Codebook determination method and device
By employing a network model-based method to determine centroids and construct an effective space for codebooks, the method addresses the inefficiencies in existing MIMO systems, enhancing performance and reducing complexity.
Patent Information
- Application Number
- CN202110697874.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-06-23
- Publication Date
- 2024-12-17
- Estimated Expiration
- 2041-06-23
Smart Images
Figure CN115589243B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of mobile communication technology, and in particular to a codebook determination method and device. Background Art
[0002] Most existing codebook schemes aim to obtain a codebook with better performance through more accurate channel information, so the processing of channel information is crucial to the performance of the system.
[0003] With the increasing number of antennas in Multiple-in Multiple-out (MIMO) systems, the performance of classical clustering algorithms will be affected by big data when processing high-dimensional data, and their processing delay and complexity will also increase greatly as the amount of data continues to increase.
[0004] Therefore, how to make full use of channel characteristics to efficiently determine the codebook scheme is of great significance. Summary of the invention
[0005] The present invention provides a codebook determination method and device, which are used to solve the technical problems in the prior art that the codebook scheme does not fully utilize channel characteristics and the codebook scheme determination process is complicated.
[0006] In a first aspect, the present invention provides a codebook determination method, comprising:
[0007] Obtaining a characteristic matrix of a channel sample sequence;
[0008] Clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0009] Based on the centroids of the characteristic matrix, constructing an effective space of a codebook;
[0010] Determine each codeword in the codebook according to the effective space;
[0011] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0012] In one embodiment, the step of acquiring a characteristic matrix of a channel sample sequence includes:
[0013] Inputting the channel sample sequence into a prediction network model to determine an initial characteristic matrix of the channel sample sequence;
[0014] Inputting the initial characteristic matrix into the reinforcement network model to determine a current reinforcement characteristic matrix;
[0015] Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches a preset threshold:
[0016] Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence;
[0017] Inputting the current characteristic matrix into the enhanced network model to determine a next enhanced characteristic matrix;
[0018] The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
[0019] In one embodiment, clustering the characteristic matrix to determine each centroid of the characteristic matrix includes:
[0020] Selecting a preset number of initial centroids in the characteristic matrix;
[0021] Determining a preset number of characteristic clusters according to the preset number of initial centroids;
[0022] Determine the target data that meets the preset conditions in each characteristic cluster, and use a preset number of target data as the centroid of the characteristic matrix;
[0023] The preset condition includes: the sum of distances to each data in the corresponding characteristic cluster is minimum.
[0024] In one embodiment, constructing the effective space of the codebook based on the centroids of the characteristic matrix includes:
[0025] Using the centroids of the angle matrix as abscissas and the centroids of the amplitude matrix as ordinates to construct the effective space of the codebook; or,
[0026] The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as abscissas and the centroids of the angle matrix as ordinates.
[0027] In a second aspect, the present invention provides a codebook determination device, comprising:
[0028] A characteristic acquisition module, used to acquire a characteristic matrix of a channel sample sequence;
[0029] A centroid determination module, used for clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0030] A space determination module, configured to construct an effective space of a codebook based on the centroids of the characteristic matrix;
[0031] A codebook determination module, configured to determine each codeword in the codebook according to the effective space;
[0032] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0033] In one embodiment, the characteristic acquisition module is specifically used to:
[0034] Inputting the channel sample sequence into a prediction network model to determine an initial characteristic matrix of the channel sample sequence;
[0035] Inputting the initial characteristic matrix into the reinforcement network model to determine a current reinforcement characteristic matrix;
[0036] Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches a preset threshold:
[0037] Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence;
[0038] Inputting the current characteristic matrix into the enhanced network model to determine a next enhanced characteristic matrix;
[0039] The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
[0040] In one embodiment, the centroid determination module is specifically used to:
[0041] Selecting a preset number of initial centroids in the characteristic matrix;
[0042] Determining a preset number of characteristic clusters according to the preset number of initial centroids;
[0043] Determine the target data that meets the preset conditions in each characteristic cluster, and use a preset number of target data as the centroid of the characteristic matrix;
[0044] The preset condition includes: the sum of distances to each data in the corresponding characteristic cluster is minimum.
