A method, apparatus and device for channel compression and recovery

Abstract

This invention provides a channel compression and recovery method, apparatus, and device. The channel compression method applied to a terminal includes: measuring a reference signal transmitted by a network device to obtain channel state information; preprocessing the channel state information to obtain a preprocessing result; compressing the channel state information using a target channel compression model based on the preprocessing result to obtain compressed channel information; and sending the compressed channel information and / or the preprocessing result to the network device. The solution of this invention improves channel compression feedback performance by preprocessing the channel state information, using a trained target channel compression model to compress the preprocessed channel state information, and performing subsequent channel recovery.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and in particular to a channel compression and recovery method, apparatus, and device. Background Technology

[0002] In multiple-input multiple-output (MIMO) systems, acquiring channel state information (CSI) is crucial for beamforming to improve transmission performance. For FDD (Frequency Division Multiplexing) systems, due to the lack of complete uplink and downlink reciprocity, base stations need to rely on terminal feedback to obtain complete downlink CSI. In NR (Normally Restricted Radiant) systems, terminal CSI feedback primarily relies on codebooks. Currently, CSI type I, type II, and type II enhanced codebooks are supported for feeding back information such as RI (rank indicator), PMI (precoding matrix indicator), and CQI (channel quality indicator).

[0003] Traditional machine learning-based channel feedback schemes treat the channel matrix as a whole and directly input it into the model for processing. Although this method is simple to implement, its performance is often limited because it lacks a design that incorporates channel characteristics. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a channel compression and recovery method, apparatus and device to improve channel processing performance and reduce channel feedback overhead.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] An embodiment of the present invention provides a channel compression method applied to a terminal, the method comprising:

[0007] The channel state information is obtained by measuring the reference signal sent by the network device;

[0008] The channel state information is preprocessed to obtain the preprocessing result;

[0009] Based on the preprocessing results, the channel state information is compressed using the target channel compression model to obtain the compressed channel information.

[0010] The compressed channel information and / or the preprocessing results are sent to the network device.

[0011] Optionally, the channel state information is preprocessed to obtain a preprocessing result, including one of the following:

[0012] The channel state information is transformed into the angle domain, time delay domain, or Doppler domain to obtain the first preprocessing result;

[0013] The channel state information or the first preprocessing result is truncated and preprocessed to obtain the second preprocessing result;

[0014] The channel state information or the first preprocessing result is subjected to cyclic shift preprocessing to obtain the third preprocessing result.

[0015] Optionally, transforming the channel state information to the angle domain, time delay domain, or Doppler domain includes:

[0016] The channel state information is transformed into the angle domain by inverse fast four-factor transform (IFFT) in the antenna dimension, into the time delay domain by inverse fast four-factor transform (IFFT) in the frequency dimension, and into the Doppler domain by sparse four-factor transform (SFFT) in the time dimension.

[0017] Optionally, the channel state information is pruned and preprocessed to obtain a second preprocessing result, including:

[0018] When performing the first pruning preprocessing on the channel state information, a first number of elements are retained, and the remaining elements are pruned to obtain the first pruning preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold; or

[0019] When performing the second pruning preprocessing on the channel state information, the positions of the second number of elements to be retained are marked in a bitmap manner, and the remaining elements are pruned to obtain the second pruning preprocessing result. The second number is a second preset value or the number of elements to be retained obtained by filtering according to a second preset threshold.

[0020] Optionally, the first cropping preprocessing result includes: the positions of a first number of elements and the remaining elements obtained after cropping;

[0021] The second cropping preprocessing result includes: marking the positions of the second number of elements to be retained in a bitmap format and the remaining elements obtained from cropping.

[0022] Optionally, the first quantity and the second quantity are pre-configured by the network device or determined by the terminal;

[0023] The first quantity and the second quantity are pre-configured to the terminal via higher-layer signaling when the network device is pre-configured;

[0024] When the first quantity and the second quantity are determined by the terminal, the first quantity and the second quantity determined by the terminal are reported to the network device.

[0025] Optionally, the channel state information is subjected to cyclic shift preprocessing to obtain a third preprocessing result, including:

[0026] If the first element of the channel meets the preset conditions, the target element of the channel is cyclically shifted to the position of the first element, and the position corresponding to the target element is recorded to obtain the third preprocessing result.

[0027] Optionally, the third preprocessing result includes: the channel obtained after cyclic shifting, and the position corresponding to the target element.

[0028] Optionally, the channel state information is subjected to Fourier transform to the angle domain, time delay domain, or Doppler domain, including:

[0029] The channel state information is transformed into the angle domain by Fast Fourier Transform (FFT), into the time delay domain by Inverse Fast Fourier Transform (IFFT), and into the Doppler domain by Sparse Fourier Transform (SFFT).

[0030] Optionally, the first preprocessing result is subjected to a cropping preprocessing to obtain a second preprocessing result, including:

[0031] Based on the first preprocessing result, a first number of elements around the maximum value of the channel delay and / or the angle are retained, and the remaining elements are truncated to obtain the second preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold.

[0032] Optionally, the first preprocessing result is subjected to cyclic shift preprocessing to obtain a third preprocessing result, including:

[0033] Based on the first preprocessing result, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then start from the element with the maximum value of the channel delay and / or the reserved angle and sequentially shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

[0034] Optionally, the channel state information includes at least one of the following:

[0035] Channel matrix;

[0036] Channel feature vector;

[0037] Channel precoding matrix.

