Channel information processing method, terminal, base station and readable storage medium

Abstract

The present invention provides a method, terminal, base station, and storage medium for processing channel information. The method includes: obtaining channel information to be processed, determining the distribution of each element in the channel information to be processed; selecting a corresponding clipping method based on the distribution to clip near-zero values ​​in the channel information to obtain target information, and compressing and quantizing the target using a preset deep learning model to obtain quantized information; sending the quantized information, a deep learning model identifier, and the clipping method to a base station, wherein the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantized information to the target information, and the clipping method is used to assist the base station in padding the target information with corresponding zero values ​​according to the clipping method to obtain corresponding channel information. The present invention can reduce channel system overhead.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of communication technologies, and in particular to a method for processing channel information, a terminal, a base station, and a readable storage medium. Background Art

[0002] In multiple-input multiple-output (MIMO) systems, obtaining channel state information (CSI) is a key requirement for beamforming to improve transmission performance. In frequency-division duplex (FDD) systems, due to the lack of complete uplink and downlink channel reciprocity, base stations rely on terminal feedback to obtain complete downlink CSI. In new radio (NR) systems, terminal CSI feedback primarily relies on a codebook.

[0003] The codebooks used in existing systems are primarily based on the Discrete Fourier Transform (DFT) matrix. This approach treats the channel as an isotropic, randomly distributed system and uses the DFT vectors to evenly divide the beam space. However, in real systems, channels do not strictly follow this isotropic distribution pattern but are instead concentrated in a specific subspace. This inability to effectively utilize information about the subspace in which the channel resides limits the performance of DFT codebooks.

[0004] Currently, existing technologies use deep neural networks to extract features from the entire channel information, and then compress and restore it to achieve the transmission of channel information. However, this method is based on the entire channel data for calculation, and the required system overhead is relatively large. Summary of the Invention

[0005] Embodiments of the present invention provide a channel information processing method, a terminal, a base station, and a readable storage medium, aiming to reduce channel system overhead.

[0006] To solve the above problems, the present invention is achieved as follows:

[0007] In a first aspect, an embodiment of the present invention provides a method for processing channel information, which is executed by a terminal. The method includes:

[0008] Acquiring channel information to be processed, and determining the distribution of each element in the channel information to be processed;

[0009] Selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compressing and quantizing the target through a preset deep learning model to obtain quantized information;

[0010] The quantization information, the deep learning model identifier and the cropping method are sent to the base station, the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

[0011] In a second aspect, an embodiment of the present invention provides a method for processing channel information, which is executed by a base station and includes:

[0012] Quantization information, deep learning model identification and cropping method sent by the receiving terminal;

[0013] Using the algorithm corresponding to the deep learning model identifier to restore the above-mentioned quantified information to target information;

[0014] The target information is zero-filled according to the clipping method to obtain restored channel information.

[0015] In a third aspect, an embodiment of the present invention further provides a terminal, comprising: an acquisition module, configured to acquire channel information to be processed and determine a distribution of each element in the channel information to be processed;

[0016] A clipping module, configured to select a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compress and quantize the target using a preset deep learning model to obtain quantized information;

[0017] A sending module is used to send the quantization information, the identifier of the deep learning model and the cropping method to a base station, wherein the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

[0018] In a fourth aspect, an embodiment of the present invention further provides a base station, including:

[0019] A receiving module, configured to receive quantization information, a deep learning model identifier, and a cropping method sent by a terminal;

[0020] A recovery module, configured to recover the quantized information into target information using an algorithm corresponding to the deep learning model identifier;

[0021] A filling module is used to fill the target information with zero values ​​according to the cropping method to obtain restored channel information.

[0022] In a fifth aspect, an embodiment of the present invention further provides a readable storage medium for storing a program, which, when executed by a processor, implements the steps in the method described in the first aspect.