[0045] In one embodiment, the space determination module is specifically used to:
[0046] Using the centroids of the angle matrix as abscissas and the centroids of the amplitude matrix as ordinates to construct the effective space of the codebook; or,
[0047] The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as abscissas and the centroids of the angle matrix as ordinates.
[0048] In a third aspect, the present invention provides an electronic device, comprising a processor and a memory storing a computer program, wherein when the processor executes the program, the steps of the codebook determination method described in the first aspect are implemented.
[0049] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the codebook determination method described in the first aspect are implemented.
[0050] The codebook determination method and device provided by the present invention can fully utilize the channel characteristics by obtaining the characteristic matrix of the channel sample sequence and determining the final codebook by performing a series of processing on the characteristic matrix, so that the generated codebook has excellent performance. In addition, the codebook determination method and device provided by the present invention can reduce the complexity of codebook calculation caused by data redundancy by clustering the obtained characteristic matrix through a clustering algorithm, and significantly improve the efficiency of codebook determination. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0052] Figure 1 It is a flowchart of the codebook determination method provided by the present invention;
[0053] Figure 2 It is a structural schematic diagram of a codebook determination device provided by the present invention;
[0054] Figure 3 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0055] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0056] Figure 1 is a flow chart of the codebook determination method provided by the present invention. Figure 1 The present invention provides a codebook determination method, which may include:
[0057] Step 110, obtaining a characteristic matrix of a channel sample sequence;
[0058] Step 120: clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0059] Step 130: constructing an effective space of a codebook based on each centroid of the characteristic matrix;
[0060] Step 140: Determine each codeword in the codebook according to the effective space;
[0061] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0062] It should be noted that the execution subject of the codebook determination method provided by the present invention can be a computer device, such as a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook or a personal digital assistant (PDA), etc.
[0063] In step 110, a real-time channel sample sequence may be obtained, and a characteristic matrix of the channel sample sequence may be determined by network model prediction; that is, an angle matrix and an amplitude matrix of the channel sample sequence may be determined by network model prediction.
[0064] In step 120, the angle matrix and the amplitude matrix may be clustered respectively by using a preset clustering algorithm to determine each centroid of the angle matrix and each centroid of the amplitude matrix.
[0065] After determining the centroids of the angle matrix and the centroids of the amplitude matrix, in step 130 , an effective space of the codebook may be constructed based on the determined centroids.
[0066] Finally, in step 140, each codeword in the codebook may be determined based on the constructed effective space. After each codeword is determined, the codewords may be combined into a required codebook.
[0067] The codebook determination method provided by the present invention can fully utilize the channel characteristics by obtaining the characteristic matrix of the channel sample sequence and determining the final codebook by performing a series of processing on the characteristic matrix, so that the generated codebook has excellent performance. In addition, the codebook determination method provided by the present invention can reduce the codebook calculation complexity caused by data redundancy by clustering the obtained characteristic matrix through a clustering algorithm, and significantly improve the codebook determination efficiency.
[0068] In one embodiment, step 110 may include:
[0069] Inputting the channel sample sequence into the prediction network model to determine the initial characteristic matrix of the channel sample sequence;
[0070] Inputting the initial characteristic matrix into the reinforcement network model to determine the current reinforcement characteristic matrix;
[0071] Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches the preset threshold:
[0072] Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence;
[0073] Inputting the current feature matrix into the reinforcement network model to determine the next reinforcement feature matrix;
[0074] The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
[0075] Optionally, the prediction network model can be used to learn the key propagation characteristics (angle and amplitude, etc.) of the channel first, and after the learning is completed, the prediction of the key propagation characteristics of the channel can be achieved. That is, for the prediction network model that has completed the learning, the channel sample sequence can be used as the input of the prediction network model, and the prediction network model will correspondingly predict the initial angle matrix and initial amplitude matrix corresponding to the channel sample sequence.
