Method and equipment for generating sample set

Abstract

A method and apparatus for generating a sample set, the method comprising: constructing a first basis vector space comprising O basis vector groups, each basis vector group comprising N1 * N2 spatial basis vectors, and a second basis vector space comprising Nsb frequency domain basis vectors, where O is a positive integer, N is a positive integer, N is a positive integer, N is a positive integer, N is a positive integer, N is a negative integer, and N is a positive integer; n1 is the number of antenna ports of a first dimension of a transmitting end of channel state information (CSI), N2 is the number of antenna ports of a second dimension of the transmitting end of the CSI, and Nsb represents the number of sub-bands; according to the first base vector space and the second base vector space, constructing a CSI sample set corresponding to each task in a plurality of tasks, the CSI sample set corresponding to each task is constructed based on a part of spatial domain base vectors in one base vector group in the first base vector space and a part of frequency domain base vectors in the second base vector space, and the CSI sample sets corresponding to the plurality of tasks are used for training a first model; the first model is used for training based on a CSI sample set of a target scene to obtain a second model, and the second model is adapted to the target scene.

Description

Method and device for generating sample set Technical Field

[0001] The embodiments of the present application relate to the field of communications, and in particular to a method and device for generating a sample set. Background Art

[0002] In the New Radio (NR) system, the terminal device needs to feed back Channel State Information (CSI) to the network device so that the network device can determine the precoding matrix for downlink transmission.

[0003] In some scenarios, meta-learning-based CSI feedback is being considered. However, building a meta-model requires massive amounts of CSI data from various scenarios. Collecting a large number of highly diverse CSI samples is challenging in terms of both cost and difficulty. Therefore, implementing meta-learning-based CSI feedback is an urgent issue that needs to be addressed.

[0004] Summary of the Invention

[0005] The present application provides a method and device for generating a sample set, which is conducive to reducing the cost and difficulty of obtaining CSI samples.

[0006] In a first aspect, a method for generating a sample set is provided, comprising: constructing a first basis vector space and a second basis vector space, wherein the first basis vector space includes O basis vector groups, each basis vector group includes N1*N2 spatial basis vectors, and the second basis vector space includes N sb frequency domain basis vectors, where O is a positive integer, N1 is the number of antenna ports in the first dimension of the transmitting end of the channel state information CSI, N2 is the number of antenna ports in the second dimension of the transmitting end of the CSI, and N sb Indicates the number of subbands;

[0007] Based on the first basis vector space and the second basis vector space, a CSI sample set corresponding to each task in a plurality of tasks is constructed, wherein the CSI sample set corresponding to each task is constructed based on some spatial domain basis vectors in a basis vector group in the first basis vector space and some frequency domain basis vectors in the second basis vector space. The CSI sample sets corresponding to the plurality of tasks are used to train a first model. The first model is used to train a second model based on the CSI sample set of a target scene, and the second model is adapted to the target scene.

[0008] In a second aspect, a device for generating a sample set is provided, which is used to execute the method in the first aspect or its implementation manners.

[0009] Specifically, the terminal device includes a functional module for executing the method in the above-mentioned first aspect or its various implementation modes.

[0010] In a third aspect, a device for generating a sample set is provided, comprising a processor and a memory, wherein the memory is configured to store a computer program, and the processor is configured to call and execute the computer program stored in the memory to perform the method of the first aspect or its implementations.

[0011] In a fourth aspect, a chip is provided for implementing the method in the above-mentioned first aspect or its various implementations.

[0012] Specifically, the chip includes: a processor, configured to call and run a computer program from a memory, so that a device equipped with the device executes the method of the first aspect or its various implementations described above.

[0013] In a fifth aspect, a computer-readable storage medium is provided for storing a computer program, which enables a computer to execute the method in the above-mentioned first aspect or its various implementations.

[0014] In a sixth aspect, a computer program product is provided, comprising computer program instructions, wherein the computer program instructions enable a computer to execute the method in the above-mentioned first aspect or its various implementations.

[0015] In a seventh aspect, a computer program is provided, which, when executed on a computer, enables the computer to execute the method in the above-mentioned first aspect or its various implementations.

[0016] Through the above technical solution, a first basis vector space corresponding to the spatial domain basis vectors and a second basis vector space corresponding to the frequency domain basis vectors can be constructed. Furthermore, based on the first basis vector space and the second basis vector space, a CSI sample set corresponding to each of the multiple tasks used to train the meta-model (i.e., the first model) is generated. By automatically generating a data set for meta-model training, the cost and difficulty of obtaining CSI data can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] FIG1 is a schematic diagram of a communication system architecture provided in an embodiment of the present application.

[0018] FIG2 is a schematic diagram of a neuron structure.

[0019] FIG3 is a schematic diagram of a neural network provided by the present application.

[0020] FIG4 is a schematic diagram of a convolutional neural network provided in this application.

[0021] FIG5 is a schematic diagram of an LSTM unit provided in this application.

[0022] FIG6 is a schematic diagram of a channel information feedback provided by the present application.

[0023] FIG7 is a schematic diagram of another channel information feedback provided in this application.

[0024] FIG8 is a schematic flowchart of a method for generating a sample set according to an embodiment of the present application.

[0025] FIG9 is a schematic block diagram of a device for generating a sample set according to an embodiment of the present application.

[0026] FIG10 is a schematic block diagram of a communication device provided according to an embodiment of the present application.

[0027] FIG11 is a schematic block diagram of a chip provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] The following will describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of the embodiments. With respect to the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0029] The technical solutions of the embodiments of the present application can be applied to various communication systems, such as: Global System of Mobile communication (GSM) system, Code Division Multiple Access (CDMA) system, Wideband Code Division Multiple Access (WCDMA) system, General Packet Radio Service (GPRS), Long Term Evolution (LTE) system, Advanced Long Term Evolution (LTE-A) system, New Radio (NR) system, NR system evolution system, LTE on unlicensed spectrum (LTE-U) system, NR on unlicensed spectrum (NR-U) system, Non-Terrestrial Networks (NTN) system, Universal Mobile Telecommunication System (UMTS), Wireless Local Area Networks (WLAN), Wireless Fidelity (Wireless Fidelity) system. Fidelity, WiFi), fifth-generation communication (5th-Generation, 5G) system or other communication systems, etc.

[0030] Generally speaking, traditional communication systems support a limited number of connections and are easy to implement. However, with the development of communication technology, mobile communication systems will not only support traditional communications, but will also support, for example, device-to-device (D2D) communication, machine-to-machine (M2M) communication, machine-type communication (MTC), vehicle-to-vehicle (V2V) communication, or vehicle-to-everything (V2X) communication, etc. The embodiments of the present application can also be applied to these communication systems.

