自配置不同的第二AI网络模型，并基于各自配置的第二AI网络模型确定接入网络侧设备的终端需要上报哪些正交基向量的系数，从而通过第一指示信息指示终端上报这些正交基向量的系数。
在实施中，终端可以在接入网络侧设备时，从网络侧设备获取上述第一指示信息，并在接入该网络侧设备的期间，都按照第一指示信息的指示来上报指定的目标正交基向量的系数。与之相对应的，网络侧设备在采用第二AI网络模型训练得到N个目标正交基向量的情况下，在接收终端上报的N个系数时，能够确定这N个系数各自对应哪一个目标正交基向量，从而基于N个系数与各自对应的正交基向量来恢复第一信道信息。
方式二，对于所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量的情况，上述第二AI网络模型可以是协议约定的或者是网络侧设备指示的，例如：网络侧设备指示第二AI网络模型的结构，终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量，此时，不同的终端可以采用各自的第二AI网络模型来训练各自的目标正交基向量，并上报训练出的目标正交基向量的系数。
可选地，在所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量之后，所述方法还包括：
所述终端向所述网络侧设备发送所述目标正交基向量。
本实施方式中，终端还可以将训练出的目标正交基向量上报给网络侧设备，例如：周期性地向网络侧设备上报目标正交基向量，此外，目标正交基向量的变化频率较低，终端可以间隔较长的周期来上报该目标正交基向量。这样，网络侧设备能够获知终端上报的系数是哪些正交基向量的系数，从而基于N个系数与各自对应的正交基向量来恢复第一信道信息。
方式三，对于所述终端根据通信协议中约定的所述目标正交基向量的情况，上述目标正交基向量可以是采用离线的第二AI网络模型训练得到的正交基向量，在实施中，本方式三与上述方式一和方式二的区别包括：本方式三中，目标正交基向量是预先采用离线的第二AI网络模型训练得到的正交基向量，不会跟随网络状态的变化而变化，因此，不需要实时地训练，且通过在通信协议中约定该目标正交基向量的方式，使得终端和网络侧设备都能够获知该目标正交基向量，因此，终端和网络侧设备之间不需要对该目标正交基向量进行交互或者指示。
步骤203、所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
需要说明的是，在实施中，上述终端向网络侧设备发送第一信道特征信息，可以是采用CSI上报的方式在CSI报告中携带该第一信道特征信息，以上报网络侧设备，其中，信道特征信息具体可以是PMI信息。当然，上述第一信道特征信息还可以采用其他任意方式上报给网络侧设备，为了便于说明，本申请实施例中，以采用CSI上报的方式上报第一信道特征信息为例进行举例说明，在此不构成具体限定。
在N个系数为终端分别采用N个第一AI网络模型确定的系数的情况下，上述N个第一AI网络模型可以是网络侧设备预先为终端配置的AI网络模型，且该AI网络模型可以包括多种类型的AI算法模块，例如：神经网络、决策树、支持向量机、贝叶斯分类器等，在此不作具体限定，且为了便于说明，以下实施例中以所述AI算法模型为神经网络模型为例进行举例说明，在此不构成具体限定。
如图3所示，神经网络模型包括输入层、隐层和输出层，其可以根据输入层获取的出入信息(X1～Xn)预测可能的输出结果(Y)。神经网络模型由大量的神经元组成，如图4所示，神经元的参数包括：输入参数a1～aK、权值w、偏置b以及激活函数σ(z)，以及与这些参数获取输出值a，其中，常见的激活函数包括S型生长曲线(Sigmoid)函数、双曲正切(tanh)函数、线性整流函数(Rectified Linear Unit，ReLU，其也称之为修正线性单元)函数等等，且上述函数σ(z)中的z可以通过以下公式计算得到：
z＝a1w1+…+akwk+aKwK+b
其中，K表示输入参数的总数。
神经网络的参数通过优化算法进行优化。优化算法就是一种能够帮我们最小化或者最大化目标函数(有时候也叫损失函数)的一类算法。而目标函数往往是模型参数和数据的数学组合。例如给定数据X和其对应的标签Y，我们构建一个神经网络模型f(.)，有了模神经网络型后，根据输入x就可以得到预测输出f(x)，并且可以计算出预测值和真实值之间的差距(f(x)-Y)，这个就是损失函数。我们的目的是找到合适的W和b，使上述的损失函数的值达到最小，损失值越小，则说明我们的模型越接近于真实情况。
目前常见的优化算法，基本都是基于误差反向传播算法。误差反向传播算法的基本思想是，学习过程由信号的正向传播与误差的反向传播两个过程组成。正向传播时，输入样本从输入层传入，经各隐层逐层处理后，传向输出层。若输出层的实际输出与期望的输出不符，则转入误差的反向传播阶段。误差反传是将输出误差以某种形式通过隐层向输入层逐层反传，并将误差分摊给各层的所有单元，从而获得各层单元的误差信号，此误差信号即作为修正各单元权值的依据。这种信号正向传播与误差反向传播的各层权值调整过程，是周而复始地进行的。权值不断调整的过程，也就是网络的学习训练过程。此过程一直进行到网络输出的误差减少到可接受的程度，或进行到预先设定的学习次数为止。
常见的优化算法有梯度下降(Gradient Descent)、随机梯度下降(Stochastic Gradient Descent，SGD)、小批量梯度下降(mini-batch gradient descent)、动量法(Momentum)、Nesterov(其表示带动量的随机梯度下降)、自适应梯度下降(Adaptive gradient descent， Adagrad)、自适应学习率调整(Adadelta)、均方根误差降速(root mean square prop，RMSprop)、自适应动量估计(Adaptive Moment Estimation，Adam)等。
这些优化算法在误差反向传播时，都是根据损失函数得到的误差/损失，对当前神经元求导数/偏导，加上学习速率、之前的梯度/导数/偏导等影响，得到梯度，将梯度传给上一层。
作为一种可选的实施方式，所述终端采用第一AI网络模型将所述第一信道信息处理成N个系数，包括：
所述终端将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，其中，所述M个第一AI网络模型与所述M个目标正交基向量一一对应，M为大于或等于N的整数；
所述终端从所述M个目标正交基向量各自对应的系数中确定N个系数；
其中，所述第一信息包括：
所述第一信道信息；或者，
所述第一信道信息投影在完备正交空间的第一系数；或者，
所述第一系数，以及所述第一系数对应的正交基向量加权后的部分或全部；或者，
所述第一信道信息经过预编码计算，得到的预编码矩阵或特定层的预编码矩阵；或者，
所述预编码矩阵投影在完备正交空间的第二系数；或者，
所述第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部。
在实施中，上述M个第一AI网络模型可以是网络侧设备配置给终端的全部第一AI网络模型。根据信道状态的实际情况，终端可能只要上报其中部分第一AI网络模型输出的系数，便能够反映目标信道的实际正交基向量的情况，此时，上述终端从所述M个目标正交基向量各自对应的系数中确定N个系数，可以是，终端从M个目标正交基向量的系数中选择幅度较大的N个系数。
在一种可能的实施方式中，输入第一AI网络模型的可以是第一信道信息，或者是第一信道信息投影在完备正交空间的结果(第一系数)，或者是该第一系数及其对应的正交基向量加权后的部分或全部，例如：离散傅里叶变换(Discrete Fourier Transform，DFT)正交空间可以过采或不过采，并将投影在该DFT正交空间的结果输入到第一AI网络模型。
在另一种可能的实施方式中，输入第一AI网络模型的还可以是第一信道信息经过预编码计算得到的预编码矩阵或特定层的预编码矩阵，或者是将所述预编码矩阵投影在完备正交空间的第二系数，或者是第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部。
例如：以空域为例，假设第一信道信息是4×32的信道矩阵，则可以通过以下编码预处理，来确定频域-空域正交基向量的系数：
经过奇异值分解(singular value decomposition，SVD)分解之后选择选择V矩阵的第一列上报，即rank＝1，第一列的PMI是一个32长度的向量，通过映射，可以将32长度 的向量投影在32点DFT正交基向量组上，32点DFT构成完备正交空间需要32个DFT正交基向量，投影完便可以得到32个系数，即PMI＝DFT1×a1+DFT2×a2+…DFT32×a32。其中，a1～a32分别是DFT1～DFT32这32个DFT正交基向量各自对应的系数。
以输入第一AI网络模型的信息包括上述第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部为例：
