Method and device for acquiring channel quality, storage medium and chip

Abstract

The invention relates to a method and device for obtaining channel quality, a storage medium and a chip. The method comprises: a terminal device receiving a pilot signal sent by a network device through a downlink channel; acquiring a first channel matrix according to the pilot signal; compressing the first channel matrix according to a channel state information (CSI) compression model and channel state information (CSI) compression parameters to obtain a compressed target channel matrix; and sending the target channel matrix to the network equipment, so that the network equipment determines the channel quality of the downlink channel according to the target channel matrix. Wherein the first channel matrix is used for representing the channel quality of a downlink channel; the CSI compression model comprises a channel encoder, and the channel encoder comprises a plurality of sub-encoders; different sub-encoders correspond to different CSI compression parameters; the target channel matrix is used for indicating the network equipment to determine the channel quality of the downlink channel.

Description

Method, device, storage medium and chip for obtaining channel quality Technical Field

[0001] The present disclosure relates to the field of communication technology, and in particular, to a method, device, storage medium, and chip for obtaining channel quality. Background Art

[0002] As a key technology for 5G (the fifth generation mobile communication technology), massive multiple-input multiple-output (mMIMO) has been widely researched and deployed in the communications field in recent years. By deploying a large number of antennas on the transmitter side, either centralized or distributed, mMIMO systems offer excellent performance in terms of system stability, energy efficiency, and interference mitigation. To fully leverage the advantages of mMIMO systems, accurate channel state information (CSI) is required on the transmitter side. For example, terminal devices can report CSI information to network equipment, enabling the network equipment to determine the downlink channel quality and select appropriate modulation and coding schemes for downlink transmission based on the channel quality, thereby improving data transmission performance. However, in mMIMO, as the number of antennas continues to increase, the CSI contains increasingly rich information, resulting in increasing resource overhead for CSI reporting.

[0003] To reduce CSI reporting overhead, related technologies employ CSI compression technology based on the Discrete Fourier Transform (DFT). Terminal devices compress CSI before reporting it to network devices. However, CSI compression reduces the accuracy of channel quality acquired by network devices. In particular, the accuracy of compressed CSI varies significantly across different scenarios, impacting data transmission efficiency.

[0004] Summary of the Invention

[0005] To overcome the above-mentioned problems existing in the related art, the present disclosure provides a method, device, storage medium and chip for obtaining channel quality.

[0006] According to a first aspect of an embodiment of the present disclosure, a method for obtaining channel quality is provided, which is applied to a terminal device. The method includes:

[0007] receiving a pilot signal sent by a network device via a downlink channel;

[0008] Acquire a first channel matrix according to the pilot signal; the first channel matrix is ​​used to characterize the channel quality of the downlink channel;

[0009] compressing the first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; wherein the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters;

[0010] The target channel matrix is ​​sent to the network device, so that the network device obtains the channel quality of the downlink channel according to the target channel matrix.

[0011] According to a second aspect of an embodiment of the present disclosure, a method for obtaining channel quality is provided, which is applied to a network device. The method includes:

[0012] receiving a target channel matrix sent by a terminal device; the target channel matrix is ​​obtained by the terminal device by compressing a first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter, where the first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal and used to characterize the channel quality of a downlink channel;

[0013] decompressing the target channel matrix according to a channel state information (CSI) decompression model and the CSI compression parameter to obtain a third channel matrix; wherein the CSI decompression model includes a channel decoder, the channel decoder includes a plurality of sub-decoders, and different sub-decoders correspond to different CSI compression parameters;

[0014] The channel quality of the downlink channel is determined according to the third channel matrix.

[0015] According to a third aspect of an embodiment of the present disclosure, there is provided an apparatus for obtaining channel quality, which is applied to a terminal device, and includes:

[0016] A first receiving module is configured to receive a pilot signal sent by a network device through a downlink channel;

[0017] a first matrix acquisition module, configured to acquire a first channel matrix according to the pilot signal; the first channel matrix is ​​used to characterize the channel quality of the downlink channel;

[0018] a target matrix acquisition module configured to compress the first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; wherein the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters;

[0019] The first sending module is configured to send the target channel matrix to the network device, so that the network device obtains the channel quality of the downlink channel according to the target channel matrix.

[0020] According to a fourth aspect of an embodiment of the present disclosure, there is provided an apparatus for obtaining channel quality, which is applied to a network device, and the apparatus includes:

[0021] A second receiving module is configured to receive a target channel matrix sent by a terminal device; the target channel matrix is ​​obtained by the terminal device by compressing a first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter, where the first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal and used to characterize the channel quality of a downlink channel;

[0022] a third matrix acquisition module configured to decompress the target channel matrix according to a channel state information (CSI) decompression model and the CSI compression parameters to obtain a third channel matrix; wherein the CSI decompression model includes a channel decoder, the channel decoder includes a plurality of sub-decoders, and different sub-decoders correspond to different CSI compression parameters;

[0023] The channel quality determination module is configured to determine the channel quality of the downlink channel according to the third channel matrix.

[0024] According to a fifth aspect of an embodiment of the present disclosure, there is provided an apparatus for obtaining channel quality, including:

[0025] processor;

[0026] a memory for storing processor-executable instructions;

[0027] The processor is configured to execute the steps of the method for obtaining channel quality provided in the first aspect of the present disclosure.

[0028] According to a sixth aspect of an embodiment of the present disclosure, there is provided an apparatus for obtaining channel quality, including:

[0029] processor;

[0030] a memory for storing processor-executable instructions;

[0031] The processor is configured to execute the steps of the method for obtaining channel quality provided in the second aspect of the present disclosure.

[0032] According to a seventh aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the steps of the method for obtaining channel quality provided in the first aspect of the present disclosure are implemented.

[0033] According to an eighth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the steps of the method for obtaining channel quality provided in the second aspect of the present disclosure are implemented.

[0034] According to a ninth aspect of an embodiment of the present disclosure, a chip is provided, comprising: a processor and an interface; the processor is configured to read instructions to execute the steps of the method for obtaining channel quality provided in the first aspect of the present disclosure.

[0035] According to a tenth aspect of an embodiment of the present disclosure, a chip is provided, comprising: a processor and an interface; the processor is used to read instructions to execute the steps of the method for obtaining channel quality provided in the second aspect of the present disclosure.

[0036] The technical solution provided by the embodiments of the present disclosure may include the following beneficial effects: a terminal device receives a pilot signal sent by a network device through a downlink channel; obtains a first channel matrix based on the pilot signal; compresses the first channel matrix based on a channel state information (CSI) compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; and sends the target channel matrix to the network device so that the network device determines the channel quality of the downlink channel based on the target channel matrix. The first channel matrix is ​​used to characterize the channel quality of the downlink channel; the CSI compression model may include a channel encoder, which includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters; and the target channel matrix is ​​used to instruct the network device to determine the channel quality of the downlink channel. In this way, different CSI compression parameters can be adapted through multiple sub-encoders, so that in scenarios where the CSI compression rate changes, a more accurate target channel matrix can be adaptively obtained and sent to the network device so that the network device can obtain more accurate channel quality, thereby improving data transmission efficiency.

[0037] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0039] Fig. 1 is a block diagram showing a communication system according to an exemplary embodiment.

[0040] Fig. 2 is a flow chart showing a method for obtaining channel quality according to an exemplary embodiment.

[0041] Fig. 3 is a flow chart showing a method for obtaining channel quality according to an exemplary embodiment.

[0042] Fig. 4 is a schematic structural diagram showing a network model for obtaining channel quality according to an exemplary embodiment.

[0043] Fig. 5 is a schematic diagram showing a feature converter in a CSI compression model according to an exemplary embodiment.

[0044] Fig. 6 is a schematic diagram showing a channel encoder in a CSI compression model according to an exemplary embodiment.

[0045] Fig. 7 is a schematic diagram showing a CSI decompression model according to an exemplary embodiment.

[0046] FIG8 is a flowchart showing a method for training a CSI compression model according to an exemplary embodiment.

[0047] FIG9 is a flowchart showing a method for training a CSI decompression model according to an exemplary embodiment.

[0048] Fig. 10 is a flowchart showing a method for obtaining channel quality according to an exemplary embodiment.

[0049] Fig. 11 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment.

[0050] Fig. 12 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment.

[0051] Fig. 13 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment.

[0052] Fig. 14 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment.

[0053] Fig. 15 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment.

[0054] Fig. 16 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment. DETAILED DESCRIPTION

[0055] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0056] It should be noted that all actions of acquiring signals, information or data in the present disclosure are carried out in compliance with the corresponding data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0057] In this disclosure, terms such as "first," "second," etc. are used to distinguish similar objects and are not necessarily to be understood as implying a particular order or precedence. In addition, in the description with reference to the accompanying drawings, the same reference numerals in different drawings represent the same elements unless otherwise indicated.

[0058] In the description of this disclosure, unless otherwise specified, "multiple" means two or more than two, and other quantifiers are similar; "at least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural; "and / or" is a type of association relationship that describes associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural.

[0059] Although operations are described in a particular order in the accompanying drawings in the disclosed embodiments, this should not be understood as requiring that the operations be performed in the particular order shown or in a serial order, or that all of the operations shown be performed to obtain the desired results. In certain circumstances, multitasking and parallel processing may be advantageous.

[0060] With the adoption of massive MIMO technology, CSI compression technology based on discrete Fourier transform (DFT) can be used to reduce CSI reporting overhead. Terminal devices compress CSI before reporting it to network devices. However, CSI compression reporting reduces the accuracy of channel quality obtained by network devices, affecting data transmission efficiency.

[0061] For example, a terminal device can use a pre-trained encoding neural network to compress CSI and report the compressed CSI to the network device. The network device also uses a decoding neural network to decompress the compressed CSI to determine the channel quality. However, due to different CSI compression rates, the requirements and parameters of the encoding neural network and decoding neural network are different. Therefore, if the same encoding neural network and decoding neural network are used in scenarios where the CSI compression rate changes, the accuracy of CSI reporting will vary significantly. When the CSI compression rate changes, retraining the encoding neural network and decoding neural network is inefficient and cannot adapt to scenarios where the CSI compression rate frequently changes.