[0031] Optionally, the communication system in the embodiment of the present application can be applied to a carrier aggregation (CA) scenario, a dual connectivity (DC) scenario, or a standalone (SA) networking scenario.

[0032] Optionally, the communication system in the embodiment of the present application can be applied to an unlicensed spectrum, where the unlicensed spectrum can also be considered as a shared spectrum; or, the communication system in the embodiment of the present application can also be applied to an authorized spectrum, where the authorized spectrum can also be considered as an unshared spectrum.

[0033] The embodiments of the present application describe various embodiments in conjunction with network devices and terminal devices, wherein the terminal device may also be referred to as user equipment (UE), access terminal, user unit, user station, mobile station, mobile station, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent or user device, etc.

[0034] The terminal device can be a station (STA) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication capabilities, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a wearable device, a terminal device in a next-generation communication system such as an NR network, or a terminal device in a future evolved Public Land Mobile Network (PLMN) network, etc.

[0035] In an embodiment of the present application, the terminal device can be deployed on land, including indoors or outdoors, handheld, wearable or vehicle-mounted; it can also be deployed on the water surface (such as ships, etc.); it can also be deployed in the air (such as airplanes, balloons and satellites, etc.).

[0036] In an embodiment of the present application, the terminal device may be a mobile phone, a tablet computer, a computer with wireless transceiver function, a virtual reality (VR) terminal device, an augmented reality (AR) terminal device, a wireless terminal device in industrial control, a wireless terminal device in self-driving, a wireless terminal device in remote medical, a wireless terminal device in a smart grid, a wireless terminal device in transportation safety, a wireless terminal device in a smart city, or a wireless terminal device in a smart home, etc.

[0037] As an example and not a limitation, in the embodiment of the present application, the terminal device may also be a wearable device. Wearable devices may also be called wearable smart devices, which are a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing, and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are fully functional, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0038] In an embodiment of the present application, the network device may be a device for communicating with a mobile device. The network device may be an access point (AP) in WLAN, a base station (BTS) in GSM or CDMA, a base station (NodeB, NB) in WCDMA, an evolved base station (eNB or eNodeB) in LTE, or a relay station or access point, or a vehicle-mounted device, a wearable device, and a network device (gNB) in an NR network, or a network device in a future evolved PLMN network or a network device in an NTN network, etc.

[0039] As an example and not a limitation, in an embodiment of the present application, the network device may have a mobile feature, for example, the network device may be a mobile device. Alternatively, the network device may be a satellite or a balloon station. For example, the satellite may be a low earth orbit (LEO) satellite, a medium earth orbit (MEO) satellite, a geostationary earth orbit (GEO) satellite, a high elliptical orbit (HEO) satellite, etc. Optionally, the network device may also be a base station set up in a location such as land or water.

[0040] In an embodiment of the present application, the network device can provide services for a cell, and the terminal device communicates with the network device through the transmission resources used by the cell (for example, frequency domain resources, or spectrum resources). The cell can be a cell corresponding to the network device (for example, a base station). The cell can belong to a macro base station or a base station corresponding to a small cell. The small cells here may include: metro cells, micro cells, pico cells, femto cells, etc. These small cells have the characteristics of small coverage and low transmission power, and are suitable for providing high-speed data transmission services.

[0041] For example, a communication system 100 used in an embodiment of the present application is shown in FIG1 . The communication system 100 may include a network device 110, which may be a device that communicates with a terminal device 120 (or a communication terminal or terminal). The network device 110 may provide communication coverage for a specific geographic area and may communicate with terminal devices within the coverage area.

[0042] FIG1 exemplarily shows a network device and two terminal devices. Optionally, the communication system 100 may include multiple network devices and each network device may include another number of terminal devices within its coverage area, which is not limited in the embodiments of the present application.

[0043] Optionally, the communication system 100 may further include other network entities such as a network controller and a mobility management entity, which is not limited in the embodiment of the present application.

[0044] It should be understood that in the embodiments of the present application, a device having a communication function in a network / system may be referred to as a communication device. Taking the communication system 100 shown in FIG1 as an example, the communication device may include a network device 110 and a terminal device 120 having a communication function. The network device 110 and the terminal device 120 may be the specific devices described above and will not be described in detail here. The communication device may also include other devices in the communication system 100, such as a network controller, a mobility management entity, and other network entities, which are not limited in the embodiments of the present application.

[0045] It should be understood that the terms "system" and "network" are often used interchangeably herein. The term "and / or" is simply a description of an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " generally indicates that the related objects are in an "or" relationship.

[0046] It should be understood that the "indication" mentioned in the embodiments of this application can be a direct indication, an indirect indication, or an indication of an association. For example, "A indicates B" can mean that A directly indicates B, for example, B can be obtained through A; it can also mean that A indirectly indicates B, for example, A indicates C, and B can be obtained through C; it can also mean that there is an association between A and B.

[0047] In the description of the embodiments of the present application, the term "corresponding" may indicate a direct or indirect correspondence between the two, or an association relationship between the two, or a relationship between indication and being indicated, configuration and being configured, etc.

[0048] In the embodiments of the present application, "pre-defined" may be implemented by pre-storing corresponding codes, tables, or other methods that can be used to indicate relevant information in a device (e.g., a terminal device or a network device). The present application does not limit the specific implementation method. For example, pre-defined may refer to information defined in a protocol.

[0049] In the embodiments of the present application, the "protocol" may refer to a standard protocol in the communication field, for example, it may include an LTE protocol, an NR protocol, and related protocols used in future communication systems, and this application does not limit this.

[0050] To facilitate understanding of the technical solutions of the embodiments of the present application, the technical solutions of the present application are described in detail below through specific embodiments. The following related technologies can be combined with the technical solutions of the embodiments of the present application as optional solutions, and they all fall within the scope of protection of the embodiments of the present application. The embodiments of the present application include at least part of the following contents.

[0051] To facilitate a better understanding of the embodiments of the present application, the channel information feedback related to the present application is explained.

[0052] In the NR system, for the Channel State Information (CSI) feedback scheme, the codebook-based eigenvector feedback is mainly used to enable the base station to obtain downlink CSI. Specifically, the base station sends a downlink Channel State Information Reference Signal (CSI-RS) to the terminal device. The terminal device uses the CSI-RS to estimate the CSI of the downlink channel and performs eigenvalue decomposition on the estimated downlink channel to obtain the eigenvector corresponding to the downlink channel. The NR system provides three codebook design schemes: Type 1 (Type I), Type 2 (Type II), and enhanced Type 2 (eType II).