[0023] An embodiment of the present invention obtains channel information to be processed to determine the distribution of each element in the channel information to be processed; selects a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compresses and quantizes the target through a preset deep learning model to obtain quantization information; sends the quantization information, the deep learning model identifier and the clipping method to a base station, the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the clipping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the clipping method to obtain corresponding channel information. In this way, before sending channel information to the base station, the terminal first determines the distribution of elements in the channel information, and then uses the corresponding clipping method to clip the near-zero values ​​in the channel information according to the distribution, and obtains the clipped target information. Then, the target information is compressed and quantized using a deep learning model to obtain quantized information. Finally, the quantized information, the identifier of the deep learning model used, and the clipping method are sent to the base station. The base station can select the corresponding algorithm based on the identifier of the deep learning model to restore the quantized information to the target information, and then fill the target information with the corresponding zero value according to the clipping method to obtain the corresponding channel information. The obtained channel information is basically the same as the information clipped by the terminal. In this way, the present application retains the valuable information in the channel information, and clips the remaining near-zero information and then compresses and quantizes it, reducing system overhead. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in describing the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0025] Figure 1 is a schematic diagram of the structure of a network system to which an embodiment of the present invention is applicable;

[0026] Figure 2 This is one of the flow charts of the method for processing channel information provided by an embodiment of the present invention;

[0027] Figure 3 Schematic diagram of energy distribution of channel information in an embodiment of the present invention;

[0028] Figure 4is a schematic diagram of energy distribution of another channel information in an embodiment of the present invention;

[0029] Figure 5 This is a second flow chart of a method for processing channel information provided by an embodiment of the present invention;

[0030] Figure 6 This is one of the structural diagrams of the terminal provided by the embodiment of the present invention;

[0031] Figure 7 This is one of the structural diagrams of the base station provided by the implementation of the present invention. DETAILED DESCRIPTION

[0032] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0033] The terms "first", "second" and the like in the embodiments of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices. In addition, "and / or" is used in the present invention to represent at least one of the connected objects, such as A and / or B and / or C, which means seven situations including single A, single B, single C, and both A and B exist, both B and C exist, both A and C exist, and both A, B and C exist.

[0034] See Figure 1 , Figure 1 is a structural diagram of a network system to which an embodiment of the present invention can be applied, such as Figure 1 As shown, it includes a terminal 11 and a base station 12.

[0035] The terminal 11 and the base station 12 can communicate with each other.

[0036] In practical applications, terminal 11 can be a mobile phone, tablet computer, laptop computer, personal digital assistant (PDA), mobile internet device (MID), wearable device, or vehicle-mounted device. Base station 12 is a public mobile communication base station, which is an interface device for mobile devices to access the Internet. It is also a form of radio station, referring to a radio transceiver station that transmits information to and from mobile phone terminals through a mobile communication exchange center within a certain radio coverage area.

[0037] The following describes a method for processing channel information provided by an embodiment of the present invention.

[0038] See also Figure 2 , Figure 2 This is one of the flow charts of the channel information processing method provided by an embodiment of the present invention. Figure 2 The channel information processing method shown can be executed by a data cell.

[0039] like Figure 2 As shown, the channel information processing method may include the following steps:

[0040] Step 200: Acquire channel information to be processed and determine the distribution of each element in the channel information to be processed;

[0041] The channel information to be processed may be any type of channel information between the terminal and the base station. For example, the channel information to be processed may be downlink channel data generated by a simulation model, or may be used to notify the terminal to report complete downlink channel data when the service is not busy.

[0042] Before sending a message to the base station, the terminal obtains channel information to be sent as channel information to be processed, and then analyzes the information to be processed to determine the distribution of each element in the channel information to be processed.

[0043] As an example, the channel information to be processed is the channel information of the frequency division duplex FDD downlink system. The original channel information of the downlink system is the frequency domain-antenna domain channel information (information to be processed) estimated by each subcarrier using the downlink reference signal. After performing fast Fourier transform FFT on the original channel information along the frequency domain and antenna domain, the time delay-antenna domain channel information with a more sparse representation can be obtained, such as Figure 3 As shown, the distribution of each element in the channel information can be determined, where the important information (valuable information) in the channel information is Figure 3 The white and gray parts in the middle and the rest of the black part are information with a near-zero value, that is, information in the channel information that has no value for this communication, and its energy is 0 or close to 0. In other words, a near-zero value means that the energy is less than or equal to a value close to 0.

[0044] Step 210: Select a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compress and quantize the target using a preset deep learning model to obtain quantized information;

[0045] Based on the channel information to be processed obtained in step S200 and the distribution of each element in the channel information, the channel information is clipped using a clipping method corresponding to the distribution, thereby clipping the near-zero values ​​of the black portion and retaining other elements with higher energy values ​​as target information. The target information is then compressed and quantized using a preset deep learning model to obtain corresponding quantized information. Specifically, the process of compression and quantization using a deep learning model in this embodiment can use existing technologies and will not be described in detail here.