一种实现方式是，对应的编码AI网络模型为32个，且每个网络模型的输入为对应的两个DFT向量及该DFT向量的系数，例如：将[DFT1×a1,DFT2×a2]输入第一个AI网络模型，将[DFT2×a2,DFT3×a3]输入第二个AI网络模型，依次类推，最后将[DFT32×a32+DFT1×a1]输入第32个AI网络模型。
另一种实现方式是，对应的编码AI网络模型为16个，且每个网络模型的输入为对应的两个DFT向量及该DFT向量的系数，例如：将[DFT1×a1,DFT2×a2]输入第一个AI网络模型，将[DFT3×a3,DFT4×a4]输入第二个AI网络模型，依次类推，最后将[DFT31×a31+DFT32×a32]输入第16个AI网络模型。
在实施中，上述编码预处理的方式可以与使用的第一AI网络模型对应，或者由通信协议约定，或者由网络侧设备指示，在此不作具体限定。
需要说明的是，输入第一AI网络模型的信息可能是上述第二系数中的全部或者部分，例如：直接把上述a1～a32输入第一AI网络模型，或者把a1～a32中幅度较大的若干个系数输入第一AI网络模型。
在第一种可能的实现方式中，在上述将第一信道信息输入至第一AI网络模型的情况下，第一AI网络模型可以用于构成对应的一个或至少两个正交基向量，这样，将第一信道信息输入至第一AI网络模型，可以利用第一AI网络模型确定第一信道信息投影在该第一AI网络模型构成的正交基向量的系数。
本实施方式中，N个第一AI网络模型可以与N个目标正交基向量一一对应，这样，某一第一AI网络模型可以构成其对应的目标正交基向量，这样，在将第一信道信息输入该第一AI网络模型时，该第一AI网络模型能够确定第一信道信息投影在其对应的目标正交基向量上的结果，从而得到与该目标正交基向量对应的系数。终端分别向N个第一AI网络模型输入第一信道信息，便可以得到第一信道信息分别投影在N个目标正交基向量上的系数，从而实现信道信息的编码，即将第一信道信息压缩编码成数据量更小的第一信道特征信息。与之相对应的，网络侧设备能够根据N个目标正交基向量以及各自对应的系数来恢复上述第一信道信息，从而能够在实现终端上报第一信道信息的基础上，降低上报第一信道信息的开销。
需要说明的是，在实际应用中，也可以存在采用完整正交基向量对应的一个AI网络模型输出M个系数，并从中选择幅度较大的N个系数的方式来确定需要上报的目标正交基向量的系数。也就是说，第一AI网络模型也可能构成至少两个正交基向量，本申请实施例中，为了便于说明，以第一AI网络模型与正交基向量一一对应为例进行说明，在此 不构成具体限定。
在第二种可能的实现方式中，上述将投影在完备正交空间的结果输入到第一AI网络模型，具体指将第一信道信息投影在完备正交空间的第一系数输入到第一AI网络模型，以使第一AI网络模型输出目标正交基向量的系数。
例如：对于一个4×32的信道矩阵，其中每一行的32个信道系数都可以投影到32个DFT正交基，共32个系数，这样，4行信道系数全部进行投影之后得到的投影系数依旧是4×32的矩阵。
在第三种可能的实现方式中，上述将投影在完备正交空间的结果输入到第一AI网络模型，具体指将该第一系数和对应的正交基向量加权后的部分或全部输入到第一AI网络模型，以使第一AI网络模型输出目标正交基向量的系数。
例如：4×32的信道矩阵，每一行的32个信道系数投影到32个DFT正交基上，可以表示为32个系数和32个正交基向量，且每个正交基向量的长度为32，对32个正交基向量进行加权之后合并为一个32×32＝1024长度的向量，然后将该1024长度的向量输入到第一AI网络模型中计算对应的系数，或者是将其中的一部分(例如：对32个正交基向量中最强的两个正交基向量进行加权求和后，得到的64长度的向量)输入到第一AI网络模型中计算该部分对应的系数。
可选地，所述第一系数包括：
所述第一信道信息投影在完备正交空间的系数中的至少部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的幅度大于或者等于预设幅度的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分按照幅度值大小的排序。
在实施中，第一系数可以是第一信道信息投影在完备正交空间的系数中的一部分，例如：系数较大的一部分，或对应的幅度较大的一部分，或者是系数较大的一部分按照幅度值大小排列后的系数序列。即第一系数可以是一个系数值，也可能是一组系数值。
值得提出的是，在实施中，第一AI网络模型输出的也可能是加权后的正交向量，这样终端还需要使用一个已知向量(即预设向量)和这个加权后的正交向量来计算目标正交基向量的系数。
可选地，所述终端将所述第一信道信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，包括：
所述终端将所述第一信道信息输入至M个第一AI网络模型，得到M个第一正交基向量，其中，所述第一正交基向量为加权后的正交向量；
所述终端根据预设向量和所述M个第一正交基向量，确定M个系数。
其中，预设向量可以是预先配置的向量，也可以是将默认的向量(例如：单位正交基向量、全0向量或全1向量等)输入第一AI网络模型后得到的结果。
可选地，所述信道特征信息上报方法还包括：
所述终端向所述网络侧设备发送所述N个系数各自对应的第一AI网络模型的标识信息。
本实施方式中，终端在采用第一AI网络模型来构成目标正交基向量，并确定该目标正交基向量的系数之后，通过向网络侧设备上报该第一AI网络模型的标识信息，可以是网络侧设备根据该第一AI网络模型的标识信息确定接收到的第一信道特征信息中的N个系数各自对应的目标正交基向量，从而基于该N个系数和N个系数各自对应的目标正交基向量来恢复第一信道信息。
作为一种可选的实施方式，在所述第一AI网络模型构成空域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设频域正交基向量对应的空域正交基向量的系数或非零系数。
其中，在终端仅配置了构成空域正交基向量的第一AI网络模型时，终端可以按照现有技术中的常规技术手段来选择若干个预设频域正交基向量(即delay)，然后针对每一个delay分别采用上述第一AI网络模型来确定空域正交基向量的系数，从而可以向网络侧设备上报每个delay的每个空域系数中的非零值或较大的若干个。
本实施方式中，在仅存在空域正交基向量的AI网络模型时，能够通过编码预处理选择预设频域正交基的方式来确定频域-空域正交基向量的系数。
作为一种可选的实施方式，在所述第一AI网络模型构成频域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设空域正交基向量的每一个极化对应的频域正交基向量的系数或非零系数；和/或，
根据目标时延域信道信息确定的预设空域正交基向量的系数，其中，所述目标时延域信道信息是采用所述第一AI网络模型对所述第一信道信息进行处理得到。
在实施中，上述预设空域正交基向量可以是采用现有技术中的常规技术手段来选择若干个空域正交基向量(即波束(beam))，在实施中，可以先选择预设空域正交基向量，然后计算每一个预设空域正交基向量的频域正交基向量的系数，或者，可以先计算目标信道的时延域信息，然后计算该时延域信息对应的预设空域正交基向量的系数。
在一种可能的实现方式中，在终端仅配置了构成频域正交基向量的第一AI网络模型时，终端可以按照现有技术中的常规技术手段来选择若干个预设空域正交基向量(即beam)，然后针对每一个beam的每一个极化对应空域正交基向量，分别采用上述第一AI网络模型来确定频域正交基向量的系数，从而可以向网络侧设备上报每个预设空域正交基向量的每个频域系数中的非零值或较大的若干个。
在另一种可能的实现方式中，在终端仅配置了构成频域正交基向量的第一AI网络模 型时，终端可以将全部信道信息输入至第一AI网络模型中，得到目标信道的时延域信道信息，然后再选择若干个beam，以及计算选择的beam的每一个极化对应的空域正交基向量的系数。
例如：对于13个子带，每个子带对应1个4×32的信道矩阵，其中，4为接收天线的数量，32为CSI-RS端口的数量，第一AI网络模型的输入维度可以是13个复数对应的长度，且输出可以是N个复数对应的长度。这样，终端分别针对每个接收天线的每个CSI-RS端口，将13个子带的系数输入第一AI网络模型，通过第一AI网络模型得到N个系数，即为该接收天线的CSI-RS端口对应的时延域信息，遍历全部接收天线的全部CSI-RS端口后，便可以得到N个时延域信道矩阵，然后再按照现有技术中的常规方案来选择若干个beam，并计算每个时延域信道矩阵在选择的每一个beam上的系数，并向网络侧设备上报该时延域信道矩阵对应的预设空域正交基向量的系数。
本实施方式中，在仅存在频域正交基向量的AI网络模型时，能够通过选择预设空域正交基的方式来确定空域-频域正交基向量或频域-空域正交基向量的系数。