[0053] For the codebook feedback of eType II, the precoding matrix to be fed back is expressed as W∈C Nt×Nsb , where C represents the complex space, Nt represents the number of transmit antenna ports, Nsb represents the number of subbands, and each column of the matrix W represents the precoding vector shared by multiple subcarriers on each subband. The eTypeII codebook first considers compressing W to be represented as W'=W1W2W f , where the diagonal block matrix W1 = [B, 0; 0, B] ∈ C Nt×2L , B∈C Nt / 2×L All columns in are a set of L orthogonal basis vectors selected in the DFT vector space, W f ∈C M×Nsb All rows in are also a set of M orthogonal basis vectors selected in the DFT vector space, W2∈C 2L×M is the projection coefficient of the precoding matrix W after projecting on the two sets of basis vectors. The eTypeII codebook considers feedback of the following information:

[0054] The indices of the L basis vectors selected in the DFT vector space to form the orthogonal basis matrix B;

[0055] The orthogonal basis matrix W is selected in the DFT vector space f The indices of the M basis vectors;

[0056] The coefficients in the projection coefficient matrix W2.

[0057] The receiving end uses W'=W1W2W according to the above feedback information f Recover the precoding matrix.

[0058] To facilitate a better understanding of the embodiments of the present application, the neural network related to the present application is described.

[0059] A neural network is a computational model consisting of multiple interconnected neuron nodes, where the connections between nodes represent weighted values ​​from input signals to output signals, called weights. Each node performs a weighted summation (SUM) on different input signals and outputs them through a specific activation function (f). Figure 2 is a schematic diagram of a neuron structure, where a1, a2, …, an represent input signals, w1, w2, …, wn represent weights, f represents the activation function, and t represents the output.

[0060] A simple neural network, shown in Figure 3, consists of an input layer, hidden layers, and an output layer. By using different connections, weights, and activation functions among multiple neurons, different outputs can be generated, thereby fitting the mapping relationship from input to output. Each node in the previous level is connected to all nodes in the next level. This neural network is a fully connected neural network, also known as a deep neural network (DNN).

[0061] Deep learning utilizes deep neural networks with multiple hidden layers, significantly improving the network's ability to learn features and fitting complex, nonlinear mappings from input to output. Consequently, it has found widespread application in speech and image processing. In addition to deep neural networks, deep learning also includes other commonly used basic structures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) for different tasks.

[0062] The basic structure of a convolutional neural network includes an input layer, multiple convolutional layers, multiple pooling layers, a fully connected layer, and an output layer, as shown in Figure 4. Each neuron in the convolution kernel of the convolutional layer is locally connected to its input, and the introduction of the pooling layer extracts the local maximum or average features of a certain layer, effectively reducing the network parameters and mining local features, enabling the convolutional neural network to converge quickly and achieve excellent performance.

[0063] RNNs are neural networks that model sequential data and have achieved remarkable success in natural language processing applications such as machine translation and speech recognition. Specifically, the network memorizes information from past moments and uses it in the calculation of current outputs. This means that nodes in the hidden layers are no longer disconnected but connected, and the input to a hidden layer includes not only the input layer but also the output of the previous hidden layer. Common RNN structures include long short-term memory (LSTM) and gated recurrent unit (GRU). Figure 5 shows a basic LSTM cell structure, which can include a tanh activation function. Unlike RNNs, which only consider the most recent state, the LSTM cell state determines which states should be retained and which should be forgotten, addressing the shortcomings of traditional RNNs in long-term memory.

[0064] To facilitate a better understanding of the embodiments of the present application, the CSI feedback method based on artificial intelligence (AI) related to the present application is described.

[0065] Given the tremendous success of AI technology in areas such as computer vision and natural language processing, the communications field has begun exploring the use of deep learning, such as deep learning, to address technical challenges that are difficult to address with traditional communications methods. The neural network architecture commonly used in deep learning is nonlinear and data-driven. It can extract features from actual channel matrix data and, at the base station, restore the channel matrix information compressed and fed back by the terminal as closely as possible. This ensures the restoration of channel information while also enabling the reduction of CSI feedback overhead on the terminal side.

[0066] Deep learning-based CSI feedback treats channel information as an image to be compressed, uses a deep learning autoencoder to compress and feed back the channel information, and reconstructs the compressed channel image at the transmitter, which can preserve the channel information to a greater extent, as shown in Figure 6.

[0067] A typical channel information feedback system is shown in Figure 7. The entire feedback system consists of an encoder and a decoder, deployed at the transmitter and receiver, respectively. After obtaining channel information through channel estimation, the transmitter compresses and encodes the channel information matrix using the encoder's neural network. The compressed bitstream is then fed back to the receiver via an air interface feedback link. The receiver uses the decoder to recover the channel information based on the feedback bitstream to obtain complete feedback channel information. The encoder shown in Figure 7 employs a stack of multiple fully connected layers, while the decoder employs convolutional layers and a residual architecture. While maintaining the same encoding and decoding framework, the network model structures within the encoder and decoder can be flexibly designed.

[0068] To facilitate a better understanding of the embodiments of the present application, meta-learning related to the present application is explained.

[0069] The key idea behind meta-learning is to "let machines learn to learn." For example, a model is first trained using a large amount of data from various scenarios and categories using a meta-learning algorithm (including but not limited to MAML and Reptile) with randomly initialized weights as a starting point, resulting in a meta-model that has learned a large amount of basic knowledge. Because this meta-model is trained on a large amount of scenario data (the data used to train the meta-model can be divided into different scenarios, referred to as different "tasks"), a model adapted to the target scenario can be trained using a small amount of data from the target scenario.

[0070] Therefore, if CSI feedback is performed based on meta-learning, it is necessary to collect a large amount of data from different scenarios to support the construction of the meta-model. Collecting such a large amount of channel data with high diversity is challenging in terms of both collection cost and difficulty.

[0071] FIG8 is a schematic flow chart of a method 200 for generating a sample set according to an embodiment of the present application. As shown in FIG8 , the method 200 includes the following contents:

[0072] S210: Construct a first basis vector space and a second basis vector space, wherein the first basis vector space includes O basis vector groups, each basis vector group includes N1*N2 spatial basis vectors, and the second basis vector space includes N sb frequency domain basis direction, O is a positive integer;

[0073] S220. Construct a CSI sample set corresponding to each of the multiple tasks based on the first basis vector space and the second basis vector space, wherein the CSI sample set corresponding to each task is constructed based on some spatial domain basis vectors in a basis vector group in the first basis vector space and some frequency domain basis vectors in the second basis vector space.

[0074] It should be understood that in the embodiment of the present application, the method 200 can be executed by a terminal device, or it can also be executed by a network device, and the present application does not limit this.

[0075] In some embodiments, N sb Indicates the number of subbands.

[0076] In some embodiments, the transmitter of the CSI has a two-dimensional array antenna, N1 is the number of antenna ports in a first dimension of the two-dimensional array antenna, and N2 is the number of antenna ports in a second dimension of the two-dimensional array antenna.