[0076] Among them, the prediction network model can be implemented by various network models, such as deep neural networks (DNN), recursive neural networks (RNN), convolutional neural networks (CNN), etc.
[0077] Optionally, the initial angle matrix and the initial amplitude matrix predicted by the prediction network model can be used as inputs of the enhanced network model, so that the enhanced network model can further enhance the accuracy of the characteristics and obtain an enhanced angle matrix and an enhanced amplitude matrix.
[0078] Among them, the reinforcement network model can be implemented by various network models, such as deep neural networks (DNN), recursive neural networks (RNN), convolutional neural networks (CNN), etc.
[0079] Strengthening the network model to enhance the accuracy of the characteristics can refer to narrowing the value range of each element in the characteristic matrix.
[0080] It should be noted that in order to ensure that the angle matrix and amplitude matrix finally obtained are as close as possible to the characteristics of the actual channel, the following steps need to be repeatedly performed:
[0081] For the angle matrix:
[0082] First, the current reinforcement angle matrix is input into the prediction network model to determine the current angle matrix;
[0083] Then the current angle matrix is input into the reinforcement network model to determine the next reinforcement angle matrix;
[0084] Repeat the above two steps until the loss function based on the prediction network model and the enhanced network model reaches a preset threshold, and then use the next enhanced angle matrix as the angle matrix of the channel sample sequence.
[0085] For the magnitude matrix:
[0086] First, the current reinforcement amplitude matrix is input into the prediction network model to determine the current amplitude matrix;
[0087] The current amplitude matrix is then input into the reinforcement network model to determine the next reinforcement amplitude matrix;
[0088] Repeat the above two steps until the loss function based on the prediction network model and the enhanced network model reaches a preset threshold, and then use the next enhanced amplitude matrix as the amplitude matrix of the channel sample sequence.
[0089] The codebook determination method provided by the present invention obtains channel characteristics by using a prediction network model and verifies the obtained characteristics by using an enhanced network model, thereby ensuring that the final output is closer to the characteristics of the actual channel, thereby making the final determined codebook have higher accuracy.
[0090] In one embodiment, step 120 may include:
[0091] Selecting a preset number of initial centroids within the feature matrix;
[0092] Determine a preset number of feature clusters according to a preset number of initial centroids;
[0093] Determine the target data that meets the preset conditions in each feature cluster, and use the preset number of target data as the centroid of the feature matrix;
[0094] The preset condition may include: the sum of distances to each data in the corresponding characteristic cluster is minimum.
[0095] For example, for an angle matrix, N uniform centroids may be selected therein as the initial centroids of the angle matrix, wherein the specific size of N may be adjusted according to actual needs, and the present invention does not specifically limit this.
[0096] Afterwards, N angle clusters can be determined based on the N initial centroids according to the nearest neighbor rule (each data is associated with the nearest centroid).
[0097] Next, target data that meets preset conditions may be selected from each angle cluster, and the target data may be used as the centroid of the corresponding angle cluster.
[0098] Finally, the N centroids (i.e., the centroids corresponding to the N angle clusters respectively) can be used as the centroid of the angle matrix.
[0099] As for the amplitude matrix, N uniform centroids can be selected from the amplitude matrix as the initial centroid of the amplitude matrix, wherein the specific size of N can be adjusted according to actual needs, and the present invention does not make any specific limitation on this.
[0100] Afterwards, N amplitude clusters can be determined based on the N initial centroids according to the nearest neighbor rule (each data is associated with the nearest centroid).
[0101] Next, target data that meets preset conditions may be selected from each amplitude cluster, and the target data may be used as the centroid of the corresponding amplitude cluster.
[0102] Finally, the N centroids (ie, the centroids corresponding to the N amplitude clusters respectively) can be used as the centroid of the amplitude matrix.
[0103] The codebook determination method provided by the present invention determines the centroids of the characteristic matrix by using the clustering algorithm, which can reduce the computational complexity caused by data redundancy, thereby improving the efficiency of determining the codebook.