作为一种可选的实施方式，在所述终端将第一信息输入至M个第一AI网络模型之前，所述方法还包括：
所述终端接收来自所述网络侧设备的所述M个第一AI网络模型。
本实施方式中，终端可以先从网络侧设备获取多个第一AI网络模型，且每一个第一AI网络模型可以构成一个或至少两个正交基向量，这样，终端在上报信道信息时，能够采用已经配置的第一AI网络模型来计算对应的正交基向量的系数。
可选地，在所述终端接收来自所述网络侧设备的所述M个第一AI网络模型之前，所述方法还包括：
所述终端向所述网络侧设备发送目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
在实施中，上述目标能力信息可以是终端能够配置的第一AI网络模型的最大数量，这样，网络侧设备可以根据终端的能力信息来确定为终端配置多少个第一AI网络模型，或者配置哪一些第一AI网络模型。
在本申请实施例中，终端获取目标信道的第一信道信息；所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。这样，终端能够采用AI网络模型来选择需要上报系数的目标正交基向量，或者，采用AI网络模型来确定指定的目标正交基向量的系数，使得信道信息的上报能够以正交基向量为粒度，所需的AI网络模型比较小，从而减少了终端与网络侧设备之间传输AI网络模型的开销， 甚至还可以在网络侧设备采用AI网络模型来确定需要上报系数的正交基向量，并指示终端上报这些系数，能够避免在终端与网络侧设备之间传输AI网络模型。
请参阅图5，本申请实施例提供的一种信道特征信息恢复方法，其执行主体可以是网络侧设备，该终端可以是如图1中列举的各种类型的网络侧设备12，或者是除了如图1所示实施例中列举的网络侧设备类型之外的其他网络侧设备，在此不作具体限定。如图5所示，该信道特征信息恢复方法可以包括以下步骤：
步骤501、网络侧设备接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数。
在实施中，上述第一信道特征信息、系数、第一AI网络模型、第一信道信息、第二AI网络模型和目标正交基向量分别与如图2所示方法实施例中的第一信道特征信息、系数、第一AI网络模型、第一信道信息、第二AI网络模型和目标正交基向量具有相同的含义，在此不再赘述。
步骤502、所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
在实施中，若终端上报的N个西施是通信协议中约定的第二网络模型离线确定的目标正交基向量的系数，则网络侧设备也可以获取该目标正交基向量，从而基于目标正交基向量和各自对应的系数来恢复所述第一信道信息。
若终端上报的N个系数是网络侧设备采用自身的第二AI网络模型训练的目标正交基向量的系数，则网络侧设备已经获知了该目标正交基向量，从而也能够基于目标正交基向量和各自对应的系数来恢复所述第一信道信息。
若终端上报的N个系数是采用终端自身的第二AI网络模型训练的目标正交基向量的系数，则终端还可以向网络侧设备上报所述目标正交基向量，以使网络侧设备基于目标正交基向量和各自对应的系数来恢复所述第一信道信息。
若终端上报的N个系数是终端采用N个第一AI网络模型确定的系数，则终端还可以向网络侧设备上报所述N个第一AI网络模型的标识信息，以使网络侧设备采用与该N个第一AI网络模型对应的解码AI网络模型来根据所述N个系数恢复所述第一信道信息，或者，使网络侧设备基于该N个第一AI网络模型对应的目标正交基向量，来确定N个系数各自对应的正交基向量，从而恢复所述第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型确定的系数的情况下，所述信道特征信息恢复方法还包括：
所述网络侧设备接收来自所述终端的所述N个系数各自对应的第一AI网络模型的标识信息。
本实施方式中，网络侧设备能够根据第一AI网络模型的标识信息来确定对应的解码 AI网络模型，和/或确定对应的目标正交基向量。
可选地，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
所述网络侧设备根据所述N个第一AI网络模型的标识信息确定所述N个系数对应的目标正交基向量；
所述网络侧设备根据所述目标正交基向量和所述N个系数，恢复所述第一信道信息。
本实施方式中，N个第一AI网络模型可以与N个目标正交基向量一一对应，这样，网络侧设备能够根据终端上报的N个第一AI网络模型的标识信息确定所述N个系数分别是第一信道信息投影在哪一个目标正交基向量上的系数，从而恢复第一信道信息。
可选地，所述第一AI网络模型与目标正交基向量一一对应，所述网络侧设备对所述第一信道特征信息进行恢复处理，包括：
所述网络侧设备采用第三AI网络模型对所述第一信道特征信息进行恢复处理，其中，所述第三AI网络模型的输入长度为M，M等于所述目标正交基向量的最大数量，M为大于或等于N的整数。
本实施方式中，上述M可以等于所述目标正交基向量的最大数量，例如：完备正交基向量，此时，终端上报的N个系数可能只包括第一信道投影在部分目标正交基向量上的系数。本实施方式中，不论终端上报的系数是全部目标正交基向量的投影系数还是部分目标正交基向量的投影系数，都是采用固定的输入长度为M的第三AI网络模型来对N个系数进行恢复处理。
可选地，在M大于N的情况下，所述网络侧设备采用第三AI网络模型对所述第一信道特征信息进行恢复处理，包括：
所述网络侧设备对所述第一信道特征信息进行第一处理，以使所述第一信道特征信息的长度由N个系数的长度调整为M个系数的长度；
所述网络侧设备采用第三AI网络模型对所述第一处理后的第一信道特征信息进行恢复处理。
在实施中，上述第一处理可以是补充默认值的处理，例如：对于M个系数中除了N个系数以外的其他系数，都默认等于0。这样，可以使终端上报的第一信道特征信息的长度符合第三AI网络模型的输入长度。
可选地，在所述N个系数为分别采用N个第一AI网络模型确定的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
所述网络侧设备采用N个第四AI网络模型对各自对应的系数进行恢复处理，所述N个系数与所述N个第四AI网络模型一一对应；
所述网络侧设备根据所述N个第四AI网络模型各自恢复得到的信道信息，确定所述第一信道信息。
在实施中，上述第四AI网络模型可以与第一AI网络模型一一对应，在实际应用中， 可以由网络侧设备联合训练上述第四AI网络模型和对应的第一AI网络模型。
本实施方式中，采用独立的N个第四AI网络模型来恢复各自对应的一组信道信息，然后将N个第四AI网络模型各自恢复的信道信息进行合并，例如：直接相加，或者采用另一个AI网络模型来合并N个第四AI网络模型各自恢复的信道信息，以得到所述第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型确定的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
所述网络侧设备根据预设正交基向量和所述第一信道特征信息，恢复第二信道信息；
所述网络侧设备采用第五AI网络模型对所第二信道信息进行矫正，得到所述第一信道信息。
本实施方式中，网络侧设备先采用预设正交基向量(例如：全0或全1的正交向量，或者单位正交基向量)和终端上报的N个系数组合出第二信道信息，然后采用第五AI网络模型来矫正该第二信道信息，得到最终恢复的第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型确定的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
所述网络侧设备根据N个第二正交基向量对所述N个系数进行加权处理，得到所述第一信道信息，其中，所述N个第二正交基向量与所述N个第一AI网络模型一一对应，且每一个所述第二正交基向量与各自对应的第一AI网络模型联合训练得到。
本实施方式中，网络侧设备能够使用与N个第一AI网络模型一一对应的第二正交基向量，然后采用该N个第二正交基向量和各自对应的系数进行加权处理，得到第一信道信息，其中，采用第二正交基向量和系数进行加权处理的过程，与现有技术中的加权处理过程相似，在此不再赘述。
可选地，在所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
所述网络侧设备采用所述第二AI网络模型确定所述N个目标正交基向量，或者，所述网络侧设备接收来自所述终端的所述目标正交基向量；
所述网络侧设备根据所述N个目标正交基向量和所述N个目标正交基向量各自对应的系数，恢复所述第一信道信息。
本实施方式中，在终端采用第二AI网络模型选择N个目标正交基向量的情况下，网络侧设备也采用相同的第二AI网络模型来确定N个目标正交基向量，或者从终端接收该N个目标正交基向量，这样，网络侧设备能够确定N个系数各自对应的目标正交基向量，并据此恢复第一信道信息。