[0077] It should be understood that this application only uses the example of the antenna at the transmitting end of the CSI being a two-dimensional array antenna for illustration. When the antenna at the transmitting end of the CSI is in other forms, the dimensions of the spatial basis vectors included in the basis vector group can be adjusted for applicability, and this application does not limit this.

[0078] In some embodiments, the CSI sample sets corresponding to the multiple tasks are used to train a first model, i.e., the first model is a meta-model, i.e., a starting point training model. Furthermore, the first model can be trained based on the CSI sample set of the target scene to obtain a second model adapted to the target scene.

[0079] In this embodiment of the present application, since the first model is trained based on a large number of task CSI sample sets, only a small number of target scenario CSI sample sets are needed to train a second model that can quickly adapt to the target scenario. Furthermore, the second model can be used to compress and feed back the CSI data for the target scenario, which helps reduce CSI feedback overhead.

[0080] In some embodiments, the multiple tasks are d tasks, and the CSI sample set corresponding to each task may include k CSI samples. Optionally, the number of CSI samples included in the CSI sample set corresponding to each task may be the same, or may be different.

[0081] Optionally, the dimension of each CSI sample can be N t ×N sb , that is, the CSI sample is represented by W∈C Nt×Nsb , where C represents the complex space, N t =2N1N2 represents the number of transmitting antenna ports.

[0082] In some embodiments, O is a positive integer greater than 1, and increasing the number of spatial basis vectors by a factor of O is beneficial to increasing the accuracy of CSI samples in the spatial domain.

[0083] In some embodiments, the matrix composed of the N1*N2 spatial basis vectors in each of the O basis vector groups is a unitary matrix, that is, the spatial basis vectors in the same basis vector group are orthogonal, and the N sb The matrix composed of frequency domain basis vectors is a unitary matrix, that is, the frequency domain basis vectors in the second basis vector space are orthogonal.

[0084] That is to say, the first basis vector space and the second basis vector space in the embodiment of the present application are complete. Therefore, the data set for training the meta-model generated based on the first basis vector space and the second basis vector space is also complete.

[0085] In an embodiment of the present application, the matrix formed by all spatial basis vectors in each basis vector group in the first basis vector space is a unitary matrix. Therefore, any vector of the same dimension as the spatial basis vector can be linearly represented using all spatial basis vectors in the basis vector group.

[0086] Similarly, the matrix formed by all frequency domain basis vectors in the second basis vector space is a unitary matrix. Therefore, any vector with the same dimension as the frequency domain basis vector can be linearly represented using all frequency domain basis vectors in the second basis vector space.

[0087] It should be understood that the embodiments of the present application are not limited to the specific implementation of constructing the O basis vector groups in the first basis vector space and constructing the second basis vector space. Any method for constructing a unitary matrix can be used as the O basis vector groups in the first basis vector space and the method for constructing the second basis vector space. This application does not limit this.

[0088] The following describes a method for constructing the first basis vector space and the second basis vector space with reference to specific examples, but the present application is not limited thereto.

[0089] In some embodiments of the present application, constructing the first basis vector space includes:

[0090] Loop through the following steps O times to obtain the spatial basis vectors of each of the O basis vector groups:

[0091] Random sampling based on the first distribution obtains the first matrix (denoted as X h ), and randomly sample based on the second distribution to obtain the second matrix (denoted as X v ), wherein the first matrix is ​​a matrix of N1*N1 dimensions (ie X h ∈C N1×N1 ), the second matrix is ​​a matrix of N2*N2 dimensions (ie X v ∈C N2×N2 );

[0092] Determine the orthogonal matrix corresponding to the first matrix (denoted as U h ) and the orthogonal matrix corresponding to the second matrix (denoted as U v );

[0093] Perform the Kronecker product of the orthogonal matrix corresponding to the first matrix and the orthogonal matrix corresponding to the second matrix to obtain the target matrix (denoted as S), that is, in, represents the Kronecker product;

[0094] A column of the target matrix is ​​used as a spatial basis vector in a basis vector group in the first basis vector space.

[0095] Optionally, the first distribution may include but is not limited to complex Gaussian distribution and complex uniform distribution.

[0096] Optionally, the second distribution may include but is not limited to complex Gaussian distribution and complex uniform distribution.

[0097] Optionally, the first distribution and the second distribution may be the same, or may be different.

[0098] In some embodiments, the first matrix is ​​obtained by random sampling based on a first distribution, so the elements in the first matrix obey the first distribution. Similarly, the second matrix is ​​obtained by random sampling based on a second distribution, so the elements in the second matrix obey the second distribution.

[0099] In some embodiments, the orthogonal matrix corresponding to the first matrix may be an orthogonal matrix determined based on the first matrix, and the orthogonal matrix corresponding to the second matrix may refer to an orthogonal matrix determined based on the second matrix. This application does not limit the specific method for determining the orthogonal matrix corresponding to the first matrix and the orthogonal matrix corresponding to the second matrix. For example, it can be obtained by singular value decomposition, or it can be obtained by Schmidt orthogonalization, etc.

[0100] In some embodiments, determining an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix includes:

[0101] Performing SVD on the first matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the first matrix;

[0102] Perform SVD on the second matrix, and use the obtained left singular matrix as the orthogonal matrix corresponding to the second matrix.

[0103] In some other embodiments, determining an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix includes:

[0104] Performing Schmidt orthogonalization on the first matrix, and using the obtained orthogonal matrix as the orthogonal matrix corresponding to the first matrix;

[0105] Perform Schmidt orthogonalization on the second matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the second matrix.

[0106] In some embodiments of the present application, constructing the second basis vector space includes:

[0107] Based on the third distribution, random sampling is performed to obtain the third matrix (denoted as X f ), the third matrix is ​​N sb *Nsb A matrix of dimensions (i.e., X f ∈C Nsb×Nsb );

[0108] Determine the orthogonal matrix corresponding to the third matrix (denoted as U f );

[0109] The orthogonal matrix corresponding to the third matrix (i.e., U f ) as a frequency domain basis vector in the second basis vector space.

[0110] Optionally, the third distribution may include but is not limited to complex Gaussian distribution and complex uniform distribution.

[0111] In some embodiments, the third matrix is ​​obtained by random sampling based on a third distribution. Therefore, elements in the third matrix obey the third distribution.

[0112] In some embodiments, the orthogonal matrix corresponding to the third matrix may be an orthogonal matrix determined based on the third matrix. This application does not limit the specific method for determining the orthogonal matrix corresponding to the third matrix. For example, it can be obtained by SVD, or by Schmidt orthogonalization.