[0062] In order to solve the above problems, the present disclosure provides a method, device, storage medium and chip for obtaining channel quality.

[0063] The following first introduces the implementation environment of the embodiments of the present disclosure.

[0064] FIG1 is a schematic diagram of a communication system according to an exemplary embodiment. As shown in FIG1 , the communication system may include a terminal device 101 and a network device 102. The communication system may be used to support 4G (the 4th Generation, fourth generation) network access technology, such as Long Term Evolution (LTE) access technology, or 5G (the 5th Generation, fifth generation) network access technology, such as New Radio Access Technology (New RAT), or other future wireless communication technologies. It should be noted that the communication system may be a communication system using FDD (Frequency Division Duplexing, frequency division duplexing) technology, or a communication system using TDD (Time Division Duplexing, time division duplexing) technology. In addition, in the communication system, the number of network devices and terminal devices may be one or more. The number of network devices and terminal devices in the communication system shown in FIG1 is only an adaptive example, and the present disclosure does not limit this.

[0065] The network device in FIG1 can be used to support terminal access, for example, it can be an evolutionary Node B (eNB or eNodeB) in LTE; or a base station in a 5G network or a future evolved public land mobile network (PLMN), a broadband network gateway (BNG), an aggregation switch or a non-3GPP (3rd Generation Partnership Project) access device, etc. Optionally, the network device in the embodiment of the present disclosure may include various forms of base stations, such as: a macro base station, a micro base station (also known as a small base station), a relay station, an access point, a 5G base station or a future base station, a satellite, a transmitting and receiving point (TRP), a transmitting point (TP), a mobile switching center, and a device that performs base station functions in device-to-device (D2D), vehicle-to-everything (V2X), and machine-to-machine (M2M) communications, etc., and the embodiment of the present disclosure does not specifically limit this. For the convenience of description, in all embodiments of the present disclosure, the devices that provide wireless communication functions for terminal devices are collectively referred to as network devices or base stations.

[0066] The terminal device in FIG1 can be an electronic device that provides voice or data connectivity, and can also be referred to as user equipment (UE), subscriber unit (Subscriber Unit), mobile station (Mobile Station), station (Station), terminal, etc. For example, the terminal device can include a smartphone, smart wearable device, smart speaker, smart tablet, wireless modem, wireless local loop (WLL) station, PDA (Personal Digital Assistant), CPE (Customer Premise Equipment), etc. With the development of wireless communication technology, devices that can access a communication system, communicate with network devices of a communication system, or communicate with other objects through a communication system can be terminal devices in the embodiments of the present disclosure, such as terminals and cars in smart transportation, household appliances in smart homes, power meter reading instruments, voltage monitoring instruments, environmental monitoring instruments in smart grids, video surveillance instruments in smart security networks, cash registers, etc. In the embodiments of the present disclosure, the terminal device can communicate with network devices, such as the network devices in FIG1. ​​Multiple terminals can also communicate with each other. The terminal may be statically fixed or mobile, which is not limited in this disclosure.

[0067] FIG2 is a flow chart of a method for obtaining channel quality according to an exemplary embodiment. The method can be applied to a terminal device. As shown in FIG2 , the method may include:

[0068] S201. A terminal device receives a pilot signal sent by a network device through a downlink channel.

[0069] For example, in the above communication system, the network device may send a pilot signal to the terminal device via a downlink channel, and the terminal device may receive the pilot signal accordingly.

[0070] In some embodiments, the pilot signal may include a Channel State Information Reference Signal (CSI-RS).

[0071] S202. The terminal device obtains a first channel matrix according to the pilot signal.

[0072] The first channel matrix can be used to characterize the channel quality of the downlink channel.

[0073] For example, the terminal device can perform channel state information CSI estimation based on the received pilot signal (such as CSI-RS) to obtain a CSI estimation matrix; and then obtain a first channel matrix representing the downlink channel quality based on the CSI estimation matrix.

[0074] S203. The terminal device compresses the first channel matrix according to the channel state information CSI compression model and the channel state information CSI compression parameters to obtain a compressed target channel matrix.

[0075] The CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters.

[0076] In some embodiments, the terminal device may determine one or more target subcoders from a plurality of subcoders based on the CSI compression parameters, and compress the first channel matrix through the target subcoder to obtain a target channel matrix.

[0077] For example, the CSI compression parameter may represent a CSI compression rate, and the value of the CSI compression rate may be any preset value, for example, 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32, or 1 / 64, etc., which is not limited in the present disclosure.

[0078] S204: The terminal device sends the target channel matrix to the network device.

[0079] By sending the target channel matrix, the network device can determine the channel quality of the downlink channel according to the target channel matrix.

[0080] Using the above method, the terminal device receives a pilot signal sent by the network device through the downlink channel; obtains a first channel matrix based on the pilot signal; compresses the first channel matrix based on the channel state information CSI compression model and the channel state information CSI compression parameter to obtain a compressed target channel matrix; and sends the target channel matrix to the network device so that the network device can determine the channel quality of the downlink channel based on the target channel matrix. The first channel matrix is ​​used to characterize the channel quality of the downlink channel; the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters; and the target channel matrix is ​​used to instruct the network device to determine the channel quality of the downlink channel. In this way, different CSI compression parameters can be adapted through multiple sub-encoders, so that in scenarios where the CSI compression rate changes, a more accurate target channel matrix can be adaptively obtained and sent to the network device, so that the network device can obtain more accurate channel quality, thereby improving data transmission efficiency.

[0081] In some embodiments, the terminal device may obtain the first channel matrix in the following manner:

[0082] First, the spatial channel matrix is ​​measured based on the pilot signal.

[0083] Secondly, the spatial domain channel matrix is ​​transformed into the angle delay domain channel matrix through discrete Fourier transform.

[0084] Thirdly, the first channel matrix is ​​determined according to the angle delay domain channel matrix.

[0085] For example, the above communication system may adopt the mMIMO technology based on OFDM (Orthogonal Frequency Division Multiplexing) with the number of subcarriers being N. s , the number of antennas used by the network device for mMIMO can be N t The pilot signal may include a channel state information reference signal CSI-RS. Taking the communication system as an example, the above steps are illustrated as follows:

[0086] First, the terminal device can measure (also called estimate) the spatial channel matrix H based on the received CSI-RS. The size of the spatial channel matrix H can be N s ×N t , the spatial channel matrix H can characterize the channel quality of each subcarrier of each antenna.

[0087] Secondly, the terminal device can transform the spatial channel matrix into the angle delay domain channel matrix through discrete Fourier transform. For example, the angle delay domain channel matrix can be obtained by the following formula (1):

[0088] H a =F d HF a (1)

[0089] Where H represents the above spatial channel matrix, and the size of H can be N s ×N t , H a represents the transformed angle delay domain channel matrix, F d and F a Indicates size N s ×N s , N t ×N t The discrete Fourier transform matrix, N s Indicates the number of subcarriers corresponding to mMIMO, N t Indicates the number of antennas corresponding to mMIMO.

[0090] Finally, the first channel matrix is ​​determined according to the principal value part of the angle delay domain channel matrix.

[0091] It should be noted that due to the influence of multipath delay, the above angle delay domain channel matrix H a Only in the first N c The row has a value, and after intercepting the main value part, the main value channel matrix in the angle delay domain is obtained. The size of the main value channel matrix can be N c ×N t .

[0092] Furthermore, the real and imaginary parts of the principal value channel matrix are separated to obtain the first channel matrix Wherein c is the dimension of the real and imaginary parts of the channel, for example, c may be 2. The first channel matrix may be used as an input of the CSI compression model to obtain a target channel matrix.

[0093] In this way, the first channel matrix representing the downlink channel quality can be obtained according to the pilot signal in the above manner.

[0094] The above-mentioned CSI compression parameters can be parameters received by the terminal device from the network device, or can be parameters preset by the terminal device.

[0095] In some embodiments, the CSI compression parameter may be determined by the terminal device based on a parameter received from the network device. For example, the terminal device may receive a first compression parameter sent by the network device and determine the CSI compression parameter based on the first compression parameter. For example, the terminal device may receive the first compression parameter via RRC (Radio Resource Control) signaling (e.g., broadcast signaling or proprietary signaling for the terminal device).

[0096] In some embodiments, the network device may pre-set a value of a CSI compression parameter, determine a first compression parameter based on the CSI compression parameter, and send the first compression parameter to the terminal device.

[0097] For example, the CSI compression parameter and the first compression parameter may be determined by a preset correspondence relationship between the first compression parameters. For example, when the CSI compression parameter represents the CSI compression rate, a first compression parameter corresponding to a value of 1 / 2 of the CSI compression parameter may be 1, a first compression parameter corresponding to a value of 1 / 4 of the CSI compression parameter may be 2, a first compression parameter corresponding to a value of 1 / 8 of the CSI compression parameter may be 3, and other values ​​may be deduced in the same manner.

[0098] In this way, after the terminal device receives the first compression parameter sent by the network device, it can determine the CSI compression parameter according to the first compression parameter.

[0099] In some embodiments, the network device may update the value of the CSI compression parameter and determine a new first compression parameter based on the updated CSI compression parameter; similarly, the terminal device may also determine a new CSI compression parameter based on the new first compression parameter when receiving the new first compression parameter.

[0100] In other embodiments, the value of the CSI compression parameter may be preset by the terminal device, for example, it may be a preset parameter value of the terminal device, or it may be a parameter value set by the terminal device according to user input.

[0101] In some embodiments, the terminal device may determine a second compression parameter based on a preset value of the CSI compression parameter and send the second compression parameter to the network device, thereby achieving consensus on the use of the same CSI compression parameter. For example, the terminal device may send the second compression parameter to the network device via RRC signaling.

[0102] Likewise, the CSI compression parameter and the second compression parameter may be determined by a preset corresponding relationship between the second compression parameters.