[0104] In one embodiment, step 130 may include:
[0105] The effective space of the codebook is constructed by using the centroids of the angle matrix as the horizontal coordinates and the centroids of the amplitude matrix as the vertical coordinates; or,
[0106] The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as the abscissa and the centroids of the angle matrix as the ordinate.
[0107] It can be understood that the codebook determination method provided by the present invention can fully utilize the channel characteristics by constructing an effective space according to the centroid of the angle matrix and the centroid of the amplitude matrix, so that the final determined codebook can be applicable to complex wireless environments and various antenna array structures.
[0108] The following is a specific example to illustrate the codebook determination method provided by the present invention:
[0109] In this example, a single-cell downlink massive MIMO communication system is considered, which is equipped with N t A base station with k antennas communicates with k single-antenna terminals simultaneously.
[0110] Assume that the uplink and downlink channels are reciprocal, that is, represents the transpose of the matrix; t is the transmission weight. Then the codebook determination method provided by the invention includes:
[0111] Step 1: Get real-time channel samples Ω through the base station H =[H 1 ,H 2 ,……,H L ], after processing the prediction network model and the strengthening network model, the angle matrix is obtained And the amplitude matrix
[0112] Step 2.1: In order to make full use of channel information and speed up the convergence of the clustering algorithm, the angle matrix and the amplitude matrix Select N uniform centroids as the initial centroids θ =[t θ，1 , t θ，2 ,......,t θ，N ]、t g =[t g，1 , t g，2 ,......,t g，N ].
[0113] Step 2.2.1. Clustering angular features: According to the nearest neighbor rule, each data object is associated with the nearest centroid. Specifically, since the i-th element To the centroid of the l-th angle cluster t θ，l The distance is the smallest, so the data object Assigned to the lth angle cluster Cluster θ，l , the cluster selection process is expressed as in
[0114] Step 2.2.2: Clustering the amplitude features: According to the nearest neighbor rule, each data object is associated with the nearest centroid. Specifically, since the i-th element To the centroid of the lth amplitude cluster t g，l The distance is the smallest, so the data object Assigned to the lth amplitude cluster Cluster g，l , the cluster selection process is expressed as in
[0115] Step 2.3.1, obtain the centroid of the angle feature cluster: take the object with the smallest sum of distances to each data in the corresponding angle cluster as the new centroid of the corresponding angle cluster. For the lth angle cluster, its cluster centroid can be expressed as Repeat this clustering process until the objects and centroids of each cluster remain unchanged. Finally, we get N angular centroids t θ .
[0116] Step 2.3.2, obtain the centroid of the amplitude feature cluster: take the object with the smallest sum of distances to each data in the corresponding amplitude cluster as the new centroid of the corresponding amplitude cluster. For the lth amplitude cluster, its cluster centroid can be expressed as Repeat this clustering process until the objects and centroids of each cluster remain unchanged. Finally, we obtain N magnitude centroids t g .
[0117] Step 2.4: Construct the effective space based on the centroid of each angle and each amplitude The effective space The horizontal coordinate of each element in is the corresponding angle centroid, and the vertical coordinate is the corresponding amplitude centroid. For example,
[0118] Step 2.5: Based on the available space Determine each codeword in the codebook. The calculation of each codeword can be obtained by the matrix C 1 express:
[0119]
[0120] Among them, : is the number of rows, which corresponds to the number of antennas; i is the number of columns, i=1, 2, ..., N; d is the distance between antennas; j is an imaginary quantity; λ is the wavelength of the signal.
[0121] In summary, the codebook determination method provided by the present invention uses a predictive network model to obtain channel characteristics, and uses an enhanced network model to verify the obtained characteristics, which can ensure that the final characteristics are closer to the actual channel. In addition, by clustering the obtained features using a clustering algorithm, the computational complexity caused by data redundancy is reduced. The codebook design provided by the present invention is more in line with the characteristics of the channel environment, improves the accuracy of channel state information feedback, and reduces the pilot contamination of large-scale multi-antenna systems. The codebook design not only improves the performance of large-scale multi-antenna systems, but also greatly reduces the computational complexity of the system.
[0122] The codebook determination device provided by the present invention is described below. The codebook determination device described below and the codebook determination method described above can be referred to each other.