[0113] In some embodiments, determining the orthogonal matrix corresponding to the third matrix includes:

[0114] Performing SVD on the third matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the third matrix; or

[0115] Perform Schmidt orthogonalization on the third matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the third matrix.

[0116] In some embodiments of the present application, the S220 may include:

[0117] Loop through the following steps d times to obtain a CSI sample set corresponding to each of the multiple tasks, where d is the number of tasks:

[0118] Selecting a basis vector group from the O basis vector groups in the first basis vector space;

[0119] According to L in the one basis vector group task spatial basis vectors and the second basis vector space M task frequency domain basis vectors to construct a CSI sample set corresponding to one of the multiple tasks, where L task , L task is a positive integer greater than 1.

[0120] It should be understood that for different tasks in the multiple tasks, when constructing the corresponding CSI sample set, the parameters L used task , M task They can be the same or different, and this application does not limit this.

[0121] In some embodiments, the one basis vector group is randomly selected from the O basis vector groups.

[0122] In some embodiments, the L task Less than N1*N2, the M task Less than N sb .

[0123] That is, L task A spatial basis vector is part of the spatial basis vectors in a basis vector group, M task The frequency domain basis vectors are partial frequency domain basis vectors in the second basis vector space.

[0124] In some embodiments, the L task The spatial basis vectors are randomly selected from the one basis vector group.

[0125] In some embodiments, the M task The frequency domain basis vectors are N in the second basis vector space. sb The frequency domain basis vectors are randomly selected.

[0126] For example, when constructing a CSI sample set corresponding to the first task among multiple tasks, a basis vector group (denoted as the first basis vector group) is randomly selected from the O basis vector groups in the first basis vector space, and N basis vector groups are randomly selected from the N basis vector groups in the second basis vector space. sb Frequency domain basis vectors are randomly selected M task frequency domain basis vectors, further based on some of the spatial domain basis vectors in the first basis vector group and the M task The CSI sample set corresponding to the first task is constructed by some frequency domain basis vectors in the frequency domain basis vectors. task spatial basis vectors) and the M task The frequency domain basis vectors can be considered as a task basis vector set corresponding to the first task (denoted as the first task basis vector set), that is, the CSI sample set corresponding to the first task is constructed based on the basis vectors in the first task basis vector set.

[0127] For another example, when constructing a CSI sample set corresponding to the second task among multiple tasks, a basis vector group (denoted as the second basis vector group) is randomly selected from the O basis vector groups in the first basis vector space, and N basis vector groups in the second basis vector space are randomly selected. sb Frequency domain basis vectors are randomly selected M taskfrequency domain basis vectors, further based on some of the spatial domain basis vectors in the second basis vector group and the M task The CSI sample set corresponding to the second task is constructed by some frequency domain basis vectors in the frequency domain basis vectors. task spatial basis vectors) and the M task The frequency domain basis vectors can be considered as the task basis vector set corresponding to the second task (denoted as the second task basis vector set), that is, the CSI sample set corresponding to the second task is constructed based on the basis vectors in the second task basis vector set.

[0128] The first basis vector space and the second basis vector space can be considered as the full set of orthogonal basis vectors. Therefore, in an embodiment of the present application, the orthogonal basis vectors (i.e., the task basis vector set) used to construct the CSI sample sets corresponding to different tasks are different subsets of the full set of orthogonal basis vectors. For example, the CSI sample set corresponding to the first task is generated based on the basis vectors in the first task basis vector set, and the CSI sample set corresponding to the second task is constructed based on the basis vectors in the second task basis vector set, which is conducive to achieving diversity of CSI samples between different tasks.

[0129] In some embodiments, the basis vectors in the task basis vector set corresponding to a task can be considered to have similar or consistent characteristics. For example, L randomly selected in the first basis vector group task The spatial basis vectors can be considered to have similar or consistent characteristics, and the L randomly selected in the second basis vector group task The spatial basis vectors can be considered to have similar or consistent characteristics. In the second basis vector space, M randomly selected task The frequency domain basis vectors can be considered to have similar or consistent characteristics. Therefore, a CSI sample set for a specific scenario (ie, a specific task) can be generated based on the basis vectors in the task basis vector set.

[0130] In some embodiments, the method according to L in the one basis vector group task spatial basis vectors and the second basis vector space M task frequency domain basis vectors, constructing a CSI sample set corresponding to one of the multiple tasks, including:

[0131] The following steps are executed k times in a loop to obtain a CSI sample set corresponding to a task, including k CSI samples, where k is greater than 1:

[0132] In the L task Randomly select L spatial basis vectors from the spatial basis vectors, and arrange the L spatial basis vectors in columns to obtain a matrix B, where L is a positive integer;

[0133] Construct a diagonal block matrix W1 = [B, 0; 0, B] based on the matrix B;

[0134] In the M task Randomly select M frequency domain basis vectors from the frequency domain basis vectors, and arrange the M frequency domain basis vectors in rows to obtain a matrix W f , where M is a positive integer;

[0135] According to the diagonal block matrix W1, the matrix W f And the random number matrix W2 generates a CSI sample in the CSI sample set corresponding to a task.

[0136] In some embodiments, L<L task , M<M task .

[0137] Therefore, in an embodiment of the present application, when constructing different CSI samples in a CSI sample set corresponding to a task, they can be based on different subsets of the task basis vector set corresponding to the task, which is conducive to achieving diversity of different samples in the same task.

[0138] In some embodiments, the random number matrix W2 is obtained by sampling based on the fourth distribution, that is, the elements in W2 obey the fourth distribution.

[0139] Since an infinite number of random samplings can be performed based on the fourth distribution, sufficient and large-scale CSI samples can be constructed.

[0140] In the embodiment of the present application, the first basis vector space and the second basis vector space can be used to provide spatial and frequency domain characteristics, and the random number matrix can be used to expand the scale of the CSI sample set. Therefore, the method for generating a sample set according to the embodiment of the present application can take into account the spatial and frequency domain characteristics and scale of the generated CSI samples.

[0141] Optionally, the fourth distribution may include but is not limited to complex Gaussian distribution and complex uniform distribution.

[0142] In some embodiments, the matrix W is based on the diagonal block matrix W1. f And the random number matrix W2, generate a CSI sample in the CSI sample set corresponding to a task, including:

[0143] The diagonal block matrix W1, matrix W f Multiply it by the random number matrix W2 to obtain the first CSI sample;

[0144] Normalize each column in the matrix of the first CSI samples to obtain target CSI samples.

[0145] In some embodiments, the first CSI sample W′=W1W2Wf =[w1,...,w Nsb ].

[0146] Optionally, the normalization process may include but is not limited to a two-norm process.

[0147] For example, target CSI sample W = [w1 / norm(w1),...,w Nsb / norm(w Nsb )], where norm() represents the two-norm.