[0046] Step 220: Send the quantization information, the deep learning model identifier and the cropping method to the base station. The deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information. The cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

[0047] After obtaining the quantization information, the terminal can send the quantization information, the identifier of the deep learning model and the cropping method to the base station. The base station can select the corresponding algorithm according to the identifier of the deep learning model to restore the quantization information to the target information, and then fill the target information with zeros according to the cropping method to restore the target information to the corresponding channel information. The channel information is basically the same as the channel information to be processed by the terminal.

[0048] An embodiment of the present invention obtains channel information to be processed to determine the distribution of each element in the channel information to be processed; selects a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compresses and quantizes the target through a preset deep learning model to obtain quantization information; sends the quantization information, the deep learning model identifier and the clipping method to a base station, the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the clipping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the clipping method to obtain corresponding channel information. In this way, before sending channel information to the base station, the terminal first determines the distribution of elements in the channel information, and then uses the corresponding clipping method to clip the near-zero values ​​in the channel information according to the distribution, and obtains the clipped target information. Then, the target information is compressed and quantized using a deep learning model to obtain quantized information. Finally, the quantized information, the identifier of the deep learning model used, and the clipping method are sent to the base station. The base station can select the corresponding algorithm based on the identifier of the deep learning model to restore the quantized information to the target information, and then fill the target information with the corresponding zero value according to the clipping method to obtain the corresponding channel information. The obtained channel information is basically the same as the information clipped by the terminal. In this way, the present application retains the valuable information in the channel information, and clips the remaining near-zero information and then compresses and quantizes it, reducing system overhead.

[0049] Optionally, in some embodiments, based on the above embodiment, the step of selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information may include:

[0050] If the channel information to be processed is distributed around the element with the highest value, information outside the preset distance from the element with the highest value is clipped, and information within the preset distance from the element with the highest value is used as target information.

[0051] This embodiment is applicable to the case where the distribution of each element in the channel information is centered around the element with the highest value, such as Figure 3 As shown, the highest energy position is the position where the delay domain is 11, and the remaining non-zero energy values ​​are distributed with the position where the delay domain is 11 as the center. In this distribution, the information outside the preset distance from the element with the highest value is clipped, and the information within the preset distance from the element with the highest value is used as the target information. For example, Figure 3 The content between 5 and 16 in the middle delay domain (6 delay units away from the center point) is retained as target information, and the other content is cropped.

[0052] Furthermore, to ensure that the target information retains all valuable content, the ratio of the energy of the target information to the energy of the channel information to be processed can be greater than or equal to a preset ratio, wherein the preset ratio can be determined based on the type of information being sent. Generally, the preset ratio is greater than or equal to 80%. The specific steps of clipping information outside the preset distance from the element with the highest value and selecting information within the preset distance from the element with the highest value as the target information may include:

[0053] Sort by the energy value of each element, starting from the element with the highest energy, and select them in sequence until the ratio of the energy value of the selected multiple elements to the total energy value of the channel information is greater than or equal to the preset ratio. Then the selected multiple elements are the target information, and the other information is cropped.

[0054] Furthermore, in order to speed up the cropping speed, multiple cropping models are set in this embodiment. For example, the cropping models include cropping 3 / 4 and leaving 1 / 4; cropping 1 / 2 and leaving 1 / 2; cropping 1 / 4 and leaving 3 / 4. The cropping model selection process includes: calculating the ratio of the energy value of the information remaining after cropping by the concentrated cropping model to the total energy, with the element with the highest energy as the center, and then selecting a cropping model with the smallest ratio.