进一步地，在所述网络侧设备采用所述第二AI网络模型确定所述N个目标正交基向量之后，所述方法还包括：
所述网络侧设备向所述终端发送第一指示信息，所述第一指示信息用于指示所述N个目标正交基向量。
本实施方式中，网络侧设备在采用第二AI网络模型选择N个目标正交基向量的情况下，直接将选择出的N个目标正交基向量发送给终端，以使终端直接计算和上报网络侧设备指示的正交基向量的系数即可，这样，无需在网络侧设备和终端之间传输第二AI网络模型。
可选地，在所述网络侧设备接收来自终端的第一信道特征信息之前，所述方法还包括：
所述网络侧设备向所述终端发送M个第一AI网络模型，所述M个第一AI网络模型包括所述N个第一AI网络模型。
本实施方式中，网络侧设备给终端配置M个第一AI网络模型，以使终端从已经配置的M个第一AI网络模型中选择至少部分来确定系数。
进一步地，在所述网络侧设备向所述终端发送M个第一AI网络模型之前，所述方法还包括：
所述网络侧设备接收来自所述终端的目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
本实施方式中，网络侧设备按照终端的能力来确定给终端配置多少个第一AI网络模型，和/或，给终端配置哪一些第一AI网络模型，以使终端配置的第一AI网络模型与其能力匹配，降低配置与终端的能力不匹配的第一AI网络模型而造成的资源浪费的问题。
本申请实施例中，网络侧设备能够对终端上报的以正交基向量系数为粒度的信道特征信息进行恢复，也就实现了以正交基向量为粒度的信道特征信息的上报，该过程中所需的编码AI网络模型和/或解码AI网络模型比较小，从而减少了终端与网络侧设备之间传输AI网络模型的开销，甚至还可以在网络侧设备采用AI网络模型来确定需要上报系数的正交基向量，并指示终端上报这些系数，能够避免在终端与网络侧设备之间传输AI网络模型。
为了便于说明本申请实施例提供的信道特征信息上报方法和信道特征信息恢复方法，以如下应用场景为例，对本申请实施例提供的信道特征信息上报方法和信道特征信息恢复方法进行结合说明：
假设终端获取13个子带的第一信道信息，每一个子带具有一个4×32的信道矩阵，即终端有4个接收天线，且每一个接收天线有32个CSI-RS端口。
此时，终端可以将13个4×32的信道矩阵输入到第一AI网络模型1中，得到系数1，其中，第一AI网络模型1可以构成目标正交基向量1，则系数1可以是将13个4×32的信道矩阵投影到目标正交基向量1上的系数。然后将13个4×32的信道矩阵输入到第一AI网络模型2中，得到系数2，并依次遍历M个第一AI网络模型，得到M个系数。
假设M等于32，终端可以从32个系数中选择12个系数，并对这12个系数进行量化 后，上报给网络侧设备，此时，终端还可向网络侧设备上报得到这12个系数的第一AI网络模型的标识信息，例如：12第一AI网络模型的标识分别是：1～9，11，13和14。
网络侧设备则在接收到上述12个系数和12第一AI网络模型的标识信息之后，根据32个基础正交基向量中选择与标识1-9,11,13,14对应的正交基向量和对应的上报的系数，得到完整的信道信息，然后网络侧设备还可以通过另一AI网络模型对该完整的信道信息进行校正，得到最终结果；或者，
网络侧设备可以将32个系数中与1-9,11,13,14位置的系数对应为上报的系数，其他位置的系数则置零，然后通过解码AI网络模型(和/或正交基向量的编码AI网络模型)基于这32个系数恢复完整信道信息；或者，
网络侧设备可以将收到的12个系数加权到对应的基础正交基向量(例如：DFT正交基向量)上，然后将12个32维度的DFT正交基向量输入解码网络，得到恢复的信道信息。
本申请实施例提供的信道特征信息上报方法，执行主体可以为信道特征信息上报装置。本申请实施例中以信道特征信息上报装置执行信道特征信息上报方法为例，说明本申请实施例提供的信道特征信息上报装置。
请参阅图6，本申请实施例提供的一种信道特征信息上报装置，可以是终端内的装置，如图6所示，该信道特征信息上报装置600可以包括以下模块：
第一获取模块601，用于获取目标信道的第一信道信息；
第二获取模块602，用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
第一发送模块603，用于向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
可选地，第二获取模块602，具体用于：
分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数；或者，
将所述第一信道信息分别投影在N个目标正交基向量上，得到N个系数，其中，所述目标正交基向量由所述第二AI网络模型确定。
可选地，所述目标正交基向量包括以下至少一项：
空域正交基向量；
频域正交基向量；
多普勒域正交基向量。
可选地，信道特征信息上报装置600还包括：
第二接收模块，用于接收来自所述网络侧设备的第一指示信息，所述第一指示信息用 于指示所述目标正交基向量，其中，所述目标正交基向量由所述网络侧设备采用所述第二AI网络模型训练得到；或者，
训练模块，用于采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量；
或者，
第二获取模块，用于获取通信协议中约定所述目标正交基向量。
可选地，信道特征信息上报装置600还包括：
第二发送模块，用于向所述网络侧设备发送所述目标正交基向量。
可选地，第二获取模块602，包括：
第一处理单元，用于将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，其中，所述M个第一AI网络模型与所述M个目标正交基向量一一对应，M为大于或等于N的整数；
第一确定单元，用于从所述M个目标正交基向量各自对应的系数中确定N个系数；
其中，所述第一信息包括：
所述第一信道信息；或者，
所述第一信道信息投影在完备正交空间的第一系数；或者，
所述第一系数，以及所述第一系数对应的正交基向量加权后的部分或全部；或者，
所述第一信道信息经过预编码计算，得到的预编码矩阵或特定层的预编码矩阵；或者，
所述预编码矩阵投影在完备正交空间的第二系数；或者，
所述第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部。
可选地，所述第一系数包括：
所述第一信道信息投影在完备正交空间的系数中的至少部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的幅度大于或者等于预设幅度的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分按照幅度值大小的排序。
可选地，信道特征信息上报装置600还包括：
第三发送模块，用于向所述网络侧设备发送所述N个系数各自对应的第一AI网络模型的标识信息。
可选地，所述第一处理单元，包括：
第一处理子单元，用于将所述第一信道信息输入至M个第一AI网络模型，得到M个第一正交基向量，其中，所述第一正交基向量为加权后的正交向量；
第一确定子单元，用于根据预设向量和所述M个第一正交基向量，确定M个系数。
可选地，在所述第一AI网络模型构成空域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设频域正交基对应的空域正交基向量的系数或非零系数。
可选地，在所述第一AI网络模型构成频域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设空域正交基向量的每一个极化对应的频域正交基向量的系数或非零系数；和/或，
根据目标时延域信道信息确定的预设空域正交基向量的系数，其中，所述目标时延域信道信息是采用所述第一AI网络模型对所述第一信道信息进行处理得到。
可选地，信道特征信息上报装置600还包括：
第三接收模块，用于接收来自所述网络侧设备的所述M个第一AI网络模型。
可选地，信道特征信息上报装置600还包括：
第四发送模块，用于向所述网络侧设备发送目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
可选地，所述系数为包括实部和虚部的复数，或者所述系数为包括至少一个幅度和/或相位的实数。
本申请实施例中的信道特征信息上报装置600可以是电子设备，例如具有操作系统的电子设备，也可以是电子设备中的部件，例如集成电路或芯片。该电子设备可以是终端，也可以为除终端之外的其他设备。示例性的，终端可以包括但不限于上述所列举的终端11的类型，其他设备可以为服务器、网络附属存储器(Network Attached Storage，NAS)等，本申请实施例不作具体限定。
本申请实施例提供的信道特征信息上报装置600能够实现图2所示方法实施例实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例提供的信道特征信息恢复方法，执行主体可以为信道特征信息恢复装置。本申请实施例中以信道特征信息恢复装置执行信道特征信息恢复方法为例，说明本申请实施例提供的信道特征信息恢复装置。