[0148] In a specific embodiment, a CSI sample set corresponding to each of the d tasks may be generated based on the following steps.

[0149] Step a, randomly selecting a basis vector group from the O basis vector groups in the first basis vector space, denoted as basis vector group Q;

[0150] Step b, randomly select L from the basis vector group Q task Spatial basis vectors, and get the basis vector group Q';

[0151] Step c, from the second basis vector space N sb Randomly select M frequency domain basis vectors task Frequency domain basis vectors, and obtain the basis vector group P;

[0152] The basis vector group Q' and the basis vector group P can be considered as a set of task basis vectors corresponding to the current task.

[0153] Step d, randomly select L from the basis vector group Q' <L task Basis vectors are arranged in columns to form a matrix B, and a diagonal block matrix W1 = [B, 0; 0, B] is constructed based on the matrix B, where B∈C N1N2×L , W1∈C 2N1N2×2L .

[0154] Step e, randomly select M from the basis vector set P <M task basis vectors, arranged in rows to form a matrix W f , where W f ∈C M×Nsb .

[0155] Step f, construct random number matrix W2∈C 2L×M , each element in the random number matrix W2 obeys a fourth distribution, such as a complex Gaussian distribution or a complex uniform distribution;

[0156] Step g, according to the diagonal block matrix W1, the matrix W f Sample and random number matrix W2 generate the first CSI sample W'=W1W2W f =[w1,...,wNsb ].

[0157] Furthermore, each column of the first CSI sample is normalized to obtain a target CSI sample. For example, each column of the first CSI sample is normalized using a two-norm to obtain a target CSI sample W = [w1 / norm(w1), ..., w Nsb / norm(w Nsb )], where norm() represents the two-norm;

[0158] Step h: Determine whether the k CSI samples in the CSI sample set corresponding to the current task have been constructed. If so, return to step a to generate a CSI sample in the CSI sample set corresponding to the next task. If not, return to step d to construct the next CSI sample in the CSI sample set corresponding to the current task.

[0159] In step i, it is determined whether the CSI sample sets corresponding to the d tasks have been constructed. If so, the process ends; if not, the process returns to step a to construct the CSI sample set corresponding to the next task.

[0160] When the CSI sample sets corresponding to the d tasks are constructed, the CSI sample set T = {T1, ..., T d}, where the CSI sample set T i Denotes the CSI sample set corresponding to task i, i = 1, ..., d. As described above, the first basis vector space and the second basis vector space are complete. Therefore, the CSI sample set T = {T1, ..., T d} is also complete.

[0161] Furthermore, based on the CSI sample set T={T1,...,T d A first model, i.e., a starting training model, can be trained. Furthermore, the first model can be trained based on the CSI sample set of the target scenario to obtain a second model adapted to the target scenario. Because the first model is trained based on a large number of CSI sample sets for tasks, only a small number of CSI sample sets for the target scenario are needed to train a second model that quickly adapts to the target scenario. Furthermore, the CSI data of the target scenario can be compressed and fed back based on the second model, which helps reduce CSI feedback overhead.

[0162] It should be understood that in an embodiment of the present application, the device for constructing the CSI sample set and the device for training the first model may be the same device, or may be different devices. When the two are different devices, the device for constructing the CSI sample set may send the constructed CSI sample set to the device for training the first model, so that the device trains the first model based on the CSI sample set.

[0163] In summary, in an embodiment of the present application, a communication device (e.g., a terminal device or a network device) can construct a first basis vector space corresponding to spatial basis vectors and a second basis vector space corresponding to frequency domain basis vectors, wherein the first basis vector space includes O basis vector groups, the matrix composed of N1*N2 spatial basis vectors in each basis vector group is a unitary matrix, and the matrix composed of frequency domain basis vectors in the second basis vector space is a unitary matrix. That is, the first basis vector space and the second basis vector space are complete, and therefore, the data set for training the meta-model generated based on the first basis vector space and the second basis vector space is also complete.

[0164] Furthermore, using different task basis vector sets to construct CSI sample sets corresponding to different tasks is conducive to achieving diversity of CSI samples between different tasks.

[0165] Furthermore, different subsets of the task basis vector set are used to construct different CSI samples in the CSI sample set corresponding to the task, which is conducive to achieving diversity of different samples in the same task.

[0166] Furthermore, since the random number matrix can perform infinite random sampling, the first basis vector space can provide the spatial domain characteristics of the CSI data, and the second basis vector space can provide the frequency domain characteristics of the CSI data. Therefore, constructing the CSI sample set corresponding to each task based on the first basis vector space, the second basis vector space, and the random number matrix can ensure the spatial and frequency domain characteristics and scale of the generated CSI samples.

[0167] The above, in conjunction with Figure 8, describes in detail the method embodiment of the present application. The following, in conjunction with Figures 9 to 11, describes in detail the device embodiment of the present application. It should be understood that the device embodiment and the method embodiment correspond to each other, and similar descriptions can refer to the method embodiment.

[0168] FIG9 shows a schematic block diagram of a device 400 for generating a sample set according to an embodiment of the present application. As shown in FIG4 , the device 400 includes:

[0169] The processing unit 410 is configured to construct a first basis vector space and a second basis vector space, wherein the first basis vector space includes O basis vector groups, each basis vector group includes N1*N2 spatial basis vectors, and the second basis vector space includes N sbfrequency domain basis vectors, where O is a positive integer, N1 is the number of antenna ports in the first dimension of the transmitting end of the channel state information CSI, N2 is the number of antenna ports in the second dimension of the transmitting end of the CSI, and N sb represents the number of subbands; and

[0170] Based on the first basis vector space and the second basis vector space, a CSI sample set corresponding to each task in a plurality of tasks is constructed, wherein the CSI sample set corresponding to each task is constructed based on some spatial domain basis vectors in a basis vector group in the first basis vector space and some frequency domain basis vectors in the second basis vector space. The CSI sample sets corresponding to the plurality of tasks are used to train a first model. The first model is used to train a second model based on the CSI sample set of a target scene, and the second model is adapted to the target scene.

[0171] In some embodiments, the matrix composed of N1*N2 spatial basis vectors in each basis vector group of the O basis vector groups is a unitary matrix, and the N sb The matrix composed of frequency domain basis vectors is a unitary matrix.

[0172] In some embodiments, the processing unit 410 is further configured to:

[0173] Loop through the following steps O times to obtain the spatial basis vectors of each of the O basis vector groups:

[0174] Performing random sampling based on a first distribution to obtain a first matrix, and performing random sampling based on a second distribution to obtain a second matrix, wherein the first matrix is ​​a matrix of N1*N1 dimension, and the second matrix is ​​a matrix of N2*N2 dimension;

[0175] Determine an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix;

[0176] Performing a Kronecker product on the orthogonal matrix corresponding to the first matrix and the orthogonal matrix corresponding to the second matrix to obtain a target matrix;

[0177] A column of the target matrix is ​​used as a spatial basis vector in a basis vector group in the first basis vector space.