[0038] Optionally, the target channel compression model is selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different pruning ratios are trained through the following process:

[0039] Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0040] The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios.

[0041] The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

[0042] Embodiments of the present invention also provide a channel recovery method applied to a network device, the method comprising:

[0043] Send a reference signal to the terminal;

[0044] The receiver receives compressed channel information and / or preprocessing results obtained by preprocessing estimated channel state information during channel compression. The compressed channel information is obtained by the terminal through channel measurement based on a reference signal, and the terminal preprocesses the channel state information to obtain a preprocessing result. Based on the preprocessing result, the channel state information is compressed using a target channel compression model.

[0045] The compressed channel information is recovered according to the target channel decompression model to obtain the recovered channel information.

[0046] Optionally, the preprocessing result includes one of the following:

[0047] The first preprocessing result obtained by transforming the channel state information into the angle domain, time delay domain, or Doppler domain;

[0048] The second preprocessing result is obtained by pruning and preprocessing the channel state information or the first preprocessing result;

[0049] The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result.

[0050] Optionally, the compressed channel information is recovered according to the target channel decompression model to obtain the recovered channel information, including:

[0051] Based on the target channel decompression model and the preprocessing results, the compressed channel information is recovered to obtain the recovered channel information.

[0052] Optionally, the target channel decompression model is selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different pruning ratios are trained through the following process:

[0053] Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0054] The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios.

[0055] The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

[0056] Embodiments of the present invention also provide a terminal, comprising:

[0057] The estimation module is used to measure the reference signal sent by the network device to obtain channel state information;

[0058] The processing module is used to preprocess the channel state information to obtain a preprocessing result; and based on the preprocessing result, to compress the channel state information using a target channel compression model to obtain compressed channel information.

[0059] The transceiver module is used to send the compressed channel information and / or the preprocessing results to the network device.

[0060] Embodiments of the present invention also provide a network device, comprising:

[0061] The transceiver module is used to send a reference signal to the terminal; and to receive compressed channel information and / or preprocessing results obtained by preprocessing channel state information during channel compression from the terminal; the compressed channel information is obtained by the terminal through channel measurement of the reference signal, and the preprocessing of the channel state information is performed to obtain the preprocessing result; and the channel is compressed using the target channel compression model based on the preprocessing result.

[0062] The processing module is used to recover the compressed channel information according to the target channel decompression model to obtain the recovered channel information.

[0063] Embodiments of the present invention also provide a communication device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above.

[0064] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods described above.

[0065] The above-described solution of the present invention has at least the following beneficial effects:

[0066] In this invention, channel state information is obtained by measuring the reference signal sent by the network device; the channel state information is preprocessed to obtain a preprocessing result; based on the preprocessing result, the channel state information is compressed using a target channel compression model to obtain compressed channel information; the compressed channel information and / or the preprocessing result are sent to the network device; thereby improving the performance of channel compression and the accuracy of subsequent channel recovery. Attached Figure Description

[0067] Figure 1 This is a flowchart of the channel method on the terminal side provided in an embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram illustrating the implementation principle of channel compression and recovery provided in an embodiment of the present invention;

[0069] Figure 3 This is a flowchart illustrating the implementation of the channel compression and recovery method provided in this embodiment of the invention.

[0070] Figure 4 This is a schematic diagram of a terminal module block provided in an embodiment of the present invention. Detailed Implementation

[0071] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0072] like Figure 1 As shown, an embodiment of the present invention provides a channel compression method applied to a terminal, the method comprising:

[0073] Step 11: Measure the reference signal sent by the network device to obtain channel state information;

[0074] Step 12: Preprocess the channel state information to obtain the preprocessing result;

[0075] Step 13: Based on the preprocessing results, compress the channel state information using the target channel compression model to obtain the compressed channel information;

[0076] Step 14: Send the compressed channel information and / or the preprocessing result to the network device.

[0077] In this embodiment, the reference signal is periodically sent to the terminal by the network device and is used for channel information measurement. The reference signal can be a CSI-RS channel state information acquisition signal, etc. The terminal obtains channel state information based on the reference signal. The terminal preprocesses the channel state information according to the characteristics of the data elements in the channel state information to remove unnecessary data element features and reduce the impact of unnecessary data features on subsequent processing steps. The preprocessing result includes, but is not limited to, the preprocessed channel information, the data element information in the preprocessed channel information, and the data element information in the channel information removed during the preprocessing process.

[0078] Based on the preprocessing results, the terminal selects a target channel compression model from a pre-trained set of target channel compression models that is compatible with the preprocessed channel state information. This target channel compression model is then used to compress and quantize the preprocessed channel state information to improve the performance of channel compression. The information in the compressed channel can exist in the form of a bitstream. The set of target channel compression models is obtained by pruning and training multiple historical channels according to different pruning ratios. The target channel compression model can be obtained by the terminal processing and training historical channel data, or by the network device processing and training historical channel data.

[0079] The terminal sends the compressed channel information and / or the preprocessing result to the network device so that the network device can recover the compressed channel.

[0080] In an optional embodiment of the present invention, the preprocessing result in step 12 above may include one of the following:

[0081] The channel state information is transformed into the angle domain, time delay domain, or Doppler domain to obtain the first preprocessing result;

[0082] The channel state information or the first preprocessing result is truncated and preprocessed to obtain the second preprocessing result;

[0083] The channel state information or the first preprocessing result is subjected to cyclic shift preprocessing to obtain the third preprocessing result.