请参阅图7，本申请实施例提供的一种信道特征信息恢复装置，可以是网络侧设备内的装置，如图7所示，该信道特征信息恢复装置700可以包括以下模块：
第一接收模块701，用于接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
第一处理模块702，用于对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，信道特征信息恢复装置700还包括：
第四接收模块，用于接收来自所述终端的所述N个系数各自对应的第一AI网络模型的标识信息。
可选地，第一处理模块702，包括：
第二确定单元，用于根据所述N个第一AI网络模型的标识信息确定所述N个系数对应的目标正交基向量；
第一恢复单元，用于根据所述目标正交基向量和所述N个系数，恢复所述第一信道信息。
可选地，所述第一AI网络模型与目标正交基向量一一对应，第一处理模块702，具体用于：
采用第三AI网络模型对所述第一信道特征信息进行恢复处理，其中，所述第三AI网络模型的输入长度为M，M等于所述目标正交基向量的最大数量，M为大于或等于N的整数。
可选地，在M大于N的情况下，第一处理模块702，包括：
第二处理单元，用于对所述第一信道特征信息进行第一处理，以使所述第一信道特征信息的长度由N个系数的长度调整为M个系数的长度；
第二恢复单元，用于采用第三AI网络模型对所述第一处理后的第一信道特征信息进行恢复处理。
可选地，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，第一处理模块702，包括：
第三恢复单元，用于采用N个第四AI网络模型对各自对应的系数进行恢复处理，所述N个系数与所述N个第四AI网络模型一一对应；
第三确定单元，用于根据所述N个第四AI网络模型各自恢复得到的信道信息，确定所述第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，第一处理模块702，包括：
第四恢复单元，用于根据预设正交基向量和所述第一信道特征信息，恢复第二信道信息；
矫正单元，用于采用第五AI网络模型对所第二信道信息进行矫正，得到所述第一信道信息。
可选地，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，第一处理模块702，具体用于：
根据N个第二正交基向量对所述N个系数进行加权处理，得到所述第一信道信息，其中，所述N个第二正交基向量与所述N个第一AI网络模型一一对应，且每一个所述第 二正交基向量与各自对应的第一AI网络模型联合训练得到。
可选地，在所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数的情况下，第一处理模块702，包括：
第四确定单元，用于采用所述第二AI网络模型确定所述N个目标正交基向量，或者，所述网络侧设备接收来自所述终端的所述目标正交基向量；
第五恢复单元，用于根据所述N个目标正交基向量和所述N个目标正交基向量各自对应的系数，恢复所述第一信道信息。
可选地，信道特征信息恢复装置700还包括：
第五发送模块，用于向所述终端发送第一指示信息，所述第一指示信息用于指示所述N个目标正交基向量。
可选地，信道特征信息恢复装置700还包括：
第六发送模块，用于向所述终端发送M个第一AI网络模型，所述M个第一AI网络模型包括所述N个第一AI网络模型。
可选地，信道特征信息恢复装置700还包括：
第五接收模块，用于接收来自所述终端的目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
本申请实施例中的信道特征信息恢复装置700可以是电子设备，例如具有操作系统的电子设备，也可以是电子设备中的部件，例如集成电路或芯片。该电子设备可以是网络侧设备，也可以为除网络侧设备之外的其他设备。示例性的，终端可以包括但不限于上述所列举的网络侧设备12的类型，其他设备可以为服务器、网络附属存储器(Network Attached Storage，NAS)等，本申请实施例不作具体限定。
本申请实施例提供的信道特征信息恢复装置700能够实现图5所示方法实施例实现的各个过程，并达到相同的技术效果，为避免重复，这里不再赘述。
可选地，如图8所示，本申请实施例还提供一种通信设备800，包括处理器801和存储器802，存储器802上存储有可在所述处理器801上运行的程序或指令，例如，该通信设备800为终端时，该程序或指令被处理器801执行时实现上述信道特征信息上报方法实施例的各个步骤，且能达到相同的技术效果。该通信设备800为网络侧设备时，该程序或指令被处理器801执行时实现上述信道特征信息恢复方法实施例的各个步骤，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供一种终端，包括处理器和通信接口，通信接口用于获取目标信道的第一信道信息；所述处理器用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述通信接口还用于向网络侧设备发送第一信 道特征信息，所述第一信道特征信息包括所述N个系数。
该终端实施例与上述终端侧方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该终端实施例中，且能达到相同的技术效果。具体地，图9为实现本申请实施例的一种终端的硬件结构示意图。
该终端900包括但不限于：射频单元901、网络模块902、音频输出单元903、输入单元904、传感器905、显示单元906、用户输入单元907、接口单元908、存储器909以及处理器910等中的至少部分部件。
本领域技术人员可以理解，终端900还可以包括给各个部件供电的电源(比如电池)，电源可以通过电源管理系统与处理器910逻辑相连，从而通过电源管理系统实现管理充电、放电、以及功耗管理等功能。图9中示出的终端结构并不构成对终端的限定，终端可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置，在此不再赘述。
应理解的是，本申请实施例中，输入单元904可以包括图形处理单元(Graphics Processing Unit，GPU)9041和麦克风9042，图形处理器9041对在视频捕获模式或图像捕获模式中由图像捕获装置(如摄像头)获得的静态图片或视频的图像数据进行处理。显示单元906可包括显示面板9061，可以采用液晶显示器、有机发光二极管等形式来配置显示面板9061。用户输入单元907包括触控面板9071以及其他输入设备9072中的至少一种。触控面板9071，也称为触摸屏。触控面板9071可包括触摸检测装置和触摸控制器两个部分。其他输入设备9072可以包括但不限于物理键盘、功能键(比如音量控制按键、开关按键等)、轨迹球、鼠标、操作杆，在此不再赘述。
本申请实施例中，射频单元901接收来自网络侧设备的下行数据后，可以传输给处理器910进行处理；另外，射频单元901可以向网络侧设备发送上行数据。通常，射频单元901包括但不限于天线、放大器、收发信机、耦合器、低噪声放大器、双工器等。
存储器909可用于存储软件程序或指令以及各种数据。存储器909可主要包括存储程序或指令的第一存储区和存储数据的第二存储区，其中，第一存储区可存储操作系统、至少一个功能所需的应用程序或指令(比如声音播放功能、图像播放功能等)等。此外，存储器909可以包括易失性存储器或非易失性存储器，或者，存储器909可以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)，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，DDRSDRAM)、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)和直接内存总线随机存取存储器(Direct Rambus RAM，DRRAM)。本申请实 施例中的存储器909包括但不限于这些和任意其它适合类型的存储器。
处理器910可包括一个或多个处理单元；可选地，处理器910集成应用处理器和调制解调处理器，其中，应用处理器主要处理涉及操作系统、用户界面和应用程序等的操作，调制解调处理器主要处理无线通信信号，如基带处理器。可以理解的是，上述调制解调处理器也可以不集成到处理器910中。
其中，射频单元901，用于获取目标信道的第一信道信息；
处理器910，用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
射频单元901，还用于向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