[0055] Furthermore, to ensure that the information after trimming retains all valuable content, a trimming model is selected from the trimmed models to satisfy the requirement that the ratio of the energy of the remaining information to the total energy is greater than or equal to a preset ratio, and that the energy value of the remaining information is the smallest among all the models. For example, if the preset ratio is 80%, the trimming model used is: trim 3 / 4, leaving 1 / 4, and the ratio of the energy of the target information to the total energy is 79%; the trimming model used is: trim 1 / 2, leaving 1 / 2, and the ratio of the energy of the target information to the total energy is 85%; the trimming model used is: trim 1 / 4, leaving 3 / 4, and the ratio of the energy of the target information to the total energy is 90%. Since 79% < 80% < 85% < 90%, the following cropping models are satisfied: crop 3 / 4, leave 1 / 4, and crop 1 / 2, leave 1 / 2. Both models can retain all valuable content. However, the cropping model of cropping 1 / 2, leave 1 / 2 leaves more content than cropping 3 / 4, leave 1 / 4. Therefore, the cropping model of cropping 3 / 4, leave 1 / 4 is selected. In this embodiment, 3 / 4 is cropped with the element with the highest energy as the center, and the remaining 1 / 4 of the content is used as the target information.

[0056] As another embodiment, based on the above embodiment, the step of selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information may also include:

[0057] If the elements in the channel information to be processed that exceed the preset value are discretely distributed, the position of the target element whose energy value exceeds the preset value is marked in a bitmap manner, the information outside the preset range centered on the target element is cropped, and the information within the preset range centered on the target element is used as the target information.

[0058] This embodiment is applicable to the case where the elements exceeding the preset value in the channel information to be processed are discretely distributed, that is, there are multiple elements with higher energy values, and these elements are discretely distributed, such as Figure 4 , white and gray points are distributed in multiple positions. At this time, the clipping is performed with multiple points with the highest energy as the center. The information within the preset range centered on multiple points with the highest energy is used as the target information, and the rest of the information is clipped.

[0059] Furthermore, to ensure that the target information retains all valuable content, the ratio of the energy of the target information to the energy of the channel information to be processed can be greater than or equal to a preset ratio, wherein the preset ratio is determined according to the type of information to be transmitted. Generally, the preset ratio is greater than or equal to 80%. The specific steps of clipping information outside the preset range centered on the target element and treating information within the preset range centered on the target element as the target information may include:

[0060] Sort by the energy value of each element, starting from the element with the highest energy, and select them in sequence until the ratio of the energy value of the selected multiple elements to the total energy value of the channel information is greater than or equal to the preset ratio. Then the selected multiple elements are the target information, and the other information is cropped.

[0061] Furthermore, in order to speed up the cropping speed, multiple cropping models can also be set in this embodiment. For example, the cropping models include cropping 3 / 4 and leaving 1 / 4; cropping 1 / 2 and leaving 1 / 2; cropping 1 / 4 and leaving 3 / 4. The cropping model selection process includes: calculating the ratio of the energy value of the information remaining after cropping by the concentrated cropping model to the total energy, with the element with the highest energy as the center, and then selecting a cropping model with the smallest ratio.

[0062] Furthermore, to ensure that the information after trimming retains all valuable content, a trimming model is selected from the trimmed models to satisfy the requirement that the ratio of the energy of the remaining information to the total energy is greater than or equal to a preset ratio, and that the energy value of the remaining information is the smallest among all the models. For example, if the preset ratio is 80%, the trimming model used is: trim 3 / 4, leaving 1 / 4, and the ratio of the energy of the target information to the total energy is 79%; the trimming model used is: trim 1 / 2, leaving 1 / 2, and the ratio of the energy of the target information to the total energy is 85%; the trimming model used is: trim 1 / 4, leaving 3 / 4, and the ratio of the energy of the target information to the total energy is 90%. Since 79% < 80% < 85% < 90%, the following cropping models are satisfied: crop 3 / 4, leave 1 / 4, and crop 1 / 2, leave 1 / 2. Both models can retain all valuable content. However, the cropping model of cropping 1 / 2, leave 1 / 2 leaves more content than the cropping model of cropping 3 / 4, leave 1 / 4. Therefore, the cropping model of cropping 3 / 4, leave 1 / 4 is selected. In this embodiment, 3 / 4 of the content is cropped with each element as the center, and the remaining 1 / 4 of the content is used as the target information.

[0063] See Figure 5 , Figure 5 This is a second flow chart of a method for processing channel information provided by an embodiment of the present invention. The method is executed by a base station and includes:

[0064] Step 500: receiving quantization information, deep learning model identifier, and cropping method sent by the terminal;

[0065] Step 510: Using the algorithm corresponding to the deep learning model identifier to restore the quantized information to target information;

[0066] Step 520: Fill the target information with zero values ​​according to the cropping method to obtain restored channel information.