[0084] In this embodiment, the channel state information can be transformed to the angle domain, time delay domain, or Doppler domain;

[0085] The channel state information can also be preprocessed by pruning;

[0086] The channel state information can also be preprocessed by cyclic shifting;

[0087] Alternatively, the channel state information can be transformed into the angle domain, time delay domain, or Doppler domain in the antenna dimension through Fourier transform, and then cropping preprocessing and / or cyclic shift preprocessing can be performed based on the Fourier transform results.

[0088] Specifically, the pruning preprocessing and cyclic shift preprocessing of the channel state information are performed on the near-zero values ​​in the estimated channel.

[0089] The near-zero values ​​are truncated and cyclically shifted to filter out interference terms in the channel. The channel state information is then transformed into the angle domain, time delay domain, or Doppler domain using Fourier transform to avoid the influence of unnecessary features in subsequent channel processing and improve the accuracy of subsequent channel recovery processing.

[0090] In an optional embodiment of the present invention, transforming the channel state information to the angle domain, time delay domain, or Doppler domain via Fourier transform may include:

[0091] The channel state information is transformed into the angle domain in the antenna dimension using Inverse Fast Fourier Transform (IFFT), into the time delay domain in the frequency dimension using Inverse Fast Fourier Transform (IFFT), and into the Doppler domain in the time dimension using Sparse Fast Fourier Transform (SFFT).

[0092] In this embodiment, when performing Fourier transform on the channel state information, the IDFT transformation can be performed on the channel state information according to the antenna dimension at row N1 points and column N2 points to the angle domain; the IDFT transformation can be performed on the channel state information according to the frequency domain dimension at N3 points to the time delay domain; and the SFFT transformation can be performed on the channel state information according to the time dimension at N4 points to the Doppler domain. Here, N1, N2, N3, and N4 can also be configured by the network device.

[0093] In an optional embodiment of the present invention, the channel state information is subjected to pruning preprocessing to obtain a second preprocessing result, which may include:

[0094] When performing the first pruning preprocessing on the channel state information, a first number of elements are retained, and the remaining elements are pruned to obtain the first pruning preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold; or...

[0095] When performing the second pruning preprocessing on the channel state information, the positions of the second number of elements to be retained are marked in a bitmap manner, and the remaining elements are pruned to obtain the second pruning preprocessing result. The second number is a second preset value or the number of elements to be retained obtained by filtering according to a second preset threshold.

[0096] In this embodiment, either the first pruning preprocessing result or the second pruning preprocessing result is the second preprocessing result; the selection of the first pruning preprocessing and the second pruning preprocessing can be based on the distribution characteristics of the data elements in the channel state information;

[0097] When data elements in the channel state information appear in clusters, that is, when some data elements are distributed around the highest value data element and form a fixed distribution area, the channel delay and / or the first number of elements around the maximum angle value are retained, and the channel state information is subjected to a first pruning preprocessing to remove near-zero values ​​outside the fixed area, and the first pruning preprocessing result is obtained. The highest value can be the highest value in the angle domain or the highest value in the delay domain of the data element.

[0098] When the data elements in the channel state information are discretely distributed and near-zero values ​​cannot be cropped out by a fixed region, the positions of the second number of elements to be retained are marked in a graphical manner, the other unmarked elements are cropped out, and a second cropping preprocessing result is obtained.

[0099] The data elements in the channel state information are pruned by comprehensively considering the distribution of data elements in the channel state information in order to improve the efficiency and accuracy of pruning; the first quantity is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold, and the second quantity is a second preset value or the number of elements to be retained obtained by filtering according to a second preset threshold. The first quantity, the second quantity, the first preset value or the first preset value and the second preset value or the second preset threshold can be pre-configured according to the channel state information or the actual application situation.

[0100] In an optional embodiment of the present invention, the first cutting preprocessing result includes: the positions of a first number of elements and the remaining elements obtained by cutting;

[0101] The second cropping preprocessing result includes: marking the positions of the second number of elements to be retained in a bitmap format and the remaining elements obtained from cropping.

[0102] In this embodiment, the terminal sends the first pruning preprocessing result and / or the second pruning preprocessing result after the channel state information is pruned to the network device, so that the network device can perform subsequent channel processing based on the first pruning preprocessing result and / or the second pruning preprocessing result;

[0103] The first pruning preprocessing result may include: when the channel state information is pruned for the first time, the channel data in the first pruned channel state information obtained, namely the positions of a first number of elements around the maximum value of the retained channel delay and / or angle, and other elements that are pruned;

[0104] The second pruning preprocessing result may include: when the channel state information is pruned for the second time, the channel data in the second pruned channel obtained, that is, the positions of the second number of elements marked in a bitmap manner that are retained, and the other elements that are pruned;

[0105] Of course, the first pruning preprocessing result may also include the first pruning channel state information obtained by performing the first pruning; the second pruning preprocessing result may also include the second pruning channel state information obtained by performing the second pruning.

[0106] Furthermore, the first quantity and the second quantity are pre-configured by the network device or determined by the terminal; when the first quantity and the second quantity are pre-configured by the network device, the network device pre-configures them to the terminal through higher-layer signaling; when the first quantity and the second quantity are determined by the terminal, the first quantity and the second quantity determined by the terminal are reported to the network device.