[0103] In some embodiments, the terminal device may determine the value of the CSI compression parameter based on its own device parameters, which may include one or more of the following: the protocol version of the terminal device, the signal quality of the terminal device, the distance between the terminal device and the network device, the uplink data volume of the terminal device, and the downlink data volume of the terminal device. Among them, the signal quality of the terminal device may include RSRP (Reference Signal Receiving Power) or SINR (Signal to Interference plus Noise Ratio).

[0104] In some embodiments, the terminal device may update the value of the CSI compression parameter and determine a new second compression parameter based on the updated CSI compression parameter. For example, upon determining that the aforementioned device parameter has changed, the terminal device may update the value of the CSI compression parameter based on the changed new device parameter. Similarly, upon receiving a new second compression parameter, the network device may determine a new CSI compression parameter based on the new second compression parameter.

[0105] In some other embodiments, the CSI compression parameter may include a third compression parameter pre-set in the terminal device and the network device respectively, and the values ​​of the third compression parameter pre-set in the terminal device and the network device may be the same.

[0106] FIG3 is a flow chart of a method for obtaining channel quality according to an exemplary embodiment. The method can be applied to a network device. As shown in FIG3 , the method may include:

[0107] S301: A network device receives a target channel matrix sent by a terminal device.

[0108] Among them, the target channel matrix is ​​obtained by the terminal device after compressing the first channel matrix according to the channel state information CSI compression model and the channel state information CSI compression parameter. The first channel matrix is ​​a matrix obtained by the terminal device according to the pilot signal for characterizing the channel quality of the downlink channel.

[0109] In some embodiments, the network device may send a pilot signal via a downlink channel so that the terminal device receives the pilot signal and obtains the first channel matrix. For example, the pilot signal may include a channel state information reference signal CSI-RS.

[0110] S302: The network device decompresses the target channel matrix according to the channel state information CSI decompression model and CSI compression parameters to obtain a third channel matrix.

[0111] The CSI decompression model may include a channel decoder, which includes multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters.

[0112] In some embodiments, the network device may determine one or more target sub-decoders from a plurality of sub-decoders according to the CSI compression parameters, and compress the first channel matrix through the target sub-decoders to obtain a target channel matrix.

[0113] For example, the CSI compression parameter may represent a CSI compression rate, and the value of the CSI compression rate may be any preset value, for example, 1 / 2, 1 / 4, 1 / 8, 1 / 16, 1 / 32, or 1 / 64, etc., which is not limited in the present disclosure.

[0114] S303: The network device determines the channel quality of the downlink channel according to the third channel matrix.

[0115] The above method is used to receive a target channel matrix sent by a terminal device; decompress the target channel matrix according to a channel state information (CSI) decompression model and CSI compression parameters to obtain a third channel matrix; and determine the channel quality of the downlink channel according to the third channel matrix. The target channel matrix is ​​obtained by the terminal device after compressing the first channel matrix according to the channel state information (CSI) compression model and the channel state information (CSI) compression parameters. The first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal to characterize the channel quality of the downlink channel. The CSI decompression model may include a channel decoder, and the channel decoder includes multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters. In this way, the network device can adapt to different CSI compression parameters through multiple sub-decoders, so that in scenarios where the CSI compression rate changes, a more accurate third channel matrix can be adaptively obtained, and a more accurate channel quality can be obtained based on the third channel matrix. The network device can determine the modulation and coding strategy corresponding to the downlink channel based on the channel quality, thereby improving data transmission efficiency.

[0116] In some embodiments, the network device may use the sub-decoder corresponding to the CSI compression parameter as a target sub-decoder; decompress the target channel matrix through the target sub-decoder to obtain a fourth channel matrix; and determine the third channel matrix based on the fourth channel matrix.

[0117] Illustratively, the sub-decoder may include a decompression layer, which may correspond to a compression layer of a sub-encoder in a channel encoder of the terminal device.

[0118] In some embodiments, the CSI decompression model may further include a CSI reconstruction module, and the network device may input the fourth channel matrix into the CSI reconstruction module to obtain a third channel matrix.

[0119] For example, the CSI reconstruction module may include a convolutional neural network (CNN).

[0120] In some embodiments, the value of the CSI compression parameter may be preset by the network device, for example, it may be a preset parameter value of the network device, or it may be a parameter value set by the network device according to user input.

[0121] In some embodiments, the network device may determine a first compression parameter based on the CSI compression parameter, and send the first compression parameter to the terminal device to instruct the terminal device to determine the CSI compression parameter based on the first compression parameter. For example, the network device may send the first compression parameter via RRC (Radio Resource Control) signaling (e.g., broadcast signaling or dedicated signaling for the terminal device).

[0122] For example, the CSI compression parameter and the first compression parameter can be determined by a pre-set correspondence relationship between the first compression parameters. For example, when the CSI compression parameter represents the CSI compression rate, the first compression parameter corresponding to a CSI compression parameter value of 1 / 2 can be 1, the first compression parameter corresponding to a CSI compression parameter value of 1 / 4 can be 2, the first compression parameter corresponding to a CSI compression parameter value of 1 / 8 can be 3, and other values ​​can be deduced in the same way. In this way, after receiving the first compression parameter sent by the network device, the terminal device can determine the CSI compression parameter based on the first compression parameter.

[0123] In some embodiments, the network device may set different CSI compression parameter values ​​for different terminal devices. For example, the value of the CSI compression parameter may be determined based on the device parameters corresponding to the terminal device, and the device parameters may include one or more of the following: the protocol version of the terminal device, the signal quality of the terminal device, the distance between the terminal device and the network device, the uplink data volume of the terminal device, and the downlink data volume of the terminal device. Among them, the signal quality of the terminal device may include RSRP (Reference Signal Receiving Power) or SINR (Signal to Interference plus Noise Ratio).

[0124] In other embodiments, the network device may update the value of the CSI compression parameter and determine a new first compression parameter based on the updated CSI compression parameter. For example, the network device may obtain a device parameter corresponding to the terminal device and, if it determines that the device parameter has changed, update the value of the CSI compression parameter.

[0125] Similarly, upon receiving a new first compression parameter, the terminal device may also determine a new CSI compression parameter based on the new first compression parameter.

[0126] In some other embodiments, the CSI compression parameter may be determined by the network device based on a parameter received from the terminal device. For example, the network device may receive a second compression parameter sent by the terminal device and determine the CSI compression parameter based on the second compression parameter. For example, the network device may receive the second compression parameter via RRC signaling.

[0127] For example, the terminal device may pre-set a value for a CSI compression parameter, determine a second compression parameter based on the CSI compression parameter, and send the second compression parameter to the network device. Similarly, the CSI compression parameter and the second compression parameter may be determined by a pre-set correspondence between the second compression parameters. Thus, upon receiving the second compression parameter sent by the terminal device, the network device may determine the CSI compression parameter based on the second compression parameter.

[0128] In some embodiments, the terminal device can update the value of the CSI compression parameter and determine a new second compression parameter based on the updated CSI compression parameter; similarly, the network device can also determine a new CSI compression parameter based on the new second compression parameter when receiving the new second compression parameter.

[0129] In some other embodiments, the CSI compression parameter may also be a third compression parameter preset in the terminal device and the network device respectively, and the values ​​of the third compression parameter preset in the terminal device and the network device may be the same.

[0130] In some embodiments, the CSI compression model and the CSI decompression model may together constitute a network model for obtaining channel quality. The network model for obtaining channel quality is described in detail below with reference to the accompanying drawings.

[0131] Fig. 4 is a schematic structural diagram showing a network model for obtaining channel quality according to an exemplary embodiment.

[0132] As shown in Figure 4, in some embodiments, the network model for obtaining channel quality may include a channel state information CSI compression model 41 and a channel state information CSI decompression model 42. The CSI compression model 41 can be deployed on a terminal device in the communication system shown in Figure 1. For example, the terminal device can run the CSI compression model 41 through software, hardware, or a combination of software and hardware; the CSI decompression model 42 can be deployed on a network device (such as a base station) in the communication system shown in Figure 1. For example, the network device can run the CSI decompression model 42 through software, hardware, or a combination of software and hardware.

[0133] The CSI compression model 41 can compress the input first channel matrix according to the CSI compression parameters and output a target channel matrix (the target channel matrix can also be called a codeword); the terminal device can send the target channel matrix to the network device, and accordingly, the network device can input the received target channel matrix into the CSI decompression model 42. The CSI decompression model 42 can decode (also called decompress) the target channel matrix according to the CSI compression parameters and output a third channel matrix. The network device can determine the channel quality of the downlink channel based on the third channel matrix.

[0134] As shown in FIG. 4 , in some embodiments, the CSI compression model 41 may include a channel encoder 411 . The channel encoder 411 may encode an input channel matrix according to a CSI compression parameter to obtain a target channel matrix.

[0135] Furthermore, the channel encoder 411 may include multiple sub-encoders, and different sub-encoders correspond to different CSI compression parameters. In this way, the terminal device can use the sub-encoder corresponding to the CSI compression parameter as the first target sub-encoder; compress the first channel matrix through the first target sub-encoder to obtain the target channel matrix. The target channel matrix can also be called a codeword

[0136] For example, when the CSI compression parameter includes a CSI compression rate, the channel encoder 411 may include sub-encoder 1 (the corresponding CSI compression parameter is 1 / 2), sub-encoder 2 (the corresponding CSI compression parameter is 1 / 4), sub-encoder 3 (the corresponding CSI compression parameter is 1 / 8), sub-encoder 3 (the corresponding CSI compression parameter is 1 / 16), sub-encoder 4 (the corresponding CSI compression parameter is 1 / 32), sub-encoder 5 (the corresponding CSI compression parameter is 1 / 64), etc.

[0137] As shown in FIG4 , in some other embodiments, the CSI compression model may include a channel encoder 411 and a feature converter (also referred to as a feature optimizer or a feature joint optimizer) 412. The input of the feature converter 412 may be the above-mentioned first channel matrix, which is used to extract key features from the first channel matrix to obtain a second channel matrix representing the key features of the CSI. The second channel matrix may be used as the input of the channel encoder 411, and the channel encoder 411 may compress the second channel matrix according to the CSI compression parameter to obtain a target channel matrix. For example, the terminal device may use the subencoder corresponding to the CSI compression parameter as the second target subencoder, and then compress the second channel matrix through the second target subencoder to obtain a target channel matrix. The target channel matrix may also be referred to as a codeword.