[0067] This embodiment illustrates the above-mentioned channel information processing process from the perspective of the base station. Specifically, after completing the channel information processing, the terminal sends the corresponding quantization information to the base station. The acquisition of the quantization information is the same as in the above-mentioned embodiments. The base station can select the corresponding algorithm according to the identifier of the deep learning model to restore the quantization information to the target information, and then fill the target information with zeros according to the cropping method to restore the target information to the corresponding channel information. The channel information is basically the same as the channel information to be processed processed by the terminal.

[0068] It should be noted that the implementation method of the channel processing method implemented by the terminal in this embodiment is the same as that in the above method embodiment. Therefore, the various processing flows can refer to the relevant descriptions in the above method embodiment, and the same beneficial effects can be achieved. To avoid repetition, they will not be repeated here.

[0069] See Figure 6 , Figure 6 This is one of the structural diagrams of a terminal provided in an embodiment of the present invention. The terminal includes:

[0070] An acquisition module 600 is configured to acquire channel information to be processed and determine the distribution of each element in the channel information to be processed;

[0071] A clipping module 610 is configured to select a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compress and quantize the target using a preset deep learning model to obtain quantized information;

[0072] The sending module 620 is used to send the quantization information, the identifier of the deep learning model and the cropping method to the base station, wherein the deep learning model identifier is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

[0073] Optionally, the clipping module 610 is further configured to clip information outside a preset distance from the element with the highest value if the channel information to be processed is distributed centered around the element with the highest value, and use information within a preset distance from the element with the highest value as target information.

[0074] Optionally, the cropping module 610 is further configured to:

[0075] Selecting a first target remaining ratio from a preset remaining ratio set;

[0076] The information outside the channel information to be processed is cropped according to the first target remaining ratio with the element with the highest value as the center, and the first target remaining ratio satisfies that the energy value of the target information corresponding to the first target remaining ratio is less than the energy value of the target information corresponding to other remaining ratios in the preset remaining ratio set.

[0077] Optionally, the cropping module 610 is further configured to:

[0078] If the elements in the channel information to be processed that exceed the preset value are discretely distributed, the position of the target element whose energy value exceeds the preset value is marked in a bitmap manner, the information outside the preset range centered on the target element is cropped, and the information within the preset range centered on the target element is used as the target information.

[0079] Optionally, the cropping module 610 is further configured to:

[0080] selecting a second target remaining ratio from a preset remaining ratio set;

[0081] The channel information to be processed is cropped with the target element as the center to obtain target information with the target element as the center and the remaining elements as the second target residual ratio, and the second target residual ratio satisfies that the energy value of the target information corresponding to the second target residual ratio is less than the energy value of the target information corresponding to other residual ratios in the preset residual ratio set.

[0082] Optionally, a ratio of the energy of the target information to the energy of the channel information to be processed is greater than or equal to a preset ratio.

[0083] It should be noted that the terminal provided in this embodiment can implement each process of the embodiment of the channel information processing method in the embodiment of the present invention and achieve the same beneficial effects. To avoid repetition, they will not be described here.

[0084] The various optional implementations introduced in the embodiments of the present invention may be implemented in combination with each other or individually if they do not conflict with each other, and the embodiments of the present invention do not limit this.

[0085] See Figure 7 , Figure 7 This is one of the structural diagrams of a base station provided by an embodiment of the present invention. An embodiment of the present invention further provides a base station, the base station including:

[0086] The receiving module 700 is configured to receive quantization information, an identifier of a deep learning model, and a cropping method sent by a terminal;

[0087] A recovery module 710 is configured to recover the quantized information into target information using an algorithm corresponding to the deep learning model identifier;

[0088] The filling module 720 is configured to perform zero-value filling on the target information according to the cropping method to obtain restored channel information.

[0089] It should be noted that the base station provided in this embodiment can realize the embodiment of the present invention. Figure 5 The various processes of the embodiment of the channel information processing method shown achieve the same beneficial effects, and will not be described again here to avoid repetition.

[0090] Those skilled in the art will appreciate that all or part of the steps in implementing the method of the above-described embodiment can be accomplished by hardware associated with program instructions, and the program can be stored in a readable medium. The present invention also provides a readable storage medium having a computer program stored thereon. When executed by a processor, the computer program can implement any step in the method embodiment corresponding to any of the above-described embodiments, and can achieve the same technical effects. To avoid repetition, the details are not repeated here.