可选地，处理器910执行的所述根据所述第一信道信息获取N个系数，包括：
分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数；或者，
将所述第一信道信息分别投影在N个目标正交基向量上，得到N个系数，其中，所述目标正交基向量由所述第二AI网络模型确定。
可选地，所述目标正交基向量包括以下至少一项：
空域正交基向量；
频域正交基向量；
多普勒域正交基向量。
可选地，射频单元901，还用于接收来自所述网络侧设备的第一指示信息，所述第一指示信息用于指示所述目标正交基向量，其中，所述目标正交基向量由所述网络侧设备采用所述第二AI网络模型训练得到；
或者，
处理器910，还用于采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量；
或者，
处理器910，还用于获取通信协议中约定所述目标正交基向量。
可选地，在处理器910执行所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量之后，射频单元901，还用于向所述网络侧设备发送所述目标正交基向量。
可选地，处理器910执行的所述分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数，包括：
将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系 数，其中，所述M个第一AI网络模型与所述M个目标正交基向量一一对应，M为大于或等于N的整数；
从所述M个目标正交基向量各自对应的系数中确定N个系数；
其中，所述第一信息包括：
所述第一信道信息；或者，
所述第一信道信息投影在完备正交空间的第一系数；或者，
所述第一系数，以及所述第一系数对应的正交基向量加权后的部分或全部；或者，
所述第一信道信息经过预编码计算，得到的预编码矩阵或特定层的预编码矩阵；或者，
所述预编码矩阵投影在完备正交空间的第二系数；或者，
所述第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部。
可选地，所述第一系数包括：
所述第一信道信息投影在完备正交空间的系数中的至少部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的幅度大于或者等于预设幅度的部分；或者，
所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分按照幅度值大小的排序。
可选地，射频单元901，还用于向所述网络侧设备发送所述N个系数各自对应的第一AI网络模型的标识信息。
可选地，处理器910执行的所述将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，包括：
将所述第一信道信息输入至M个第一AI网络模型，得到M个第一正交基向量，其中，所述第一正交基向量为加权后的正交向量；
根据预设向量和所述M个第一正交基向量，确定M个系数。
可选地，在所述第一AI网络模型构成空域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设频域正交基对应的空域正交基向量的系数或非零系数。
可选地，在所述第一AI网络模型构成频域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
分别与预设空域正交基向量的每一个极化对应的频域正交基向量的系数或非零系数；和/或，
根据目标时延域信道信息确定的预设空域正交基向量的系数，其中，所述目标时延域信道信息是采用所述第一AI网络模型对所述第一信道信息进行处理得到。
可选地，在处理器910执行所述将第一信息输入至M个第一AI网络模型之前，射频 单元901，还用于接收来自所述网络侧设备的所述M个第一AI网络模型。
可选地，射频单元901在执行所述接收来自所述网络侧设备的所述M个第一AI网络模型之前，还用于：
向所述网络侧设备发送目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
可选地，所述系数为包括实部和虚部的复数，或者所述系数为包括至少一个幅度和/或相位的实数。
本申请实施例提供的终端900，能够执行如图6所示信道特征信息上报装置600中的各模块执行的各个过程，且能够取得相同的有益效果，为避免重复，在此不再赘述。
本申请实施例还提供一种网络侧设备，包括处理器和通信接口，通信接口用于接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；所述处理器用于对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
该网络侧设备实施例与上述网络侧设备方法实施例对应，上述方法实施例的各个实施过程和实现方式均可适用于该网络侧设备实施例中，且能达到相同的技术效果。
具体地，本申请实施例还提供了一种网络侧设备。如图10所示，该网络侧设备1000包括：天线1001、射频装置1002、基带装置1003、处理器1004和存储器1005。天线1001与射频装置1002连接。在上行方向上，射频装置1002通过天线1001接收信息，将接收的信息发送给基带装置1003进行处理。在下行方向上，基带装置1003对要发送的信息进行处理，并发送给射频装置1002，射频装置1002对收到的信息进行处理后经过天线1001发送出去。
以上实施例中网络侧设备执行的方法可以在基带装置1003中实现，该基带装置1003包括基带处理器。
基带装置1003例如可以包括至少一个基带板，该基带板上设置有多个芯片，如图10所示，其中一个芯片例如为基带处理器，通过总线接口与存储器1005连接，以调用存储器1005中的程序，执行以上方法实施例中所示的网络设备操作。
该网络侧设备还可以包括网络接口1006，该接口例如为通用公共无线接口(Common Public Radio Interface，CPRI)。
具体地，本申请实施例的网络侧设备1000还包括：存储在存储器1005上并可在处理器1004上运行的指令或程序，处理器1004调用存储器1005中的指令或程序执行图7所示各模块执行的方法，并达到相同的技术效果，为避免重复，故不在此赘述。
本申请实施例还提供一种可读存储介质，所述可读存储介质上存储有程序或指令，该 程序或指令被处理器执行时实现如图2或图5所示方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
其中，所述处理器为上述实施例中所述的终端中的处理器。所述可读存储介质，包括计算机可读存储介质，如计算机只读存储器ROM、随机存取存储器RAM、磁碟或者光盘等。
本申请实施例另提供了一种芯片，所述芯片包括处理器和通信接口，所述通信接口和所述处理器耦合，所述处理器用于运行程序或指令，实现如图2或图5所示方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
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。
本申请实施例另提供了一种计算机程序/程序产品，所述计算机程序/程序产品被存储在存储介质中，所述计算机程序/程序产品被至少一个处理器执行以实现如图2或图5所示方法实施例的各个过程，且能达到相同的技术效果，为避免重复，这里不再赘述。
本申请实施例还提供了一种通信系统，包括：终端及网络侧设备，所述终端可用于执行如第一方面所述的信道特征信息上报方法的步骤，所述网络侧设备可用于执行如第三方面所述的信道特征信息恢复方法的步骤。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。此外，需要指出的是，本申请实施方式中的方法和装置的范围不限按示出或讨论的顺序来执行功能，还可包括根据所涉及的功能按基本同时的方式或按相反的顺序来执行功能，例如，可以按不同于所描述的次序来执行所描述的方法，并且还可以添加、省去、或组合各种步骤。另外，参照某些示例所描述的特征可在其他示例中被组合。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。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可以以计算机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(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本申请各个实施例所述的方法。
上面结合附图对本申请的实施例进行了描述，但是本申请并不局限于上述的具体实施方式，上述的具体实施方式仅仅是示意性的，而不是限制性的，本领域的普通技术人员在本申请的启示下，在不脱离本申请宗旨和权利要求所保护的范围情况下，还可做出很多形式，均属于本申请的保护之内。