[0107] In an optional embodiment of the present invention, performing cyclic shift preprocessing on the channel state information to obtain a third preprocessing result may include:

[0108] If the first element of the channel meets the preset conditions, the target element of the channel is cyclically shifted to the position of the first element, and the position corresponding to the target element is recorded to obtain the third preprocessing result.

[0109] Furthermore, the third preprocessing result includes: the channel obtained after cyclic shifting, and the position corresponding to the target element.

[0110] In this embodiment, by performing cyclic shift preprocessing on the estimated channel, the conditions for pruning preprocessing are not met. The cyclic shift preprocessing removes unnecessary data elements, thereby ensuring the accuracy of subsequent channel compression recovery.

[0111] In an optional embodiment of the present invention, the first preprocessing result is subjected to a cropping preprocessing to obtain a second preprocessing result, including:

[0112] Based on the first preprocessing result, a first number of elements surrounding the maximum channel delay and / or the maximum angle are retained, and the remaining elements are pruned to obtain a second preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold. The third pruning preprocessing result may include: the positions of the first number of elements surrounding the maximum channel delay and / or the maximum angle, and the remaining elements obtained by pruning.

[0113] In an optional embodiment of the present invention, a third preprocessing result is obtained by performing a cyclic shift preprocessing on the first preprocessing result, including:

[0114] Based on the first preprocessing result, if it is determined that the first element of the channel is not the maximum value of the channel delay and / or the maximum value of the retained angle, then the channel delay and / or the maximum value of the retained angle are sequentially and cyclically shifted from the element with the maximum value of the channel delay and / or the maximum value of the retained angle to the position of the first element, and the position corresponding to the maximum value of the channel delay and / or the maximum value of the retained angle is recorded to obtain the third preprocessing result. That is, the channel state information is first subjected to Fourier transform to the angle domain, delay domain or Doppler domain, and then cyclic shift preprocessing is performed based on the result of Fourier transform.

[0115] In specific implementation, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then starting from the element with the maximum value of the channel delay and / or the reserved angle, a sequential cyclic shift is performed, cyclically shifting the maximum value of the channel delay and / or the reserved angle to the position of the first element, and recording the position corresponding to the maximum value of the channel delay and / or the reserved angle, thus obtaining the third preprocessing result. Here, the third preprocessing result may include: the channel obtained after cyclic shifting, and the position corresponding to the target element with the maximum value of the channel delay and / or the reserved angle.

[0116] Of course, when the preprocessing result includes the first preprocessing result, the second preprocessing result, and the third preprocessing result, that is, the channel state information is first subjected to Fourier transform to the angle domain, the time delay domain, or the Doppler domain, then the Fourier transform result is truncated, and then the truncated result is subjected to cyclic shift preprocessing.

[0117] Alternatively, the channel state information can be first transformed into the angle domain, time delay domain, or Doppler domain using Fourier transform, and then the result of the Fourier transform can be preprocessed by cyclic shifting, followed by pruning of the cyclic shift. In specific implementation, the pruning and cyclic shifting processes described above are followed.

[0118] It should be noted that the method by which the terminal side preprocesses the channel state information can be configured by the network device through higher-layer signaling or pre-agreed upon.

[0119] In an optional embodiment of the present invention, the channel state information is described, and the channel state information includes at least one of the following: a channel matrix; a channel feature vector; and a channel precoding matrix.

[0120] In an optional embodiment of the present invention, the target channel compression model is described, wherein the target channel compression model is a target machine learning model selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different compression ratios are trained through the following process:

[0121] Step 01: Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0122] Step 02: Perform multiple cropping processes on the downlink channels in the downlink channel dataset with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios;

[0123] Step 03: Based on the cropping results of multiple different preset cropping ratios, input them into the preset model for training to obtain multiple machine learning models with different compression ratios.

[0124] In this embodiment, it should be understood that the multiple machine learning models with different compression ratios can be trained by the terminal or by the network device. The terminal can use the simulation model to generate frequency domain reused FDD downlink channel data or notify the terminal to measure and report FDD downlink channel data when there are few services and sufficient air interface resources, and establish an FDD downlink channel dataset. Alternatively, any downlink channel data acquisition method that meets the requirements can be used to complete the process.

[0125] The terminal trains a machine learning model for channel compression and recovery based on the established FDD downlink channel dataset. The machine learning model includes a compression network model deployed at the feedback transmitter and a recovery network model deployed at the feedback receiver. During the training process, the downlink channels in the downlink channel dataset can be pre-processed by pruning according to the characteristics of the current channel dataset.

[0126] Since the sparsity of the actual channel delay domain or other dimensions that are pruned varies, multiple models need to be trained according to the pruning ratio. For example, models can be trained for different categories such as no pruning, pruning to 3 / 4 of the elements, pruning to 1 / 2 of the elements, pruning to 1 / 4 of the elements, and other proportions of elements retained. The trained models at this time process the pruned channel. Therefore, different models can be trained for the same dataset and different pruning ratios. Multiple models suitable for different pruning situations are combined into a model set for selection in subsequent processes.

[0127] In an optional embodiment of the present invention, the channel compression method described above may further include:

[0128] Based on the channel pruning preprocessing method and the pruning ratio, the machine learning models with different pruning ratios are updated;

[0129] In this embodiment, by detecting the compression performance of the target channel compression model, when the model performance drops below a preset compression threshold due to changes in the channel environment, a corresponding model update mechanism is triggered. Based on the channel pruning method and pruning ratio, the machine learning models with different pruning ratios are updated to ensure the performance of channel feedback.