[0091] The storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0092] The above is a preferred implementation of the embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for processing channel information, executed by a terminal, characterized in that: The method comprises: Acquiring channel information to be processed, and determining the distribution of each element in the channel information to be processed; Selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compressing and quantizing the target information through a preset deep learning model to obtain quantized information; The quantization information, the identifier of the deep learning model and the cropping method are sent to the base station. The identifier of the deep learning model is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information. The cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

2. The method according to claim 1, characterized in that The step of selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information includes: If the channel information to be processed is distributed around the element with the highest value, information outside the preset distance from the element with the highest value is clipped, and information within the preset distance from the element with the highest value is used as target information.

3. The method according to claim 2, characterized in that The step of cutting out information outside the preset distance from the element with the highest value and taking information within the preset distance from the element with the highest value as target information includes: Selecting a first target remaining ratio from a preset remaining ratio set; Information other than the channel information to be processed is cropped according to the first target remaining ratio with the element with the highest value as the center, and the first target remaining ratio satisfies that the energy value of the target information corresponding to the first target remaining ratio is less than the energy value of the target information corresponding to other remaining ratios in the preset remaining ratio set.

4. The method according to claim 1, wherein The step of selecting a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information also includes: If the elements in the channel information to be processed that exceed the preset value are discretely distributed, the position of the target element whose energy value exceeds the preset value is marked in a bitmap manner, the information outside the preset range centered on the target element is cropped, and the information within the preset range centered on the target element is used as the target information.

5. The method according to claim 4, characterized in that The steps of marking the position of the target element whose energy value exceeds the preset value in a bitmap manner, clipping the information outside the preset range centered on the target element, and using the information within the preset range centered on the target element as the target information include: selecting a second target remaining ratio from a preset remaining ratio set; The channel information to be processed is cropped with the target element as the center to obtain target information with the target element as the center and the remaining elements as the second target residual ratio, and the second target residual ratio satisfies that the energy value of the target information corresponding to the second target residual ratio is less than the energy value of the target information corresponding to other residual ratios in the preset residual ratio set.

6. The method according to any one of claims 1 to 5, characterized in that The ratio of the energy of the target information to the energy of the channel information to be processed is greater than or equal to a preset ratio.

7. A method for processing channel information, executed by a base station, characterized in that: The method comprises: Quantization information, a deep learning model identifier, and a clipping method sent by a receiving terminal, wherein the quantization information is obtained by the terminal compressing and quantizing target information using a preset deep learning model, and the target information is obtained by the terminal clipping near-zero values ​​in the channel information by selecting a corresponding clipping method based on the distribution of each element in the channel information to be processed; Restoring the quantized information to the target information using an algorithm corresponding to the identifier of the deep learning model; The target information is zero-filled according to the clipping method to obtain restored channel information.

8. A terminal, characterized in that: include: an acquisition module, configured to acquire channel information to be processed and determine the distribution of each element in the channel information to be processed; a clipping module, configured to select a corresponding clipping method according to the distribution to clip near-zero values ​​in the channel information to obtain target information, and compress and quantize the target information through a preset deep learning model to obtain quantized information; A sending module is used to send the quantization information, the identifier of the deep learning model and the cropping method to a base station, wherein the identifier of the deep learning model is used to assist the base station in using a corresponding algorithm to restore the quantization information to the target information, and the cropping method is used to assist the base station in filling the target information with corresponding zero values ​​according to the cropping method to obtain corresponding channel information.

9. A base station, characterized in that: include: a receiving module, configured to receive quantization information, a deep learning model identifier, and a clipping method sent by a terminal, wherein the quantization information is obtained by the terminal compressing and quantizing target information using a preset deep learning model, and the target information is obtained by the terminal clipping near-zero values ​​in the channel information by selecting a corresponding clipping method based on the distribution of each element in the channel information to be processed; A recovery module, configured to recover the quantized information into the target information using an algorithm corresponding to the identifier of the deep learning model; A filling module is used to fill the target information with zero values ​​according to the cropping method to obtain restored channel information.

10. A readable storage medium for storing a program, characterized in that: When the program is executed by a processor, the steps of the channel information processing method according to any one of claims 1 to 6 are implemented; or the steps of the channel information processing method according to claim 7 are implemented.

Images