[0178] In some embodiments, the processing unit 410 is further configured to:

[0179] Performing singular value decomposition (SVD) on the first matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the first matrix;

[0180] Perform SVD on the second matrix, and use the obtained left singular matrix as the orthogonal matrix corresponding to the second matrix.

[0181] In some embodiments, the processing unit 410 is further configured to:

[0182] Performing Schmidt orthogonalization on the first matrix, and using the obtained orthogonal matrix as the orthogonal matrix corresponding to the first matrix;

[0183] Perform Schmidt orthogonalization on the second matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the second matrix.

[0184] In some embodiments, the first distribution is a complex Gaussian distribution or a complex uniform distribution;

[0185] The second distribution is a complex Gaussian distribution or a complex uniform distribution.

[0186] In some embodiments, the processing unit 410 is further configured to:

[0187] Random sampling is performed based on the third distribution to obtain a third matrix, and the third matrix is ​​N sb *N sb Matrix of dimensions;

[0188] Determine an orthogonal matrix corresponding to the third matrix;

[0189] A column of the orthogonal matrix corresponding to the third matrix is ​​used as a frequency domain basis vector in the second basis vector space.

[0190] In some embodiments, the processing unit 410 is further configured to:

[0191] Performing SVD on the third matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the third matrix; or

[0192] Perform Schmidt orthogonalization on the third matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the third matrix.

[0193] In some embodiments, the third distribution is a complex Gaussian distribution or a complex uniform distribution.

[0194] In some embodiments, the processing unit 410 is further configured to:

[0195] Loop through the following steps d times to obtain a CSI sample set corresponding to each of the multiple tasks, where d is the number of tasks:

[0196] Selecting a basis vector group from the O basis vector groups in the first basis vector space;

[0197] According to L in the one basis vector group task spatial basis vectors and the second basis vector space Mtask frequency domain basis vectors to construct a CSI sample set corresponding to one of the multiple tasks, where L task , M task is a positive integer greater than 1.

[0198] In some embodiments, the one basis vector group is randomly selected from the O basis vector groups.

[0199] In some embodiments, the L task The spatial basis vectors are randomly selected from the one basis vector group;

[0200] The M task The frequency domain basis vectors are N in the second basis vector space. sb The frequency domain basis vectors are randomly selected.

[0201] In some embodiments, the L task Less than N1*N2, the M task Less than N sb .

[0202] In some embodiments, the processing unit 410 is further configured to:

[0203] The following steps are executed k times in a loop to obtain a CSI sample set corresponding to a task, including k CSI samples, where k is greater than 1:

[0204] In the L task Randomly select L spatial basis vectors from the spatial basis vectors, and arrange the L spatial basis vectors in columns to obtain a matrix B, where L is a positive integer;

[0205] Construct a diagonal block matrix W1 = [B, 0; 0, B] based on the matrix B;

[0206] In the M task Randomly select M frequency domain basis vectors from the frequency domain basis vectors, and arrange the M frequency domain basis vectors in rows to obtain a matrix W f , where M is a positive integer;

[0207] According to the diagonal block matrix W1, the matrix W f And the random number matrix W2 generates a CSI sample in the CSI sample set corresponding to a task.

[0208] In some embodiments, the L<L task , M<M task .

[0209] In some embodiments, the elements in the random number matrix W2 obey a fourth distribution.

[0210] In some embodiments, the fourth distribution is a complex Gaussian distribution or a complex uniform distribution.

[0211] In some embodiments, the processing unit 410 is further configured to:

[0212] The diagonal block matrix W1, matrix W f Multiply it by the random number matrix W2 to obtain the first CSI sample;

[0213] Normalize each column in the matrix of the first CSI samples to obtain target CSI samples.

[0214] In some embodiments, the first CSI sample W′=W1W2W f =[w1,...,w Nsb ],

[0215] The target CSI sample W=[w1 / norm(w1),...,w Nsb / norm(w Nsb )], where norm() represents the two-norm.

[0216] Alternatively, in some embodiments, the communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system on chip. The processing unit may be one or more processors.

[0217] It should be understood that the above and other operations and / or functions of each unit in the device 400 according to the embodiment of the present application are respectively used to implement the corresponding processes in the method 200 shown in Figure 8. For the sake of brevity, they are not repeated here.

[0218] Figure 10 is a schematic structural diagram of a communication device 600 provided in an embodiment of the present application. The communication device 600 shown in Figure 10 includes a processor 610, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.

[0219] Optionally, as shown in FIG10 , the communication device 600 may further include a memory 620. The processor 610 may call and execute a computer program from the memory 620 to implement the method in the embodiment of the present application.

[0220] The memory 620 may be a separate device independent of the processor 610 , or may be integrated into the processor 610 .

[0221] Optionally, as shown in FIG10 , the communication device 600 may further include a transceiver 630 , and the processor 610 may control the transceiver 630 to communicate with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices.

[0222] The transceiver 630 may include a transmitter and a receiver. The transceiver 630 may further include an antenna, and the number of antennas may be one or more.

[0223] Optionally, the communication device 600 may specifically be a network device in an embodiment of the present application, and the communication device 600 may implement the corresponding processes implemented by the network device in each method in the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0224] Optionally, the communication device 600 may specifically be a mobile terminal / terminal device in an embodiment of the present application, and the communication device 600 may implement the corresponding processes implemented by the mobile terminal / terminal device in each method in the embodiment of the present application. For the sake of brevity, they will not be repeated here.

[0225] Figure 11 is a schematic structural diagram of a chip according to an embodiment of the present application. The chip 700 shown in Figure 11 includes a processor 710, which can call and run a computer program from a memory to implement the method according to the embodiment of the present application.

[0226] Optionally, as shown in FIG11 , the chip 700 may further include a memory 720. The processor 710 may call and execute a computer program from the memory 720 to implement the method in the embodiment of the present application.

[0227] The memory 720 may be a separate device independent of the processor 710 , or may be integrated into the processor 710 .

[0228] Optionally, the chip 700 may further include an input interface 730. The processor 710 may control the input interface 730 to communicate with other devices or chips, and specifically, may obtain information or data sent by other devices or chips.

[0229] Optionally, the chip 700 may further include an output interface 740. The processor 710 may control the output interface 740 to communicate with other devices or chips, and specifically, may output information or data to other devices or chips.