[0130] Embodiments of the present invention also provide a channel recovery method applied to a network device, the method comprising:

[0131] Step 21: Send a reference signal to the terminal;

[0132] Step 22: Receive the compressed channel information fed back by the terminal and / or the preprocessing result obtained by preprocessing the channel state information during channel compression; the compressed channel information is obtained by the terminal through channel measurement of the reference signal, and the channel state information is preprocessed by pruning to obtain the preprocessing result; and based on the preprocessing result, the channel state information is compressed using the target channel compression model.

[0133] Step 23: Recover the compressed channel according to the target channel decompression model to obtain the recovered channel.

[0134] In this embodiment, the reference signal sent by the network device to the terminal can be downlink channel data generated by a simulation model, or it can be complete downlink channel data reported by the network device to the terminal when the service is not busy. The reference signal is mainly used by the terminal to obtain channel state information through channel measurement. After the network device sends the reference signal, it further receives the compressed channel information sent by the terminal and / or the preprocessing result obtained by preprocessing the channel state information when compressing the channel. The compressed channel is then restored according to the target channel decompression model to improve the channel recovery performance. The compressed channel information is the channel state information obtained by the terminal through channel measurement based on the reference signal, and the channel state information is then cropped and compressed. The target decompression model is a model selected from multiple pre-trained machine learning models with different compression ratios that is adapted to the compressed channel information.

[0135] In an optional embodiment of the present invention, the preprocessing result includes one of the following:

[0136] The first preprocessing result is obtained by performing a Fourier transform on the channel state information to the angle domain, time delay domain, or Doppler domain;

[0137] The second preprocessing result is obtained by pruning and preprocessing the channel state information or the first preprocessing result;

[0138] The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result.

[0139] In this embodiment, the preprocessing result is obtained by the base station through pruning preprocessing, cyclic shift preprocessing, or Fourier transform of the channel state information, or through Fourier transform followed by pruning and / or cyclic shifting.

[0140] The third preprocessing result is obtained by the base station from the channel state information through Fourier transform;

[0141] The second preprocessing result is obtained by the base station retaining a first number of elements around the maximum value of the channel delay and / or the maximum angle in the channel state information, and pruning the remaining elements; or by the base station marking the positions of the second number of retained elements in the channel state information in a bitmap manner, and pruning the remaining elements.

[0142] The third preprocessing result is obtained by the base station sequentially and cyclically shifting the channel state information starting from the element with the maximum value of channel delay and / or reserved angle, cyclically shifting the maximum value of channel delay and / or reserved angle to the position of the first element, and recording the position corresponding to the maximum value of channel delay and / or reserved angle.

[0143] In an optional embodiment of the present invention, step 23 may include:

[0144] Step 231: Based on the target channel decompression model and the preprocessing results, the compressed channel information is recovered to obtain the recovered channel information.

[0145] In this embodiment, the pruning preprocessing result is obtained by the terminal pruning the channel state to near zero value according to different pruning methods; after receiving the compressed channel information and / or the preprocessing result, the network device first uses the target decompression model to perform recovery processing on the compressed channel information to obtain the first recovered channel information, and then performs information recovery before pruning on the first recovered channel information according to the pruning preprocessing result to obtain the recovered second recovered channel information.

[0146] In an optional embodiment of the present invention, the target channel decompression model is described, wherein the target channel compression model is a target machine learning model selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different compression ratios are trained through the following process:

[0147] Step 011: Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0148] Step 012: Perform multiple cropping processes on the downlink channels in the downlink channel dataset with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios;

[0149] Step 013: Based on the cropping results of multiple different preset cropping ratios, input them into the preset model for training to obtain multiple machine learning models with different compression ratios.

[0150] In this embodiment, it should be understood that the multiple machine learning models with different compression ratios can be trained by the terminal or by the network device; the training method of the multiple deep neural network models with different compression ratios is the same as the training method on the terminal side, and will not be described again here.

[0151] In an optional embodiment of the present invention, the channel recovery method applied to a network device may further include:

[0152] Configure multiple deep neural network models with different compression ratios to the terminal.

[0153] In this embodiment, when the multiple deep neural network models with different compression ratios are trained on the network device side, the network device configures the trained multiple deep neural network models with different compression ratios to the terminal so that the terminal can compress the channel. Of course, the multiple machine learning models with different pruning ratios can also be trained by the terminal.

[0154] The following will illustrate the above channel compression and recovery method with a specific example, such as... Figure 2 The diagram shown is a schematic of the channel compression and recovery method; as follows: Figure 3 As shown, the specific implementation process is as follows:

[0155] Step 31: The network device performs data collection, including but not limited to: generating downlink channel data using a simulation model; or notifying the terminal to report complete downlink channel data when traffic is not heavy, etc.; the network device trains the model based on the data received from the mobile phone and establishes a model set according to different values ​​of specific parameters. When establishing the model set, it should include different models trained for different pruning ratios; it should be understood that the model can also be obtained through on-the-terminal training.

[0156] Step 32: The base station sends CSI channel state information to the terminal through the downlink control channel, and sends downlink reference signals for channel information measurement to the terminal. The CSI channel state information includes the channel resources occupied by feedback, feedback overhead, and the energy ratio threshold for pruning and retention.

[0157] Step 33: The terminal estimates the downlink channel based on the reference signal, selects a pruning method according to the distribution characteristics of the data elements in the estimated channel, and prunes the estimated channel; further, the terminal selects a suitable target compression model from the model set according to the pruning result to compress the pruned channel and obtain a compressed channel, wherein the information in the compressed channel exists in the form of a bit stream.