[0138] FIG5 is a schematic diagram of a feature converter in a CSI compression model according to an exemplary embodiment. As shown in FIG5 , the feature converter 412 includes a feature extraction network 4121 , an attention mechanism network 4122 , and a feature restoration network 4123 .

[0139] In some embodiments, the terminal device may input the first channel matrix into a feature converter, perform key feature extraction on the first channel matrix, and obtain a second channel matrix representing the CSI key features.

[0140] For example, the second channel matrix may be obtained by the following steps:

[0141] S51. Input the first channel matrix into a feature extraction network to obtain a plurality of first feature maps.

[0142] For example, the feature extraction network converts the first channel matrix H c Convert to the first feature map Where f represents the number of first feature maps extracted. f can be any positive integer greater than 1.

[0143] In some embodiments, the feature extraction network may include a two-dimensional convolutional layer, and the size of the convolution kernel may be f×m×m, where f represents the number of first feature maps and m×m represents the length and width of the convolution window used by the convolution kernel. The feature extraction network may use a two-dimensional normalization layer to normalize the output of the convolution layer, and the activation function of the feature extraction network may include Sigmoid, ReLU, LeakyReLU, PReLU, or ELU.

[0144] For example, the activation function may adopt a LeakyReLU (leaky linear rectifier function) activation function, which may include the following formula (2):

[0145]

[0146] Wherein, x represents an input vector, such as the first feature map of the input; γ represents a preset coefficient. For example, the preset coefficient can be any value less than 1, for example, the preset coefficient can be 0.3; LeakyReLU(x) represents the output value of the Leaky ReLU activation function.

[0147] S52. Input the multiple first feature maps into the attention mechanism network to obtain a second feature map.

[0148] The second feature map may include key feature information in the first feature map.

[0149] In some embodiments, a maximum pooling operation can be performed on multiple first feature maps through an attention mechanism network to obtain a maximum pooling feature map; an average pooling operation can be performed on multiple first feature maps through an attention mechanism network to obtain an average pooling feature map; then, the second feature map is determined based on the maximum pooling feature map and the average pooling feature map.

[0150] It should be noted that among the multiple first feature maps extracted by the feature extraction network, some contain feature information that is extremely helpful for CSI reconstruction and can be called "key feature maps." However, some first feature maps contain feature information that has little effect on CSI reconstruction and can be called "non-essential feature maps." The attention mechanism network can extract key feature maps from multiple feature maps, thereby allowing the codewords generated by the subsequent encoder to contain more key features. For example:

[0151] The attention mechanism network can perform a maximum pooling operation on the first feature map F to obtain the maximum pooling feature map M∈ For example, the maximum pooling operation can be performed using the following formula (3):

[0152] m i =max{F 1,i ,F 2,i ,F 3,i ,…,F f,i},i={1,2,…,N c N t} (3)

[0153] Among them, m i represents the i-th element in the maximum pooling feature map M, f represents the number of the first feature map, F 1,i represents the i-th element in the first feature map, F 2,i represents the i-th element in the second first feature map, F 2,i Represents the i-th element in the second first feature map, F f,i represents the i-th element in the f-th first feature map, N c ×N t Indicates the size of the first channel matrix.

[0154] In this way, each element in the maximum pooling feature map M is composed of the maximum elements at the corresponding positions in multiple first feature maps.

[0155] The attention mechanism network can also perform an average pooling operation on the first feature map F to obtain an average pooling feature map For example, the average pooling operation can be performed using the following formula (4):

[0156]

[0157] Among them, v i represents the i-th element in the average pooling feature map V, f represents the number of the first feature map, F 1,i represents the i-th element in the first feature map, F 2,irepresents the i-th element in the second first feature map, F 2,i Represents the i-th element in the second first feature map, F f-1,i represents the i-th element in the f-1th first feature map, F f,i represents the i-th element in the f-th first feature map, N c ×N t Indicates the size of the first channel matrix.

[0158] In this way, each element in the average pooling feature map V is composed of the mean of all elements at the corresponding positions of multiple first feature maps.

[0159] Furthermore, the second feature map can be determined according to the maximum pooling feature map and the average pooling feature map.

[0160] In some embodiments, the maximum pooling feature map and the average pooling feature map can be input into the fusion sub-network to obtain a fused fusion feature map; and the second feature map can be calculated based on the fusion feature map and the first feature map.

[0161] In some embodiments, the maximum pooling feature map M and the average pooling feature map V can be spliced ​​to obtain a joint feature map The joint feature map C is passed through the fusion network to obtain the fusion feature map

[0162] In some embodiments, the fusion network may adopt a two-dimensional convolution layer with a convolution kernel size of m×n×n. The fusion network may also include a two-dimensional normalization layer and an activation function, and the activation function may include a Sigmoid activation function.

[0163] Then, multiply the fusion feature map D with the first feature matrix F to obtain the second feature map The second feature map F′ may also be referred to as an optimized feature map.

[0164] In this way, the feature information in the key feature map is highlighted in the second feature map F′, while the feature information in the unnecessary feature map is weakened, so that the key features of CSI can be represented.

[0165] S53: Input the second feature map into a feature restoration network to obtain a second channel matrix.

[0166] For example, the second feature map F′ can be restored to the second channel matrix by the feature restoration network

[0167] In some embodiments, the feature restoration network may include a two-dimensional convolutional layer, a two-dimensional normalization layer and an activation function. The activation function may be a LeakyReLU activation function or other activation functions in related technologies, which is not limited in this disclosure.

[0168] In some embodiments, the feature restoration network and the feature extraction network may use the same activation function to avoid feature distortion.

[0169] Thus, the second channel matrix H obtained in this way is e The key feature information is highlighted and the unnecessary feature information is weakened, so that the key features of CSI can be characterized.

[0170] Figure 6 is a schematic diagram of a channel encoder in a CSI compression model according to an exemplary embodiment. As shown in Figure 6, the channel encoder 411 may include multiple sub-encoders, such as sub-encoder 1, sub-encoder 2, ..., sub-encoder T-1, and sub-encoder T in the figure, where T is the number of sub-encoders.

[0171] In some embodiments, the channel encoder may first convert the input second channel matrix H e Preprocessing is performed, for example, dimension transformation can be performed, and the dimension size of the transformed second channel matrix can be The second channel matrix after dimension transformation is input into the above sub-encoder for processing.

[0172] In some embodiments, the CSI compression parameters may include a plurality of preset CSI compression rates, and the set of preset CSI compression rates may include σ={σ1, σ2, . . . , σ T}, sort the preset CSI compression rates in the set in descending order from large to small. For each compression rate σ in the set j , j = {1, 2, ..., T}, the channel encoder has a sub-encoder corresponding to the preset CSI compression rate, that is, the number of sub-encoders is equal to the number of preset CSI compression rates.

[0173] For example, the preset CSI compression rate σ1 corresponding to subcoder 1 may be 1 / 2, the preset CSI compression rate σ2 corresponding to subcoder 2 may be 1 / 4, the preset CSI compression rate σ3 corresponding to subcoder 3 may be 1 / 8, the preset CSI compression rate σ4 corresponding to subcoder 4 may be 1 / 16, the preset CSI compression rate σ5 corresponding to subcoder 5 may be 1 / 32, the preset CSI compression rate σ6 corresponding to subcoder 6 may be 1 / 64, and so on. It should be noted that the values ​​of the preset CSI compression rates here are all examples, and this disclosure does not limit the specific values.

[0174] In some embodiments, the set of preset CSI compression rates may include σ = {1 / 4, 1 / 16, 1 / 32, 1 / 64}, i.e., the number of sub-encoders T = 4, and the preset CSI compression rates in the set are arranged in descending order. For each preset CSI compression rate in the set, there is a corresponding sub-encoder in the adaptive encoder that matches it, i.e., the number of sub-encoders is equal to the number of dynamic compression rates. For example, the preset CSI compression rate σ1 corresponding to sub-encoder 1 may be 1 / 4, the preset CSI compression rate σ2 corresponding to sub-encoder 2 may be 1 / 16, the preset CSI compression rate σ3 corresponding to sub-encoder 3 may be 1 / 32, and the preset CSI compression rate σ4 corresponding to sub-encoder 4 may be 1 / 64.

[0175] In some embodiments, each sub-encoder may include a compression layer, which may include a fully connected layer. The size of the compression layer of each sub-encoder may be determined according to the CSI compression parameter (e.g., a preset CSI compression rate) and the size of the first channel matrix. For example, the size of the first channel matrix is ​​c×N c ×N t , the preset CSI compression rate is σ1, then the size of the compression layer of the sub-encoder corresponding to σ1 is (c×N c ×N t )×d1, where d1 can be c×N c ×N t ×σ1.

[0176] For example, the size of the first channel matrix is ​​2×32×32, and the preset CSI compression rate is 1 / 4, then the size of the compression layer of the sub-encoder corresponding to the preset CSI compression rate may be 2048×512.

[0177] In other embodiments, each sub-encoder may include a compression layer and a coding switch. In this way, in actual use, the target sub-encoder (for example, the first target sub-encoder or the second target sub-encoder in the above embodiment) corresponding to the CSI compression parameter (for example, the CSI compression rate) can be determined, and the coding switch of the target sub-encoder is closed, and the switches of other sub-encoders are disconnected, so that the compression layer of the target sub-encoder can be used to compress the input second channel matrix, and the undetermined channel matrix of the target sub-encoder is output, and the undetermined channel matrix can be used as the target channel matrix M.

[0178] In some embodiments, the preset CSI compression rate with the largest value is used as the maximum CSI compression rate, and the compression layer of the sub-encoder corresponding to the maximum CSI compression rate (for example, sub-encoder 1 in Figure 5) is used as the maximum compression layer. The output of the maximum compression layer can be used as the input of the compression layer of other sub-encoders. In this way, the efficiency of the compression operation can be improved.