Claims (28)

  1. 一种信道特征信息上报方法，包括：
    终端获取目标信道的第一信道信息；
    所述终端根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一人工智能AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
    所述终端向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
  2. 根据权利要求1所述的方法，其中，所述终端根据所述第一信道信息获取N个系数，包括：
    所述终端分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数；或者，
    所述终端将所述第一信道信息分别投影在N个目标正交基向量上，得到N个系数，其中，所述目标正交基向量由所述第二AI网络模型确定。
  3. 根据权利要求2所述的方法，其中，所述方法还包括：
    所述终端接收来自所述网络侧设备的第一指示信息，所述第一指示信息用于指示所述目标正交基向量，其中，所述目标正交基向量由所述网络侧设备采用所述第二AI网络模型训练得到；
    或者，
    所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量；
    或者，
    所述终端获取通信协议中约定所述目标正交基向量。
  4. 根据权利要求3所述的方法，其中，在所述终端采用第二AI网络模型，基于所述第一信道信息训练得到所述目标正交基向量之后，所述方法还包括：
    所述终端向所述网络侧设备发送所述目标正交基向量。
  5. 根据权利要求2所述的方法，其中，所述终端分别采用N个所述第一AI网络模型对所述第一信道信息进行处理，得到N个系数，包括：
    所述终端将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，其中，所述M个第一AI网络模型与所述M个目标正交基向量一一对应，M为大于或等于N的整数；
    所述终端从所述M个目标正交基向量各自对应的系数中确定N个系数；
    其中，所述第一信息包括：
    所述第一信道信息；或者，
    所述第一信道信息投影在完备正交空间的第一系数；或者，
    所述第一系数，以及所述第一系数对应的正交基向量加权后的部分或全部；或者，
    所述第一信道信息经过预编码计算，得到的预编码矩阵或特定层的预编码矩阵；或者，
    所述预编码矩阵投影在完备正交空间的第二系数；或者，
    所述第二系数，以及所述第二系数对应的正交基向量加权后的部分或全部。
  6. 根据权利要求5所述的方法，其中，所述第一系数包括：
    所述第一信道信息投影在完备正交空间的系数中的至少部分；或者，
    所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分；或者，
    所述第一信道信息投影在完备正交空间的系数中的幅度大于或者等于预设幅度的部分；或者，
    所述第一信道信息投影在完备正交空间的系数中的大于或者等于预设系数的部分按照幅度值大小的排序。
  7. 根据权利要求5所述的方法，所述方法还包括：
    所述终端向所述网络侧设备发送所述N个系数各自对应的第一AI网络模型的标识信息。
  8. 根据权利要求5所述的方法，其中，所述终端将第一信息输入至M个第一AI网络模型，得到M个目标正交基向量各自对应的系数，包括：
    所述终端将所述第一信道信息输入至M个第一AI网络模型，得到M个第一正交基向量，其中，所述第一正交基向量为加权后的正交向量；
    所述终端根据预设向量和所述M个第一正交基向量，确定M个系数。
  9. 根据权利要求5所述的方法，其中，在所述第一AI网络模型构成空域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括：
    分别与预设频域正交基对应的空域正交基向量的系数或非零系数；
    和/或，
    在所述第一AI网络模型构成频域正交基向量的情况下，所述M个目标正交基向量各自对应的系数，包括以下至少一项：
    分别与预设空域正交基向量的每一个极化对应的频域正交基向量的系数或非零系数；
    根据目标时延域信道信息确定的预设空域正交基向量的系数，其中，所述目标时延域信道信息是采用所述第一AI网络模型对所述第一信道信息进行处理得到。
  10. 根据权利要求5所述的方法，其中，在所述终端将第一信息输入至M个第一AI网络模型之前，所述方法还包括：
    所述终端接收来自所述网络侧设备的所述M个第一AI网络模型。
  11. 根据权利要求10所述的方法，其中，在所述终端接收来自所述网络侧设备的所述M个第一AI网络模型之前，所述方法还包括：
    所述终端向所述网络侧设备发送目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
  12. 一种信道特征信息上报装置，应用于终端，所述装置包括：
    第一获取模块，用于获取目标信道的第一信道信息；
    第二获取模块，用于根据所述第一信道信息获取N个系数，其中，所述N个系数为分别采用N个第一AI网络模型确定的系数，所述N个系数与所述N个第一AI网络模型一一对应，或者，所述N个系数为采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
    第一发送模块，用于向网络侧设备发送第一信道特征信息，所述第一信道特征信息包括所述N个系数。
  13. 一种信道特征信息恢复方法，包括：
    网络侧设备接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
    所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
  14. 根据权利要求13所述的方法，其中，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，所述方法还包括：
    所述网络侧设备接收来自所述终端的所述N个系数各自对应的第一AI网络模型的标识信息。
  15. 根据权利要求14所述的方法，其中，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
    所述网络侧设备根据所述N个第一AI网络模型的标识信息确定所述N个系数对应的目标正交基向量；
    所述网络侧设备根据所述目标正交基向量和所述N个系数，恢复所述第一信道信息。
  16. 根据权利要求14所述的方法，其中，所述第一AI网络模型与目标正交基向量一一对应，所述网络侧设备对所述第一信道特征信息进行恢复处理，包括：
    所述网络侧设备采用第三AI网络模型对所述第一信道特征信息进行恢复处理，其中，所述第三AI网络模型的输入长度为M，M等于所述目标正交基向量的最大数量，M为大于或等于N的整数。
  17. 根据权利要求16所述的方法，其中，在M大于N的情况下，所述网络侧设备采用第三AI网络模型对所述第一信道特征信息进行恢复处理，包括：
    所述网络侧设备对所述第一信道特征信息进行第一处理，以使所述第一信道特征信息的长度由N个系数的长度调整为M个系数的长度；
    所述网络侧设备采用第三AI网络模型对所述第一处理后的第一信道特征信息进行恢复处理。
  18. 根据权利要求13所述的方法，其中，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
    所述网络侧设备采用N个第四AI网络模型对各自对应的系数进行恢复处理，所述N个系数与所述N个第四AI网络模型一一对应；
    所述网络侧设备根据所述N个第四AI网络模型各自恢复得到的信道信息，确定所述第一信道信息。
  19. 根据权利要求13所述的方法，其中，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
    所述网络侧设备根据预设正交基向量和所述第一信道特征信息，恢复第二信道信息；
    所述网络侧设备采用第五AI网络模型对所第二信道信息进行矫正，得到所述第一信道信息。
  20. 根据权利要求13所述的方法，其中，在所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
    所述网络侧设备根据N个第二正交基向量对所述N个系数进行加权处理，得到所述第一信道信息，其中，所述N个第二正交基向量与所述N个第一AI网络模型一一对应，且每一个所述第二正交基向量与各自对应的第一AI网络模型联合训练得到。
  21. 根据权利要求13所述的方法，其中，在所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数的情况下，所述网络侧设备对所述第一信道特征信息进行恢复处理，得到所述第一信道信息，包括：
    所述网络侧设备采用所述第二AI网络模型确定所述N个目标正交基向量，或者，所述网络侧设备接收来自所述终端的所述目标正交基向量；
    所述网络侧设备根据所述N个目标正交基向量和所述N个目标正交基向量各自对应的系数，恢复所述第一信道信息。
  22. 根据权利要求21所述的方法，其中，在所述网络侧设备采用所述第二AI网络模型确定所述N个目标正交基向量之后，所述方法还包括：
    所述网络侧设备向所述终端发送第一指示信息，所述第一指示信息用于指示所述N个目标正交基向量。
  23. 根据权利要求13所述的方法，其中，在所述网络侧设备接收来自终端的第一信道 特征信息之前，所述方法还包括：
    所述网络侧设备向所述终端发送M个第一AI网络模型，所述M个第一AI网络模型包括所述N个第一AI网络模型。
  24. 根据权利要求23所述的方法，其中，在所述网络侧设备向所述终端发送M个第一AI网络模型之前，所述方法还包括：
    所述网络侧设备接收来自所述终端的目标能力信息，所述目标能力信息用于指示所述终端支持的第一AI网络模型的最大数量，其中，M小于或者等于所述终端支持的第一AI网络模型的最大数量。
  25. 一种信道特征信息恢复装置，应用于网络侧设备，所述装置包括：
    第一接收模块，用于接收来自终端的第一信道特征信息，其中，所述第一信道特征信息包括N个系数，所述N个系数为分别采用N个第一AI网络模型对第一信道信息进行处理得到的系数，或者，所述N个系数为所述第一信道信息投影在采用第二AI网络模型选择的N个目标正交基向量的系数，N为大于或者等于1的整数；
    第一处理模块，用于对所述第一信道特征信息进行恢复处理，得到所述第一信道信息。
  26. 一种终端，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求1至11中任一项所述的信道特征信息上报方法的步骤。
  27. 一种网络侧设备，包括处理器和存储器，所述存储器存储可在所述处理器上运行的程序或指令，所述程序或指令被所述处理器执行时实现如权利要求13至24中任一项所述的信道特征信息恢复方法的步骤。
  28. 一种可读存储介质，所述可读存储介质上存储程序或指令，所述程序或指令被处理器执行时实现如权利要求1至11中任一项所述的信道特征信息上报方法，或者实现如权利要求13至24中任一项所述的信道特征信息恢复方法的步骤。
PCT/CN2023/084962 2022-04-01 2023-03-30 信道特征信息上报及恢复方法、终端和网络侧设备 WO2023185978A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210349421.XA CN116939705A (zh) 2022-04-01 2022-04-01 信道特征信息上报及恢复方法、终端和网络侧设备
CN202210349421.X 2022-04-01