[0158] Step 34: The terminal sends feedback information to the network device. The feedback information may include: cropping method information; cropping result information; target compression model or target compression model selection information; and compression channel information.

[0159] Step 35: Based on the feedback information, the network device first recovers the truncated channel using the target compression model and the compressed channel information; then, based on the truncating method information and truncating result information, it recovers the original channel before truncating and uses it for subsequent transmission signal processing.

[0160] Step 36: The network device and the terminal exchange information and update the compression model according to the channel pruning method and channel pruning ratio.

[0161] In the above embodiments of the present invention, through interaction between the network device and the terminal, the terminal performs preprocessing on the estimated channel to filter out interference terms in the channel, effectively improving the subsequent recovery accuracy; then, a compression model with a corresponding pruning ratio is used to compress the pruned estimated channel, and the result is fed back to the network device. The network device first recovers the pruned channel, and then obtains the original channel based on the pruning information, thereby improving the performance of the channel compression recovery method.

[0162] Embodiments of the present invention also provide a terminal, such as Figure 4 As shown, the terminal 40 includes:

[0163] The estimation module 41 is used to measure the reference signal sent by the network device to obtain channel state information;

[0164] Processing module 42 is used to preprocess the channel state information to obtain a preprocessing result; and based on the preprocessing result, to compress the channel state information using a target channel compression model to obtain compressed channel information.

[0165] The transceiver module 43 is used to send the compressed channel information and / or the preprocessing result to the network device.

[0166] Optionally, the processing module 42 is used to preprocess the channel state information to obtain a preprocessing result, including one of the following:

[0167] The first preprocessing result obtained by transforming the channel state information into the angle domain, time delay domain, or Doppler domain;

[0168] The second preprocessing result is obtained by pruning and preprocessing the channel state information or the first preprocessing result;

[0169] The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result.

[0170] Optionally, transforming the channel state information to the angle domain, time delay domain, or Doppler domain includes:

[0171] The channel state information is transformed into the angle domain by inverse fast Fourier transform (IFFT) in the antenna dimension, into the time delay domain by inverse fast Fourier transform (IFFT) in the frequency dimension, and into the Doppler domain by sparse fourier transform (SFFT) in the time dimension.

[0172] Optionally, the processing module 42 is specifically used to: when performing the first pruning preprocessing on the channel state information, retain a first number of elements, prune the remaining elements, and obtain the first pruning preprocessing result, wherein the first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold; or

[0173] When performing the second pruning preprocessing on the channel state information, the positions of the second number of elements to be retained are marked in a bitmap manner, and the remaining elements are pruned to obtain the second pruning preprocessing result. The second number is a second preset value or the number of elements to be retained obtained by filtering according to a second preset threshold.

[0174] Optionally, the first cropping preprocessing result includes: the positions of a first number of elements and the remaining elements obtained after cropping;

[0175] The second cropping preprocessing result includes: marking the positions of the second number of elements to be retained in a bitmap format and the remaining elements obtained from cropping.

[0176] Optionally, the first quantity and the second quantity are pre-configured by the network device or determined by the terminal;

[0177] The first quantity and the second quantity are pre-configured to the terminal via higher-layer signaling when the network device is pre-configured;

[0178] When the first quantity and the second quantity are determined by the terminal, the first quantity and the second quantity determined by the terminal are reported to the network device.

[0179] Optionally, the processing module 42 is specifically used to, if the first element of the channel meets the preset conditions, cyclically shift the target element of the channel to the position of the first element, and record the position corresponding to the target element to obtain the third preprocessing result.

[0180] Optionally, the third preprocessing result includes: the channel obtained after cyclic shifting, and the position corresponding to the target element.

[0181] Optionally, the first preprocessing result is subjected to a cropping preprocessing to obtain a second preprocessing result, including:

[0182] Based on the first preprocessing result, a first number of elements around the maximum value of the channel delay and / or the angle are retained, and the remaining elements are truncated to obtain the second preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold.

[0183] Optionally, the first preprocessing result is subjected to cyclic shift preprocessing to obtain a third preprocessing result, including:

[0184] Based on the first preprocessing result, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then start from the element with the maximum value of the channel delay and / or the reserved angle and sequentially shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

[0185] Optionally, the channel state information includes at least one of the following: a channel matrix; a channel feature vector; and a channel precoding matrix.

[0186] Optionally, the target channel compression model is selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different pruning ratios are trained through the following process:

[0187] Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0188] The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios.

[0189] The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

[0190] It should be noted that this terminal is the terminal corresponding to the channel compression method applied to the terminal described above. All implementation methods in the above method embodiments are applicable to this embodiment of the terminal and can achieve the same technical effect.

[0191] Embodiments of the present invention also provide a network device, the network device comprising:

[0192] The transceiver module is used to send a reference signal to the terminal; and to receive compressed channel information and / or preprocessing results obtained by preprocessing channel state information during channel compression from the terminal; the compressed channel information is obtained by the terminal through channel measurement of the reference signal, and the preprocessing of the channel state information is performed to obtain the preprocessing result; and the channel is compressed using the target channel compression model based on the preprocessing result.

[0193] The processing module is used to recover the compressed channel information according to the target channel decompression model to obtain the recovered channel information.