[0179] For example, the preset CSI compression rate σ1 is the maximum compression rate, and the size of the compression layer 1 of the sub-coder 1 corresponding to σ1 can be (c×N c ×N t )×d1, where d1 can be c×N c ×N t The output of compression layer 1 can be used as the input of the compression layer corresponding to other sub-encoders. The size of the compression layer corresponding to other sub-encoders can be d1×d k , d k It can be c×N c ×N t ×σ k Where k represents the number of sub-encoders, σ k Indicates the preset CSI compression rate corresponding to the kth sub-encoder, c×N c ×N t Indicates the size of the first channel matrix.

[0180] For example, if the size of the first channel matrix is ​​2×32×32 and the maximum compression rate σ1 is 1 / 2, the size of the compression layer 1 of the subcoder 1 corresponding to σ1 may be 2048×1024; if the preset CSI compression rate σ2 corresponding to the subcoder 2 is 1 / 4, the size of the compression layer 2 of the subcoder 2 may be 1024×512; if the preset CSI compression rate σ3 corresponding to the subcoder 3 is 1 / 16, the size of the compression layer 3 of the subcoder 3 may be 1024×256.

[0181] In this way, the compression layers in sub-encoder 2 to sub-encoder T in FIG6 can further reduce the dimensionality.

[0182] FIG7 is a schematic diagram showing a CSI decompression model according to an exemplary embodiment. As shown in FIG7 , the CSI decompression model 42 may include a channel decoder 421 .

[0183] In some embodiments, the channel decoder 421 may include multiple sub-decoders, with different sub-decoders corresponding to different CSI compression parameters. Each sub-decoder may decompress the received target channel matrix through a decompression layer to obtain a third channel matrix. The decompression layer may include a fully connected layer.

[0184] As shown in FIG7 , the channel decoder 421 may include multiple sub-decoders, for example, sub-decoder 1, sub-decoder 2, ..., sub-decoder T-1, sub-decoder T, where T is the number of sub-decoders.

[0185] In some embodiments, the CSI compression parameters may include a plurality of preset CSI compression rates, and the set of preset CSI compression rates may include σ={σ1, σ2, . . . , σT}, sort the preset CSI compression rates in the set in descending order from large to small. For each compression rate σ in the set j , j = {1, 2, ..., T}, each channel decoder has a sub-decoder corresponding to the preset CSI compression rate, that is, the number of sub-decoders is equal to the number of preset CSI compression rates.

[0186] For example, the preset CSI compression rate σ1 corresponding to sub-decoder 1 may be 1 / 2, the preset CSI compression rate σ2 corresponding to sub-decoder 2 may be 1 / 4, the preset CSI compression rate σ3 corresponding to sub-decoder 3 may be 1 / 8, the preset CSI compression rate σ4 corresponding to sub-decoder 4 may be 1 / 16, the preset CSI compression rate σ5 corresponding to sub-decoder 5 may be 1 / 32, the preset CSI compression rate σ6 corresponding to sub-decoder 6 may be 1 / 64, and so on. It should be noted that the values ​​of the preset CSI compression rates here are all examples, and this disclosure does not limit the specific values.

[0187] In some embodiments, each sub-decoder may include a decompression layer and a decoding switch. In this way, in actual use, a target sub-decoder corresponding to a CSI compression parameter (e.g., a CSI compression rate) can be determined, and the decoding switch of the target sub-decoder is closed, while the switches of other sub-decoders are opened. Thus, the decompression layer of the target sub-decoder can be used to decompress the input target channel matrix, and the pending channel matrix corresponding to the target sub-decoder is output. The pending channel matrix can be used as a fourth channel matrix; and the third channel matrix can be further determined based on the fourth channel matrix.

[0188] In some embodiments, the CSI decompression model may further include a CSI reconstruction module 422. The CSI reconstruction module may include a CNN (Convolutional Neural Networks). For example, the CSI reconstruction module includes two CNN networks, each CNN network includes 5 convolutional layers, and the convolution kernel sizes of each convolutional layer are c×k×k, f1×l×l, f2×l×l, f2×n×n, c×m×m, (f1, f2, k, l, m, n are all preset values, and different values ​​can be preset according to different convolutional layers). The step size of each convolution layer is t, and a normalization layer and a LeakyReLU activation function can be used. Then, the output element value of the second CNN module is mapped to the [0,1] interval through a Sigmoid activation function layer. In this way, the CSI reconstruction module can output a third channel matrix corresponding to the target channel matrix

[0189] It should be noted that the CSI reconstruction module can output different third channel matrices for different CSI compression parameters (e.g., CSI compression rates). For example, the set of preset CSI compression rates in the CSI compression parameters may include σ={σ1,σ2,…,σ T}, then the corresponding output third channel matrix can be

[0190] In this way, through the CSI decompression model, the network device can decompress the received target channel matrix to obtain a third channel matrix, so as to determine the channel quality of the downlink channel according to the third channel matrix.

[0191] In some embodiments, the above-mentioned CSI compression model and CSI decompression model can be obtained through offline training. For example, the above-mentioned CSI compression model and CSI decompression model can be jointly trained to obtain parameters of the CSI compression model and the CSI decompression model so that the CSI compression model and the CSI decompression model can match.

[0192] The training of the above model can be performed on a terminal device or on a network device, which will be explained below with reference to the accompanying drawings.

[0193] FIG8 is a flow chart of a method for training a CSI compression model according to an exemplary embodiment. The training method can be applied to a terminal device. As shown in FIG8 , the training method may include:

[0194] S801: The terminal device obtains a first sample channel matrix for training.

[0195] The first sample channel matrix is ​​a matrix obtained by the terminal device according to the received pilot signal and used to characterize the downlink channel quality.

[0196] In some embodiments, in an FDD downlink mMIMO system, N MIMO arrays may be configured at half-wavelength intervals in a ULA (Uniform Linear Array) manner on the network device (e.g., base station) side. t = 32 antennas, with a single antenna configured on the terminal device. Using the COST2100 channel model, 150,000 spatial CSI matrix samples were generated in a 5.3 GHz indoor microcell scenario. A training set of 100,000 samples, a validation set of 30,000 samples, and a test set of 20,000 samples were obtained. The training set can be used as the first sample channel matrix described above.

[0197] S802. The terminal device trains the first target network model according to the first sample channel matrix to obtain a CSI compression model.

[0198] Among them, the first target network model includes a first target compression model and a first target decompression model. The network structure of the first target compression model is the same as the network structure of the CSI compression model. For example, both can include a channel encoder, and the channel encoder includes multiple sub-encoders. Different sub-encoders correspond to different CSI compression parameters; the first target decompression model includes a channel decoder, and the channel decoder includes multiple sub-decoders. Different sub-decoders correspond to different CSI compression parameters.

[0199] In some embodiments, the model structure of the above-mentioned first target compression model can be the same as the CSI compression model shown in Figure 4. For example, the first target compression model may include a channel encoder and a feature converter. The structure of the feature converter can be as shown in Figure 5, and the structure of the channel encoder can be as shown in Figure 6; the model structure of the above-mentioned first target decompression model can be the same as the CSI decompression model shown in Figure 7, and the above-mentioned model structures will not be repeated here.

[0200] In some embodiments, the first model training step can be executed in a loop until the trained first target network model is determined to meet the first preset stop iteration condition based on the first sample channel matrix and the first predicted channel matrix, and the first target compression model in the trained first target network model is used as the CSI compression model.

[0201] The first predicted channel matrix is ​​a matrix output after the first sample channel matrix is ​​input into the first target network model.

[0202] For example, the first compression model parameters corresponding to the first target compression model may be used as model parameters of the CSI compression model.

[0203] The first model training step may include:

[0204] S81. Input a first sample channel matrix into a first target compression model, and compress the first sample channel matrix through multiple sub-encoders to obtain a first target sample channel matrix.

[0205] S82: Input the first target sample channel matrix into a first target decompression model, and compress the first sample channel matrix through multiple sub-decoders to obtain a first predicted channel matrix.

[0206] S83. When it is determined that the first target network model does not meet the first preset stop iteration condition based on the first sample channel matrix and the first predicted channel matrix, a first loss value is determined based on the first sample channel matrix and the first predicted channel matrix, and the parameters of the first target network model are updated based on the first loss value to obtain the trained first target network model, and the trained first target network model is used as the new first target network model.

[0207] It should be noted that the above-mentioned first preset stopping iteration condition can be determined according to the loss function used in the training process.

[0208] In some embodiments, the encoding switches corresponding to all sub-encoders may be closed, and the decoding switches corresponding to all sub-decoders may also be closed. The training errors under multiple CSI compression parameters may be jointly optimized. The loss function used in the training process may include the following formula (5):

[0209]

[0210] Wherein, σ1 represents the first preset CSI compression rate, σ2 represents the second preset CSI compression rate, σ T represents the Tth preset CSI compression rate, T represents the number of preset CSI compression rates in the CSI compression parameter (it can also represent the number of sub-encoders or sub-decoders), represents the loss value corresponding to the preset CSI compression rate σ1, and Loss represents the first loss value of the above-mentioned first sample channel matrix and the first predicted channel matrix.

[0211] By jointly optimizing the training errors corresponding to multiple CSI compression parameters (e.g., preset CSI compression rates), the network can optimize the training errors corresponding to all preset CSI compression rates during a single training process, thereby improving the network's adaptability to dynamic changes in CSI compression rates. After training, the parameters of the CSI compression model can be obtained.

[0212] In some embodiments, the terminal device can also obtain the trained parameters of the first target decompression model through the above training process, and the trained parameters of the first target decompression model can be used as parameters of the CSI decompression model on the network device side.

[0213] In some embodiments, after determining the model parameters through the training method, the terminal device may obtain first decompression model parameters corresponding to the first target decompression model in the trained first target network model, and send the first decompression model parameters to the network device to instruct the network device to determine a CSI decompression model based on the first decompression model parameters. The CSI decompression model is used by the network device to determine the channel quality of the downlink channel based on the target channel matrix.

[0214] For example, the terminal device may send the first decompression model parameters to the network device via signaling or data packets.