Publications (1)

Publication Number Publication Date
WO2023185978A1 true WO2023185978A1 (zh) 2023-10-05

Family

ID=88199355

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/084962 WO2023185978A1 (zh) 2022-04-01 2023-03-30 信道特征信息上报及恢复方法、终端和网络侧设备

Country Status (2)

Country Link
CN (1) CN116939705A (zh)
WO (1) WO2023185978A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117807383B (zh) * 2024-03-01 2024-07-02 深圳市大数据研究院 信道状态信息恢复方法及装置、设备、存储介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111277360A (zh) * 2019-01-11 2020-06-12 维沃移动通信有限公司 信道状态信息csi报告的传输方法、终端及网络设备
CN111277379A (zh) * 2019-01-11 2020-06-12 维沃移动通信有限公司 无线通信方法和终端设备
WO2021142631A1 (en) * 2020-01-14 2021-07-22 Nokia Shanghai Bell Co., Ltd. Method, device and computer readable medium of communication
WO2022040660A1 (en) * 2020-08-18 2022-02-24 Qualcomm Incorporated Neural network based channel state information feedback report size determination

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111277360A (zh) * 2019-01-11 2020-06-12 维沃移动通信有限公司 信道状态信息csi报告的传输方法、终端及网络设备
CN111277379A (zh) * 2019-01-11 2020-06-12 维沃移动通信有限公司 无线通信方法和终端设备
WO2021142631A1 (en) * 2020-01-14 2021-07-22 Nokia Shanghai Bell Co., Ltd. Method, device and computer readable medium of communication
WO2022040660A1 (en) * 2020-08-18 2022-02-24 Qualcomm Incorporated Neural network based channel state information feedback report size determination

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
VIVO: "Study on AI/ML based air interface enhancement in Rel-18", 3GPP TSG RAN MEETING #91E, RP-210321, 15 March 2021 (2021-03-15), XP051985674 *

Also Published As

Publication number Publication date
CN116939705A (zh) 2023-10-24

Similar Documents

Publication Publication Date Title
WO2023246618A1 (zh) 信道矩阵处理方法、装置、终端及网络侧设备
US20230412430A1 (en) Inforamtion reporting method and apparatus, first device, and second device
WO2023185978A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2023179476A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2023179473A1 (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2023179474A1 (zh) 信道特征信息辅助上报及恢复方法、终端和网络侧设备
WO2024032606A1 (zh) 信息传输方法、装置、设备、系统及存储介质
WO2023185980A1 (zh) 信道特征信息传输方法、装置、终端及网络侧设备
WO2024037380A1 (zh) 信道信息处理方法、装置、通信设备及存储介质
WO2024088162A1 (zh) 信息传输方法、信息处理方法、装置和通信设备
WO2024217495A1 (zh) 信息处理方法、信息处理装置、终端及网络侧设备
WO2024104126A1 (zh) 更新ai网络模型的方法、装置和通信设备
WO2023185995A1 (zh) 信道特征信息传输方法、装置、终端及网络侧设备
CN117411527A (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
WO2023179570A1 (zh) 信道特征信息传输方法、装置、终端及网络侧设备
WO2023134628A1 (zh) 传输方法、装置和设备
WO2024088161A1 (zh) 信息传输方法、信息处理方法、装置和通信设备
WO2024055974A1 (zh) Cqi传输方法、装置、终端及网络侧设备
WO2024007949A1 (zh) Ai模型处理方法、装置、终端及网络侧设备
WO2023179460A1 (zh) 信道特征信息传输方法、装置、终端及网络侧设备
WO2023179540A1 (zh) 信道预测方法、装置及无线通信设备
WO2024055993A1 (zh) Cqi传输方法、装置、终端及网络侧设备
CN117335849A (zh) 信道特征信息上报及恢复方法、终端和网络侧设备
US20240356595A1 (en) Uplink precoding method and apparatus
WO2024041420A1 (zh) 测量反馈处理方法、装置、终端及网络侧设备

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: 23778343

Country of ref document: EP

Kind code of ref document: A1