[0194] Optionally, the preprocessing result includes one of the following:

[0195] The first preprocessing result obtained by transforming the channel state information into the angle domain, time delay domain, or Doppler domain;

[0196] The second preprocessing result is obtained by pruning and preprocessing the channel state information or the first preprocessing result;

[0197] The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result.

[0198] Optionally, the processing module is used to recover the compressed channel information according to the target channel decompression model, and the recovered channel information includes:

[0199] Based on the target channel decompression model and the preprocessing results, the compressed channel information is recovered to obtain the recovered channel information.

[0200] Optionally, the target channel decompression model is selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different pruning ratios are trained through the following process:

[0201] Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels;

[0202] The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios.

[0203] The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

[0204] It should be noted that this network device is the network device corresponding to the channel recovery method applied to network devices described above. All implementation methods in the above method embodiments are applicable to the embodiments of this network device and can achieve the same technical effect.

[0205] All implementation methods described above are applicable to the embodiments of this terminal and can achieve the same technical effect.

[0206] Embodiments of the present invention also provide a communication device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0207] Embodiments of the present invention also provide a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0208] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0209] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0210] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0211] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0212] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0213] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0214] Furthermore, it should be noted that in the apparatus and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent solutions of the present invention. Moreover, the steps performing the above-described series of processes can naturally be executed in the order described, but are not necessarily required to be executed in chronological order; some steps can be executed in parallel or independently of each other. Those skilled in the art will understand that all or any step or component of the method and apparatus of the present invention can be implemented in any computing device (including processors, storage media, etc.) or network of computing devices, in hardware, firmware, software, or a combination thereof. This is something that those skilled in the art can achieve by using their basic programming skills after reading the description of the present invention.

[0215] Therefore, the object of the present invention can also be achieved by running a program or a set of programs on any computing device. The computing device can be a known general-purpose device. Therefore, the object of the present invention can also be achieved simply by providing a program product containing program code implementing the method or apparatus. That is, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any known storage medium or any storage medium developed in the future. It should also be noted that in the apparatus and method of the present invention, it is obvious that the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered equivalent to the present invention. Furthermore, the steps performing the above series of processes can naturally be performed in the order described, but are not necessarily required to be performed in chronological order. Some steps can be performed in parallel or independently of each other.

[0216] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A channel compression method, characterized in that, Applied to a terminal, the method includes: The channel state information is obtained by measuring the reference signal sent by the network device; The channel state information is preprocessed to obtain the preprocessing result; Based on the preprocessing results, the channel state information is compressed using the target channel compression model to obtain the compressed channel information. The compressed channel information and / or the preprocessing result are sent to the network device; The channel state information is preprocessed to obtain a preprocessing result, including: The channel state information is transformed into the angle domain, time delay domain, or Doppler domain to obtain the first preprocessing result; The channel state information or the first preprocessing result is subjected to cyclic shift preprocessing to obtain the third preprocessing result; The step of performing cyclic shift preprocessing on the channel state information to obtain a third preprocessing result includes: If the first element of the channel meets the preset condition, the target element of the channel is cyclically shifted to the position of the first element, and the position corresponding to the target element is recorded to obtain the third preprocessing result; The step of performing a cyclic shift preprocessing on the first preprocessing result to obtain a third preprocessing result includes: Based on the first preprocessing result, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then start from the element with the maximum value of the channel delay and / or the reserved angle and sequentially shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

2. The channel compression method according to claim 1, characterized in that, The channel state information is preprocessed to obtain a preprocessing result, which also includes: The channel state information or the first preprocessing result is truncated and preprocessed to obtain the second preprocessing result.

3. The channel compression method according to claim 2, characterized in that, Transforming the channel state information to the angle domain, time delay domain, or Doppler domain includes: The channel state information is transformed into the angle domain by inverse fast four-factor transform (IFFT) in the antenna dimension, into the time delay domain by inverse fast four-factor transform (IFFT) in the frequency dimension, and into the Doppler domain by sparse four-factor transform (SFFT) in the time dimension.

4. The channel compression method according to claim 2, characterized in that, The channel state information is pruned and preprocessed to obtain a second preprocessing result, including: When performing the first pruning preprocessing on the channel state information, a first number of elements are retained, and the remaining elements are pruned to obtain the first pruning preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold; or When performing the second pruning preprocessing on the channel state information, the positions of the second number of elements to be retained are marked in a bitmap manner, and the remaining elements are pruned to obtain the second pruning preprocessing result. The second number is a second preset value or the number of elements to be retained obtained by filtering according to a second preset threshold.

5. The channel compression method according to claim 4, characterized in that, The first preprocessing result includes: the positions of a first number of elements and the remaining elements obtained after cropping; The second cropping preprocessing result includes: marking the positions of the second number of elements to be retained in a bitmap format and the remaining elements obtained from cropping.

6. The channel compression method according to claim 4, characterized in that, The first quantity and the second quantity are pre-configured by the network device or determined by the terminal; The first quantity and the second quantity are pre-configured to the terminal via higher-layer signaling when the network device is pre-configured; When the first quantity and the second quantity are determined by the terminal, the first quantity and the second quantity determined by the terminal are reported to the network device.

7. The channel compression method according to claim 1, characterized in that, The third preprocessing result includes: the channel obtained after cyclic shifting, and the position corresponding to the target element.