[0123] Figure 2 Schematic diagram of the structure of the codebook determination device provided by the present invention. Figure 2 The present invention provides a codebook determination device, which may include:
[0124] A characteristic acquisition module 210, used to acquire a characteristic matrix of a channel sample sequence;
[0125] A centroid determination module 220, configured to cluster the characteristic matrix to determine each centroid of the characteristic matrix;
[0126] A space determination module 230, configured to construct an effective space of a codebook based on the centroids of the characteristic matrix;
[0127] A codebook determination module 240, configured to determine each codeword in the codebook according to the effective space;
[0128] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0129] The codebook determination device provided by the present invention can fully utilize the channel characteristics by obtaining the characteristic matrix of the channel sample sequence and determining the final codebook by performing a series of processing on the characteristic matrix, so that the generated codebook has excellent performance. In addition, the codebook determination device provided by the present invention clusters the obtained characteristic matrix through a clustering algorithm, which can reduce the complexity of codebook calculation caused by data redundancy and significantly improve the efficiency of codebook determination.
[0130] In one embodiment, the characteristic acquisition module 210 is specifically used for:
[0131] Inputting the channel sample sequence into a prediction network model to determine an initial characteristic matrix of the channel sample sequence;
[0132] Inputting the initial characteristic matrix into the reinforcement network model to determine a current reinforcement characteristic matrix;
[0133] Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches a preset threshold:
[0134] Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence;
[0135] Inputting the current characteristic matrix into the enhanced network model to determine a next enhanced characteristic matrix;
[0136] The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
[0137] In one embodiment, the centroid determination module 220 is specifically used to:
[0138] Selecting a preset number of initial centroids in the characteristic matrix;
[0139] Determining a preset number of characteristic clusters according to the preset number of initial centroids;
[0140] Determine the target data that meets the preset conditions in each characteristic cluster, and use a preset number of target data as the centroid of the characteristic matrix;
[0141] The preset condition includes: the sum of distances to each data in the corresponding characteristic cluster is minimum.
[0142] In one embodiment, the space determination module 230 is specifically used to:
[0143] Using the centroids of the angle matrix as abscissas and the centroids of the amplitude matrix as ordinates to construct the effective space of the codebook; or,
[0144] The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as abscissas and the centroids of the angle matrix as ordinates.
[0145] It should be noted here that the codebook determination device provided in the embodiment of the present invention can implement all the method steps implemented in the above method embodiment and can achieve the same technical effect. The parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.
[0146] Figure 3 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 3 As shown, the electronic device may include: a processor 310, a communication interface 320, a memory 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 may call a computer program in the memory 330 to execute the steps of the codebook determination method, for example, including:
[0147] Obtaining a characteristic matrix of a channel sample sequence;
[0148] Clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0149] Based on the centroids of the characteristic matrix, constructing an effective space of a codebook;
[0150] Determine each codeword in the codebook according to the effective space;
[0151] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0152] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0153] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, the computer can perform the steps of the codebook determination method provided by the above methods, for example, including:
[0154] Obtaining a characteristic matrix of a channel sample sequence;
[0155] Clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0156] Based on the centroids of the characteristic matrix, constructing an effective space of a codebook;
[0157] Determine each codeword in the codebook according to the effective space;
[0158] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0159] On the other hand, an embodiment of the present application further provides a processor-readable storage medium, wherein the processor-readable storage medium stores a computer program, wherein the computer program is used to enable the processor to execute the steps of the methods provided in the above embodiments, for example, including:
[0160] Obtaining a characteristic matrix of a channel sample sequence;
[0161] Clustering the characteristic matrix to determine each centroid of the characteristic matrix;
[0162] Based on the centroids of the characteristic matrix, constructing an effective space of a codebook;
[0163] Determine each codeword in the codebook according to the effective space;
[0164] The characteristic matrix includes an angle matrix and an amplitude matrix.