[0215] In some other embodiments of the present disclosure, the above-mentioned training method can be executed on a network device, and the terminal device can receive the second compression model parameters sent by the network device; and determine the CSI compression model based on the second compression model parameters.

[0216] For example, the second compression model parameters may be used as model parameters of the CSI compression model.

[0217] For example, the terminal device may receive the second compression model parameters sent by the network device through signaling or data packets, and determine the parameters corresponding to the CSI compression model based on the second compression model parameters.

[0218] Figure 9 is a flow chart of a method for training a CSI decompression model according to an exemplary embodiment, which can be applied to a network device. As shown in Figure 9, the training method may include:

[0219] S901: A network device obtains a second sample channel matrix for training.

[0220] The second sample channel matrix may be a matrix obtained by the terminal device according to the received pilot signal and used to characterize the downlink channel quality.

[0221] In some embodiments, in an FDD downlink mMIMO system, N can be configured at half-wavelength intervals in a ULA manner on the network device (eg, base station). t = 32 antennas, with a single antenna configured on the terminal device. Using the COST2100 channel model, 150,000 spatial CSI matrix samples were generated in a 5.3 GHz indoor microcell scenario. These samples were divided into a training set of 100,000 samples, a validation set of 30,000 samples, and a test set of 20,000 samples. The training set can be used as the first sample channel matrix.

[0222] S902: The network device trains a second target network model according to the second sample channel matrix to obtain a CSI decompression model.

[0223] Among them, the second target network model includes a second target compression model and a second target decompression model. The network structure of the second target decompression model is the same as the network structure of the CSI decompression model. For example, both include a channel decoder, which may include multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters; the second target compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders, and different sub-encoders correspond to different CSI compression parameters.

[0224] In some embodiments, the model structure of the above-mentioned second target compression model can be the same as the CSI compression model shown in Figure 4. For example, the second target compression model may include a channel encoder and a feature converter. The structure of the feature converter can be as shown in Figure 5, and the structure of the channel encoder can be as shown in Figure 6; the model structure of the above-mentioned second target decompression model can be the same as the CSI decompression model shown in Figure 7, and the above-mentioned model structures will not be repeated here.

[0225] In some embodiments, the second model training step can be executed in a loop until the trained second target network model is determined to meet the second preset stop iteration condition based on the second sample channel matrix and the second predicted channel matrix, and the second target decompression model in the trained second target network model is used as the CSI decompression model; the second predicted channel matrix is ​​the matrix output after the second sample channel matrix is ​​input into the second target network model.

[0226] The second model training step may include:

[0227] S91, inputting the second sample channel matrix into a second target compression model, and compressing the second sample channel matrix through multiple sub-encoders to obtain a second target sample channel matrix;

[0228] S92, inputting the second target sample channel matrix into a second target decompression model, and decompressing the second target sample channel matrix through a plurality of sub-decoders to obtain a second predicted channel matrix;

[0229] S93. When it is determined that the second target network model does not meet the second preset stop iteration condition according to the second sample channel matrix and the second predicted channel matrix, a second loss value is determined according to the second sample channel matrix and the second predicted channel matrix, and the parameters of the second target network model are updated according to the second loss value to obtain a trained second target network model, and the trained second target network model is used as a new second target network model.

[0230] It should be noted that the above second preset stopping iteration condition can also be determined according to the loss function used in the training process.

[0231] In some embodiments, the encoding switches corresponding to all sub-encoders may be closed, and the decoding switches corresponding to all sub-decoders may also be closed. The training errors under multiple CSI compression parameters may be jointly optimized. The loss function used in the training process may also include the above formula (5), which will not be repeated here.

[0232] By jointly optimizing the training errors corresponding to multiple CSI compression parameters (e.g., preset CSI compression rates), the network can optimize the training errors corresponding to all preset CSI compression rates during a single training session, thereby improving the network's adaptability to dynamic changes in CSI compression rates. After training, the parameters of the CSI decompression model can be obtained.

[0233] In some embodiments, the network device can also obtain the trained parameters of the second target compression model through the above training process, and the trained parameters of the second target compression model can be used as the parameters of the CSI compression model on the terminal device side.

[0234] In some embodiments, the network device may obtain second compression model parameters corresponding to a second target compression model in a trained second target network model, and transmit the second compression model parameters to a terminal device to instruct the terminal device to determine a CSI compression model based on the second compression model parameters. The CSI compression model is used by the terminal device to obtain a target channel matrix based on the first channel matrix.

[0235] For example, the network device may send the second compression model parameters to the terminal device via signaling or data packets.

[0236] In some other embodiments of the present disclosure, the above-mentioned training method can be executed in a terminal device, and the network device can receive the first decompression model parameters sent by the terminal device; and determine the CSI decompression model based on the first decompression model parameters.

[0237] For example, the network device may receive the first decompression model parameters sent by the terminal device through signaling or data packets, and determine the parameters corresponding to the CSI decompression model based on the first decompression model parameters.

[0238] FIG10 shows a method for obtaining channel quality according to an exemplary embodiment. As shown in FIG10 , the method may include:

[0239] S1001. A network device sends a pilot signal via a downlink channel.

[0240] S1002. The terminal device receives a pilot signal through a downlink channel and obtains a first channel matrix according to the pilot signal.

[0241] The first channel matrix is ​​used to characterize the channel quality of the downlink channel;

[0242] S1003. The terminal device compresses the first channel matrix according to the channel state information CSI compression model and the channel state information CSI compression parameters to obtain a compressed target channel matrix.

[0243] In some embodiments, the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters.

[0244] In some embodiments, the CSI compression model may include a channel encoder and a feature converter.

[0245] S1004: The terminal device sends the target channel matrix to the network device.

[0246] S1005. The network device receives the target channel matrix sent by the terminal device, and decompresses the target channel matrix according to the channel state information CSI decompression model and CSI compression parameters to obtain a third channel matrix.

[0247] The CSI decompression model includes a channel decoder, which includes multiple sub-decoders. Different sub-decoders correspond to different CSI compression parameters.

[0248] S1006: The network device determines the channel quality of the downlink channel according to the third channel matrix.

[0249] It should be noted that the above-mentioned CSI compression model and the above-mentioned CSI decompression model can be any model structure provided in the above-mentioned embodiments, which will not be repeated here.

[0250] In this way, different CSI compression parameters can be adapted through multiple sub-encoders and multiple sub-decoders, so that in scenarios where the CSI compression rate changes, the terminal device can adaptively obtain a more accurate target channel matrix and send it to the network device, so that the network device can obtain more accurate channel quality, thereby improving data transmission efficiency.

[0251] FIG11 is a block diagram of an apparatus 1100 for obtaining channel quality according to an exemplary embodiment, which can be applied to a terminal device. As shown in FIG11 , the apparatus 1100 may include:

[0252] The first receiving module 1101 is configured to receive a pilot signal sent by a network device through a downlink channel;

[0253] A first matrix acquisition module 1102 is configured to acquire a first channel matrix according to the pilot signal; the first channel matrix is ​​used to represent the channel quality of the downlink channel;

[0254] The target matrix acquisition module 1103 is configured to compress the first channel matrix according to a CSI compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; wherein the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters;

[0255] The first sending module 1104 is configured to send the target channel matrix to the network device, so that the network device determines the channel quality of the downlink channel according to the target channel matrix.

[0256] In some embodiments, the target matrix acquisition module 1103 is configured to use the subcoder corresponding to the CSI compression parameter as a first target subcoder; and compress the first channel matrix through the first target subcoder to obtain a target channel matrix.

[0257] In some embodiments, the CSI compression model further includes a feature converter; the target matrix acquisition module 1103 is configured to input the first channel matrix into the feature converter, extract key features of the first channel matrix, and obtain a second channel matrix representing CSI key features; and compress the second channel matrix according to the CSI compression parameters and the channel encoder to obtain the target channel matrix.

[0258] In some embodiments, the target matrix acquisition module 1103 is configured to use the subcoder corresponding to the CSI compression parameter as a second target subcoder; and compress the second channel matrix through the second target subcoder to obtain a target channel matrix.

[0259] In some embodiments, the feature converter includes a feature extraction network, an attention mechanism network and a feature restoration network; the target matrix acquisition module 1103 is configured to input the first channel matrix into the feature extraction network to obtain multiple first feature maps; input multiple first feature maps into the attention mechanism network to obtain a second feature map; the second feature map includes key feature information in the first feature map; and input the second feature map into the feature restoration network to obtain a second channel matrix.

[0260] In some embodiments, the target matrix acquisition module 1103 is configured to perform a maximum pooling operation on the plurality of first feature maps through the attention mechanism network to obtain a maximum pooling feature map; perform an average pooling operation on the plurality of first feature maps through the attention mechanism network to obtain an average pooling feature map; and determine a second feature map based on the maximum pooling feature map and the average pooling feature map.

[0261] In some embodiments, the attention mechanism network includes a fusion sub-network; the target matrix acquisition module 1103 is configured to input the maximum pooling feature map and the average pooling feature map into the fusion sub-network to obtain a fused fusion feature map; and a second feature map is calculated based on the fusion feature map and the first feature map.

[0262] In some embodiments, the first matrix acquisition module 1102 is configured to measure and obtain a spatial channel matrix based on the pilot signal; transform the spatial channel matrix into an angle delay domain channel matrix through discrete Fourier transform; and determine the first channel matrix based on the angle delay domain channel matrix.

[0263] FIG12 is a block diagram of an apparatus 1100 for obtaining channel quality according to an exemplary embodiment. As shown in FIG12 , the apparatus 1100 may further include a first training module 1105. The first training module 1105 is configured to train the CSI compression model obtained by:

[0264] Obtaining a first sample channel matrix for training; the first sample channel matrix is ​​a matrix obtained by the terminal device based on the received pilot signal and used to characterize the downlink channel quality;

[0265] Training a first target network model according to the first sample channel matrix to obtain the CSI compression model;

[0266] Among them, the first target network model includes a first target compression model and a first target decompression model, the network structure of the first target compression model is the same as the network structure of the CSI compression model, the first target decompression model includes a channel decoder, and the channel decoder includes multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters.