8. The channel compression method according to claim 2, characterized in that, The first preprocessing result is subjected to a cropping preprocessing to obtain a second preprocessing result, including: Based on the first preprocessing result, a first number of elements around the maximum value of the channel delay and / or the angle are retained, and the remaining elements are truncated to obtain the second preprocessing result. The first number is a first preset value or the number of elements to be retained obtained by filtering according to a first preset threshold.

9. The channel compression method according to claim 1, characterized in that, The channel state information includes at least one of the following: Channel matrix; Channel feature vector; Channel precoding matrix.

10. The channel compression method according to claim 1, characterized in that, The target channel compression model is selected from a plurality of pre-trained machine learning models with different pruning ratios; the plurality of machine learning models with different pruning ratios are trained through the following process: Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels; The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios. The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

11. A channel recovery method, characterized in that, Applied to network devices, the method includes: Send a reference signal to the terminal for channel estimation; The compressed channel information received from the receiving terminal and / or the preprocessing result obtained by preprocessing the channel state information during channel compression; the compressed channel information is obtained by the terminal through channel measurement based on the reference signal, and the channel state information is preprocessed to obtain the preprocessing result, and the channel state information is compressed using the target channel compression model based on the preprocessing result. The compressed channel information is recovered according to the target channel decompression model to obtain the recovered channel information. The preprocessing results include: The first preprocessing result obtained by transforming the channel state information into the angle domain, time delay domain, or Doppler domain; The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result; The third preprocessing result, obtained by performing cyclic shift preprocessing on the channel state information, includes: If the first element of the channel meets the preset condition, the target element of the channel is cyclically shifted to the position of the first element, and the position corresponding to the target element is recorded to obtain the third preprocessing result; The third preprocessing result is obtained by performing a cyclic shift preprocessing on the first preprocessing result, including: Based on the first preprocessing result, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then start from the element with the maximum value of the channel delay and / or the reserved angle and sequentially shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

12. The channel recovery method according to claim 11, characterized in that, The preprocessing results also include: The second preprocessing result is obtained by pruning and preprocessing the channel state information or the first preprocessing result.

13. The channel recovery method according to claim 12, characterized in that, The compressed channel information is recovered based on the target channel decompression model, and the recovered channel information includes: Based on the target channel decompression model and the preprocessing results, the compressed channel information is recovered to obtain the recovered channel information.

14. The channel recovery method according to claim 11, characterized in that, The target channel decompression model is selected from a pre-trained set of machine learning models with different pruning ratios; the machine learning models with different pruning ratios are trained through the following process: Obtain the downlink channel dataset generated by the simulation model or reported by the terminal, which is formed by frequency domain multiplexing downlink channels; The downlink channels in the downlink channel dataset are subjected to multiple cropping processes with different preset cropping ratios to obtain multiple cropping results with different preset cropping ratios. The cropping results from multiple preset cropping ratios are input into preset machine learning models for training, resulting in multiple machine learning models with different cropping ratios.

15. A terminal, characterized in that, include: The estimation module is used to measure the reference signal sent by the network device to obtain channel state information; The processing module is used to preprocess the channel state information to obtain a preprocessing result; Based on the preprocessing results, the channel state information is compressed using the target channel compression model to obtain the compressed channel information. The transceiver module is used to send the compressed channel information and / or the preprocessing result to the network device; The processing module is used to preprocess the channel state information to obtain a preprocessing result, including: The channel state information is transformed into the angle domain, time delay domain, or Doppler domain to obtain the first preprocessing result; The channel state information or the first preprocessing result is subjected to cyclic shift preprocessing to obtain the third preprocessing result; The processing module is also used to, if the first element of the channel meets the preset conditions, cyclically shift the target element of the channel to the position of the first element, and record the position corresponding to the target element to obtain the third preprocessing result; The processing module is further configured to, based on the first preprocessing result, determine if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then sequentially and cyclically shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

16. A network device, characterized in that, include: The transceiver module is used to send reference signals to the terminal; It also receives compressed channel information fed back by the terminal and / or preprocessing results obtained by preprocessing channel state information during channel compression; The compressed channel information is obtained by the terminal through channel measurement of the reference signal to obtain channel state information, preprocessing the channel state information to obtain a preprocessing result, and then compressing the channel state information using the target channel compression model based on the preprocessing result. The processing module is used to recover the compressed channel information according to the target channel decompression model to obtain the recovered channel information; The preprocessing results include: The first preprocessing result obtained by transforming the channel state information into the angle domain, time delay domain, or Doppler domain; The third preprocessing result is obtained by performing cyclic shift preprocessing on the channel state information or the first preprocessing result; The third preprocessing result, obtained by performing cyclic shift preprocessing on the channel state information, includes: If the first element of the channel meets the preset condition, the target element of the channel is cyclically shifted to the position of the first element, and the position corresponding to the target element is recorded to obtain the third preprocessing result; The third preprocessing result is obtained by performing a cyclic shift preprocessing on the first preprocessing result, including: Based on the first preprocessing result, if the first element of the channel is not the maximum value of the channel delay and / or the reserved angle, then start from the element with the maximum value of the channel delay and / or the reserved angle and sequentially shift the maximum value of the channel delay and / or the reserved angle to the position of the first element, and record the position corresponding to the maximum value of the channel delay and / or the reserved angle to obtain the third preprocessing result.

17. A communication device, characterized in that, include: A processor, a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method as described in any one of claims 1 to 10 or the method as described in any one of claims 11 to 14.

18. A computer-readable storage medium, characterized in that, Store instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 10 or the method as described in any one of claims 11 to 14.

Images