[0165] The processor-readable storage medium can be any available medium or data storage device that can be accessed by the processor, including but not limited to magnetic storage (such as floppy disks, hard disks, magnetic tapes, magneto-optical disks (MO), etc.), optical storage (such as CD, DVD, BD, HVD, etc.), and semiconductor storage (such as ROM, EPROM, EEPROM, non-volatile memory (NANDFLASH), solid-state drive (SSD)), etc.
[0166] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0167] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A codebook determination method, characterized in that: include: Obtaining a characteristic matrix of a channel sample sequence; Clustering the characteristic matrix to determine each centroid of the characteristic matrix; Based on the centroids of the characteristic matrix, constructing an effective space of a codebook; Determine each codeword in the codebook according to the effective space; Wherein, the characteristic matrix includes an angle matrix and an amplitude matrix; The constructing the effective space of the codebook based on each centroid of the characteristic matrix includes: Using the centroids of the angle matrix as abscissas and the centroids of the amplitude matrix as ordinates to construct the effective space of the codebook; or, The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as abscissas and the centroids of the angle matrix as ordinates.
2. The codebook determination method according to claim 1, characterized in that: The acquiring a characteristic matrix of a channel sample sequence comprises: Inputting the channel sample sequence into a prediction network model to determine an initial characteristic matrix of the channel sample sequence; Inputting the initial characteristic matrix into the reinforcement network model to determine a current reinforcement characteristic matrix; Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches a preset threshold: Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence; Inputting the current characteristic matrix into the enhanced network model to determine a next enhanced characteristic matrix; The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
3. The codebook determination method according to claim 1, characterized in that: The clustering of the characteristic matrix to determine each centroid of the characteristic matrix includes: Selecting a preset number of initial centroids in the characteristic matrix; Determining a preset number of characteristic clusters according to the preset number of initial centroids; Determine the target data that meets the preset conditions in each characteristic cluster, and use a preset number of target data as the centroid of the characteristic matrix; The preset condition includes: the sum of distances to each data in the corresponding characteristic cluster is minimum.
4. A codebook determination device, characterized in that: include: A characteristic acquisition module, used to acquire a characteristic matrix of a channel sample sequence; A centroid determination module, used for clustering the characteristic matrix to determine each centroid of the characteristic matrix; A space determination module, configured to construct an effective space of a codebook based on the centroids of the characteristic matrix; A codebook determination module, configured to determine each codeword in the codebook according to the effective space; Wherein, the characteristic matrix includes an angle matrix and an amplitude matrix; The space determination module is specifically used for: Using the centroids of the angle matrix as abscissas and the centroids of the amplitude matrix as ordinates to construct the effective space of the codebook; or, The effective space of the codebook is constructed by taking the centroids of the amplitude matrix as abscissas and the centroids of the angle matrix as ordinates.
5. The codebook determination device according to claim 4, characterized in that: The characteristic acquisition module is specifically used for: Inputting the channel sample sequence into a prediction network model to determine an initial characteristic matrix of the channel sample sequence; Inputting the initial characteristic matrix into the reinforcement network model to determine a current reinforcement characteristic matrix; Repeat the following steps until the loss function based on the prediction network model and the reinforcement network model reaches a preset threshold: Inputting the current enhanced characteristic matrix into the prediction network model to determine the current characteristic matrix of the channel sample sequence; Inputting the current characteristic matrix into the enhanced network model to determine a next enhanced characteristic matrix; The next enhanced characteristic matrix is used as the characteristic matrix of the channel sample sequence.
6. The codebook determination device according to claim 4, characterized in that: The centroid determination module is specifically used for: Selecting a preset number of initial centroids in the characteristic matrix; Determining a preset number of characteristic clusters according to the preset number of initial centroids; Determine the target data that meets the preset conditions in each characteristic cluster, and use a preset number of target data as the centroid of the characteristic matrix; The preset condition includes: the sum of distances to each data in the corresponding characteristic cluster is minimum.
7. An electronic device comprising a processor and a memory storing a computer program, characterized in that: When the processor executes the computer program, the steps of the codebook determination method according to any one of claims 1 to 3 are implemented.
8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the codebook determination method according to any one of claims 1 to 3 are implemented.