[0267] In some embodiments, the first training module 1105 is configured to loop through the first model training step until it is determined that the trained first target network model satisfies a first preset stop iteration condition based on the first sample channel matrix and the first predicted channel matrix, and the first target compression model in the trained first target network model is used as the CSI compression model; the first predicted channel matrix is ​​a matrix output after the first sample channel matrix is ​​input into the first target network model;

[0268] The first model training step includes:

[0269] Inputting the first sample channel matrix into the first target compression model, and compressing the first sample channel matrix through a plurality of sub-encoders to obtain a first target sample channel matrix;

[0270] Inputting the first target sample channel matrix into the first target decompression model, and decompressing the first target sample channel matrix through a plurality of sub-decoders to obtain a first predicted channel matrix;

[0271] When it is determined according to the first sample channel matrix and the first predicted channel matrix that the first target network model does not meet the first preset stop iteration condition, a first loss value is determined according to the first sample channel matrix and the first predicted channel matrix, and the parameters of the first target network model are updated according to the first loss value to obtain the trained first target network model, and the trained first target network model is used as the new first target network model.

[0272] In some embodiments, the first sending module 1104 is further configured to obtain first decompression model parameters corresponding to the first target decompression model in the trained first target network model; and send the first decompression model parameters to the network device to instruct the network device to determine the CSI decompression model based on the first decompression model parameters, and the CSI decompression model is used by the network device to determine the channel quality of the downlink channel based on the target channel matrix.

[0273] In some embodiments, the first receiving module 1101 is further configured to receive second compression model parameters sent by the network device; and determine the CSI compression model according to the second compression model parameters.

[0274] In some embodiments, the first receiving module 1101 is further configured to receive a first compression parameter sent by a network device; and determine the CSI compression parameter according to the first compression parameter.

[0275] FIG13 is a block diagram of an apparatus 1300 for obtaining channel quality according to an exemplary embodiment, which can be applied to a network device. As shown in FIG13 , the apparatus 1300 may include:

[0276] The second receiving module 1301 is configured to receive a target channel matrix sent by a terminal device; the target channel matrix is ​​obtained by the terminal device by compressing a first channel matrix according to a CSI compression model and a channel state information (CSI) compression parameter, where the first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal and used to characterize the channel quality of a downlink channel;

[0277] The third matrix acquisition module 1302 is configured to decompress the target channel matrix according to a CSI decompression model and the CSI compression parameter to obtain a third channel matrix; wherein the CSI decompression model includes a channel decoder, and the channel decoder includes multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters;

[0278] The channel quality determination module 1303 is configured to determine the channel quality of the downlink channel according to the third channel matrix.

[0279] In some embodiments, the third matrix acquisition module 1302 is configured to use the sub-decoder corresponding to the CSI compression parameter as a target sub-decoder; decompress the target channel matrix through the target sub-decoder to obtain a fourth channel matrix; and determine the third channel matrix based on the fourth channel matrix.

[0280] In some embodiments, the CSI decompression model further includes a CSI reconstruction module; the third matrix acquisition module 1302 is configured to input the fourth channel matrix into the CSI reconstruction module to obtain the third channel matrix.

[0281] FIG14 is a block diagram of an apparatus 1300 for obtaining channel quality according to an exemplary embodiment. As shown in FIG14 , the apparatus 1300 may further include a second training module 1304. The second training module 1304 is configured to train the CSI decompression model obtained by:

[0282] Obtaining a second sample channel matrix for training; the second sample channel matrix is ​​a matrix obtained by the terminal device based on the received pilot signal and used to characterize the downlink channel quality;

[0283] Training a second target network model according to the second sample channel matrix to obtain the CSI compression model;

[0284] Among them, the second target network model includes a second target compression model and a second target decompression model, the network structure of the second target decompression model is the same as the network structure of the CSI decompression model, the second target compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders, and different sub-encoders correspond to different CSI compression parameters.

[0285] In some embodiments, the second training module 1304 is configured to loop through the second model training step until it is determined that the trained second target network model satisfies a second preset stop iteration condition based on the second sample channel matrix and the second predicted channel matrix, and the second target decompression model in the trained second target network model is used as the CSI decompression model; the second predicted channel matrix is ​​a matrix output after the second sample channel matrix is ​​input into the second target network model;

[0286] The second model training step includes:

[0287] Inputting the second sample channel matrix into the second target compression model, and compressing the second sample channel matrix through a plurality of sub-encoders to obtain a second target sample channel matrix;

[0288] Inputting the second target sample channel matrix into the second target decompression model, and decompressing the second target sample channel matrix through a plurality of sub-decoders to obtain a second predicted channel matrix;

[0289] When it is determined according to the second sample channel matrix and the second predicted channel matrix that the second target network model does not meet the second preset stop iteration condition, a second loss value is determined according to the second sample channel matrix and the second predicted channel matrix, and the parameters of the second target network model are updated according to the second loss value to obtain a trained second target network model, and the trained second target network model is used as a new second target network model.

[0290] FIG15 is a block diagram showing an apparatus 1300 for obtaining channel quality according to an exemplary embodiment. As shown in FIG15 , the apparatus 1300 may further include:

[0291] The second sending module 1305 is configured to obtain second compression model parameters corresponding to the second target compression model in the trained second target network model; and send the second compression model parameters to the terminal device to instruct the terminal device to determine the CSI compression model based on the second compression model parameters, and the CSI compression model is used by the terminal device to obtain the target channel matrix based on the first channel matrix.

[0292] In some embodiments, the second receiving module 1301 is further configured to receive first decompression model parameters sent by the terminal device; and determine a CSI decompression model based on the first decompression model parameters.

[0293] In some embodiments, the second sending module 1305 is configured to determine a first compression parameter based on the CSI compression parameter; and send the first compression parameter to the terminal device.

[0294] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0295] Fig. 16 is a block diagram showing an apparatus for obtaining channel quality according to an exemplary embodiment. The apparatus 2000 for obtaining channel quality may be a terminal device in the communication system shown in Fig. 1 or a network device in the communication system.

[0296] 16 , the apparatus 2000 may include one or more of the following components: a processing component 2002 , a memory 2004 , and a communication component 2006 .

[0297] The processing component 2002 can control the overall operation of the device 2000, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 2002 can include one or more processors 2020 to execute instructions to complete all or part of the steps of the above-mentioned method for obtaining channel quality. In addition, the processing component 2002 can include one or more modules to facilitate interaction between the processing component 2002 and other components. For example, the processing component 2002 can include a multimedia module to facilitate interaction between the multimedia component and the processing component 2002.

[0298] The memory 2004 is configured to store various types of data to support the operations of the device 2000. Examples of such data include instructions for any application or method operating on the device 2000, contact data, phone book data, messages, pictures, videos, etc. The memory 2004 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0299] The communication component 2006 is configured to facilitate wired or wireless communication between the device 2000 and other devices. The device 2000 can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G, 5G or 6G communication technologies, or a combination thereof. In an exemplary embodiment, the communication component 2006 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 2006 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0300] In an exemplary embodiment, the apparatus 2000 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-mentioned method for obtaining channel quality.

[0301] The above-mentioned device 2000 can be an independent electronic device or a part of an independent electronic device. For example, in one embodiment, the electronic device can be an integrated circuit (IC) or a chip, wherein the integrated circuit can be a single IC or a collection of multiple ICs; the chip can include but is not limited to the following types: GPU (Graphics Processing Unit), CPU (Central Processing Unit), FPGA (Field Programmable Gate Array), DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), SOC (System on Chip, SoC), etc. The above-mentioned integrated circuit or chip can be used to execute executable instructions (or codes) to implement the above-mentioned method for obtaining channel quality. The executable instructions can be stored in the integrated circuit or chip, or obtained from other devices or equipment. For example, the integrated circuit or chip may include a processor, a memory, and an interface for communicating with other devices. The executable instruction may be stored in the processor, and when the executable instruction is executed by the processor, the above-mentioned method for obtaining channel quality is implemented; alternatively, the integrated circuit or chip may receive the executable instruction through the interface and transmit it to the processor for execution, so as to implement the above-mentioned method for obtaining channel quality.

[0302] In an exemplary embodiment, the present disclosure further provides a computer-readable storage medium having computer program instructions stored thereon. When the program instructions are executed by a processor, the steps of the method for obtaining channel quality provided by the present disclosure are implemented. For example, the computer-readable storage medium may be a non-transitory computer-readable storage medium including instructions, for example, the aforementioned memory 2004 including instructions. The aforementioned instructions may be executed by the processor 2020 of the apparatus 2000 to perform the aforementioned method for obtaining channel quality. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

[0303] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program executable by a programmable device, and has a code portion for executing the above method for obtaining channel quality when executed by the programmable device.

[0304] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the present disclosure. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0305] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for obtaining channel quality, characterized in that: Applied to a terminal device, the method includes: receiving a pilot signal sent by a network device via a downlink channel; Acquire a first channel matrix according to the pilot signal; the first channel matrix is ​​used to characterize the channel quality of the downlink channel; compressing the first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; wherein the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters; The target channel matrix is ​​sent to the network device, so that the network device determines the channel quality of the downlink channel according to the target channel matrix.

2. The method according to claim 1, characterized in that The compressing the first channel matrix according to the channel state information CSI compression model and the channel state information CSI compression parameter to obtain a target channel matrix includes: Using the sub-encoder corresponding to the CSI compression parameter as the first target sub-encoder; The first channel matrix is ​​compressed by the first target sub-encoder to obtain a target channel matrix.

3. The method according to claim 1, characterized in that The CSI compression model also includes a feature converter; The compressing the first channel matrix according to the channel state information CSI compression model and the channel state information CSI compression parameter to obtain a compressed target channel matrix includes: Inputting the first channel matrix into the feature converter, performing key feature extraction on the first channel matrix, and obtaining a second channel matrix representing CSI key features; The second channel matrix is ​​compressed according to the CSI compression parameter and the channel encoder to obtain the target channel matrix.

4. The method according to claim 3, characterized in that The compressing the second channel matrix according to the CSI compression parameter and the channel encoder to obtain the target channel matrix includes: Using the sub-encoder corresponding to the CSI compression parameter as the second target sub-encoder; The second channel matrix is ​​compressed by the second target sub-encoder to obtain a target channel matrix.

5. The method according to claim 3, characterized in that The feature converter includes a feature extraction network, an attention mechanism network, and a feature restoration network; inputting the first channel matrix into the feature converter, extracting key features of the first channel matrix, and obtaining a second channel matrix representing CSI key features includes: Inputting the first channel matrix into the feature extraction network to obtain a plurality of first feature maps; Inputting the plurality of first feature maps into the attention mechanism network to obtain a second feature map; the second feature map includes key feature information in the first feature map; The second feature map is input into the feature restoration network to obtain a second channel matrix.

6. The method according to claim 5, characterized in that Inputting the plurality of first feature maps into the attention mechanism network to obtain a second feature map includes: Performing a maximum pooling operation on the plurality of first feature maps through the attention mechanism network to obtain a maximum pooling feature map; Performing an average pooling operation on the plurality of first feature maps through the attention mechanism network to obtain an average pooled feature map; Determine a second feature map according to the maximum pooling feature map and the average pooling feature map.

7. The method according to claim 6, characterized in that The attention mechanism network includes a fusion subnetwork; and determining the second feature map according to the maximum pooling feature map and the average pooling feature map includes: Inputting the maximum pooling feature map and the average pooling feature map into the fusion sub-network to obtain a fused fusion feature map; A second feature map is determined based on the fused feature map and the first feature map.

8. The method according to claim 1, characterized in that The acquiring a first channel matrix according to the pilot signal includes: According to the pilot signal, a spatial channel matrix is ​​measured; Transforming the spatial domain channel matrix into an angle delay domain channel matrix through discrete Fourier transform; The first channel matrix is ​​determined according to the angle delay domain channel matrix.

9. The method according to any one of claims 1 to 8, characterized in that The CSI compression model is trained in the following way: Obtaining a first sample channel matrix for training; the first sample channel matrix is ​​a matrix obtained by the terminal device based on the received pilot signal and used to characterize the downlink channel quality; Training a first target network model according to the first sample channel matrix to obtain the CSI compression model; Among them, the first target network model includes a first target compression model and a first target decompression model, the network structure of the first target compression model is the same as the network structure of the CSI compression model, the first target decompression model includes a channel decoder, and the channel decoder includes multiple sub-decoders, and different sub-decoders correspond to different CSI compression parameters.

10. The method according to claim 9, characterized in that The training of the first target network model according to the first sample channel matrix includes: The first model training step is performed in a loop until it is determined that the trained first target network model satisfies a first preset stop iteration condition based on the first sample channel matrix and the first predicted channel matrix, and the first target compression model in the trained first target network model is used as the CSI compression model; the first predicted channel matrix is ​​a matrix output after the first sample channel matrix is ​​input into the first target network model; The first model training step includes: Inputting the first sample channel matrix into the first target compression model, and compressing the first sample channel matrix through a plurality of sub-encoders to obtain a first target sample channel matrix; Inputting the first target sample channel matrix into the first target decompression model, and decompressing the first target sample channel matrix through a plurality of sub-decoders to obtain a first predicted channel matrix; When it is determined according to the first sample channel matrix and the first predicted channel matrix that the first target network model does not meet the first preset stop iteration condition, a first loss value is determined according to the first sample channel matrix and the first predicted channel matrix, and the parameters of the first target network model are updated according to the first loss value to obtain the trained first target network model, and the trained first target network model is used as the new first target network model.

11. The method according to claim 9, characterized in that The method further comprises: Obtaining first decompression model parameters corresponding to the first target decompression model in the trained first target network model; The first decompression model parameters are sent to the network device to instruct the network device to determine a channel state information CSI decompression model based on the first decompression model parameters. The CSI decompression model is used by the network device to determine the channel quality of the downlink channel based on the target channel matrix.

12. The method according to any one of claims 1 to 8, characterized in that The method further comprises: receiving a second compression model parameter sent by the network device; A CSI compression model is determined according to the second compression model parameters.

13. The method according to any one of claims 1 to 8, characterized in that The method further comprises: receiving a first compression parameter sent by a network device; The CSI compression parameter is determined according to the first compression parameter.

14. A method for obtaining channel quality, characterized in that: Applied to a network device, the method includes: receiving a target channel matrix sent by a terminal device; the target channel matrix is ​​obtained by the terminal device by compressing a first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter, where the first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal and used to characterize the channel quality of a downlink channel; decompressing the target channel matrix according to a channel state information (CSI) decompression model and the CSI compression parameter to obtain a third channel matrix; wherein the CSI decompression model includes a channel decoder, the channel decoder includes a plurality of sub-decoders, and different sub-decoders correspond to different CSI compression parameters; The channel quality of the downlink channel is determined according to the third channel matrix.

15. The method according to claim 14, characterized in that The decompressing the target channel matrix according to the channel state information CSI decompression model and the CSI compression parameter to obtain a third channel matrix includes: Using the sub-decoder corresponding to the CSI compression parameter as the target sub-decoder; decompressing the target channel matrix by the target sub-decoder to obtain a fourth channel matrix; The third channel matrix is ​​determined according to the fourth channel matrix.

16. The method according to claim 15, characterized in that The CSI decompression model further includes a CSI reconstruction module; and determining the third channel matrix according to the fourth channel matrix includes: The fourth channel matrix is ​​input into the CSI reconstruction module to obtain the third channel matrix.

17. The method according to any one of claims 14 to 16, characterized in that The CSI decompression model is trained in the following way: Obtaining a second sample channel matrix for training; the second sample channel matrix is ​​a matrix obtained by the terminal device based on the received pilot signal and used to characterize the downlink channel quality; Training a second target network model according to the second sample channel matrix to obtain the CSI decompression model; Among them, the second target network model includes a second target compression model and a second target decompression model, the network structure of the second target decompression model is the same as the network structure of the CSI decompression model, the second target compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders, and different sub-encoders correspond to different CSI compression parameters.

18. The method according to claim 17, characterized in that The training of the second target network model according to the second sample channel matrix includes: The second model training step is performed in a loop until it is determined that the trained second target network model satisfies a second preset stop iteration condition based on the second sample channel matrix and the second predicted channel matrix, and the second target decompression model in the trained second target network model is used as the CSI decompression model; the second predicted channel matrix is ​​a matrix output after the second sample channel matrix is ​​input into the second target network model; The second model training step includes: Inputting the second sample channel matrix into the second target compression model, and compressing the second sample channel matrix through a plurality of sub-encoders to obtain a second target sample channel matrix; Inputting the second target sample channel matrix into the second target decompression model, and decompressing the second target sample channel matrix through a plurality of sub-decoders to obtain a second predicted channel matrix; When it is determined according to the second sample channel matrix and the second predicted channel matrix that the second target network model does not meet the second preset stop iteration condition, a second loss value is determined according to the second sample channel matrix and the second predicted channel matrix, and the parameters of the second target network model are updated according to the second loss value to obtain a trained second target network model, and the trained second target network model is used as a new second target network model.

19. The method according to claim 17, wherein The method further comprises: Obtaining second compression model parameters corresponding to the second target compression model in the trained second target network model; The second compression model parameters are sent to the terminal device to instruct the terminal device to determine a CSI compression model based on the second compression model parameters, and the CSI compression model is used by the terminal device to obtain a target channel matrix based on the first channel matrix.

20. The method according to any one of claims 14 to 16, characterized in that The method further comprises: Receiving a first decompression model parameter sent by the terminal device; Determine a CSI decompression model according to the first decompression model parameters.

21. The method according to any one of claims 14 to 16, characterized in that The method further comprises: Determining a first compression parameter according to the CSI compression parameter; The first compression parameter is sent to the terminal device.

22. A device for obtaining channel quality, characterized in that: Applied to a terminal device, the device includes: A first receiving module is configured to receive a pilot signal sent by a network device through a downlink channel; a first matrix acquisition module, configured to acquire a first channel matrix according to the pilot signal; the first channel matrix is ​​used to characterize the channel quality of the downlink channel; a target matrix acquisition module configured to compress the first channel matrix according to a channel state information (CSI) compression model and a channel state information (CSI) compression parameter to obtain a compressed target channel matrix; wherein the CSI compression model includes a channel encoder, and the channel encoder includes multiple sub-encoders; different sub-encoders correspond to different CSI compression parameters; The first sending module is configured to send the target channel matrix to the network device, so that the network device determines the channel quality of the downlink channel according to the target channel matrix.

23. A device for obtaining channel quality, characterized in that: Applied to network equipment, the device includes: A second receiving module is configured to receive a target channel matrix sent by a terminal device; the target channel matrix is ​​obtained by the terminal device by compressing a first channel matrix according to a CSI compression model and a channel state information (CSI) compression parameter, where the first channel matrix is ​​a matrix obtained by the terminal device according to a pilot signal and used to characterize the channel quality of a downlink channel; a third matrix acquisition module configured to decompress the target channel matrix according to a channel state information (CSI) decompression model and the CSI compression parameters to obtain a third channel matrix; wherein the CSI decompression model includes a channel decoder, the channel decoder includes a plurality of sub-decoders, and different sub-decoders correspond to different CSI compression parameters; The channel quality determination module is configured to determine the channel quality of the downlink channel according to the third channel matrix.

24. A device for obtaining channel quality, characterized in that: The device comprises: processor; a memory for storing processor-executable instructions; The processor is configured to execute the steps of the method described in any one of claims 1 to 13, or the processor is configured to execute the steps of the method described in any one of claims 14 to 21.

25. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the steps of the method described in any one of claims 1 to 13 are implemented; or when the computer program instructions are executed by a processor, the steps of the method described in any one of claims 14 to 21 are implemented.

26. A chip, characterized in that: Comprising a processor and an interface; the processor is used to read instructions to execute the steps of the method described in any one of claims 1 to 13, or the processor is used to read instructions to execute the steps of the method described in any one of claims 14 to 21.

Images