[0230] Optionally, the chip can be applied to the network device in the embodiments of the present application, and the chip can implement the corresponding processes implemented by the network device in each method of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0231] Optionally, the chip can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the chip can implement the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0232] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0233] It should be understood that the processor of the embodiments of the present application may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method embodiment can be completed by hardware integrated logic circuits in the processor or software instructions. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0234] It is understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0235] It should be understood that the above-mentioned memories are exemplary but not restrictive. For example, the memories in the embodiments of the present application may also be static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM RAM (DR RAM), etc. In other words, the memories in the embodiments of the present application are intended to include, but are not limited to, these and any other suitable types of memories.

[0236] An embodiment of the present application also provides a computer-readable storage medium for storing a computer program.

[0237] Optionally, the computer-readable storage medium can be applied to the network device in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0238] Optionally, the computer-readable storage medium can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the computer program enables the computer to execute the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0239] An embodiment of the present application also provides a computer program product, including computer program instructions.

[0240] Optionally, the computer program product can be applied to the network device in the embodiments of the present application, and the computer program instructions enable the computer to execute the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not repeated here.

[0241] Optionally, the computer program product can be applied to the mobile terminal / terminal device in the embodiments of the present application, and the computer program instructions enable the computer to execute the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0242] The embodiment of the present application also provides a computer program.

[0243] Optionally, the computer program can be applied to the network device in the embodiments of the present application. When the computer program runs on a computer, the computer executes the corresponding processes implemented by the network device in the various methods of the embodiments of the present application. For the sake of brevity, they are not described here.

[0244] Optionally, the computer program can be applied to the mobile terminal / terminal device in the embodiments of the present application. When the computer program runs on the computer, the computer executes the corresponding processes implemented by the mobile terminal / terminal device in the various methods of the embodiments of the present application. For the sake of brevity, they will not be repeated here.

[0245] Those skilled in the art will appreciate that the units and algorithm steps of each example 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 performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0246] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0247] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0248] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0249] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0250] If the functions are implemented in the form of 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 the present application, 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. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0251] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for generating a sample set, characterized in that: include: Construct a first basis vector space and a second basis vector space, wherein the first basis vector space includes O basis vector groups, each basis vector group includes N1*N2 spatial basis vectors, and the second basis vector space includes N sb frequency domain basis vectors, where O is a positive integer, N1 is the number of antenna ports in the first dimension of the transmitting end of the channel state information CSI, N2 is the number of antenna ports in the second dimension of the transmitting end of the CSI, and N sb Indicates the number of subbands; Based on the first basis vector space and the second basis vector space, a CSI sample set corresponding to each task in a plurality of tasks is constructed, wherein the CSI sample set corresponding to each task is constructed based on some spatial domain basis vectors in a basis vector group in the first basis vector space and some frequency domain basis vectors in the second basis vector space. The CSI sample sets corresponding to the plurality of tasks are used to train a first model. The first model is used to train a second model based on the CSI sample set of a target scene, and the second model is adapted to the target scene.

2. The method according to claim 1, characterized in that The matrix composed of N1*N2 spatial basis vectors in each basis vector group of the O basis vector groups is a unitary matrix. sb The matrix composed of frequency domain basis vectors is a unitary matrix.

3. The method according to claim 1 or 2, characterized in that The constructing of the first basis vector space includes: Loop through the following steps O times to obtain the spatial basis vectors of each of the O basis vector groups: Performing random sampling based on a first distribution to obtain a first matrix, and performing random sampling based on a second distribution to obtain a second matrix, wherein the first matrix is ​​a matrix of N1*N1 dimension, and the second matrix is ​​a matrix of N2*N2 dimension; Determine an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix; Performing a Kronecker product on the orthogonal matrix corresponding to the first matrix and the orthogonal matrix corresponding to the second matrix to obtain a target matrix; A column of the target matrix is ​​used as a spatial basis vector in a basis vector group in the first basis vector space.

4. The method according to claim 3, characterized in that The determining of an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix includes: Performing singular value decomposition (SVD) on the first matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the first matrix; Perform SVD on the second matrix, and use the obtained left singular matrix as the orthogonal matrix corresponding to the second matrix.

5. The method according to claim 3, characterized in that The determining of an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix includes: Performing Schmidt orthogonalization on the first matrix, and using the obtained orthogonal matrix as the orthogonal matrix corresponding to the first matrix; Perform Schmidt orthogonalization on the second matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the second matrix.

6. The method according to any one of claims 3 to 5, characterized in that The first distribution is a complex Gaussian distribution or a complex uniform distribution; The second distribution is a complex Gaussian distribution or a complex uniform distribution.

7. The method according to claim 6, characterized in that The constructing of the second basis vector space includes: Random sampling is performed based on the third distribution to obtain a third matrix, and the third matrix is ​​N sb *N sb Matrix of dimensions; Determine an orthogonal matrix corresponding to the third matrix; A column of the orthogonal matrix corresponding to the third matrix is ​​used as a frequency domain basis vector in the second basis vector space.

8. The method according to claim 7, characterized in that Determining the orthogonal matrix corresponding to the third matrix includes: Performing SVD on the third matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the third matrix; or Perform Schmidt orthogonalization on the third matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the third matrix.

9. The method according to claim 7 or 8, characterized in that The third distribution is a complex Gaussian distribution or a complex uniform distribution.

10. The method according to any one of claims 1 to 9, characterized in that The constructing, according to the first basis vector space and the second basis vector space, a CSI sample set corresponding to each of the multiple tasks includes: Loop through the following steps d times to obtain a CSI sample set corresponding to each of the multiple tasks, where d is the number of tasks: Selecting a basis vector group from the O basis vector groups in the first basis vector space; According to L in the one basis vector group task spatial basis vectors and the second basis vector space M task frequency domain basis vectors to construct a CSI sample set corresponding to one of the multiple tasks, where L task , M task is a positive integer greater than 1.

11. The method according to claim 10, characterized in that The one basis vector group is randomly selected from the O basis vector groups.

12. The method according to claim 10 or 11, characterized in that The L task The spatial basis vectors are randomly selected from the one basis vector group; The M task The frequency domain basis vectors are N in the second basis vector space. sb The frequency domain basis vectors are randomly selected.

13. The method according to any one of claims 10 to 12, characterized in that The L task Less than N1*N2, the M task Less than N sb .

14. The method according to any one of claims 10 to 13, characterized in that According to L in the one basis vector group task spatial basis vectors and the M in the second basis vector space task frequency domain basis vectors, constructing a CSI sample set corresponding to one of the multiple tasks, including: The following steps are executed k times in a loop to obtain a CSI sample set corresponding to a task, including k CSI samples, where k is greater than 1: In the L task Randomly select L spatial basis vectors from the spatial basis vectors, and arrange the L spatial basis vectors in columns to obtain a matrix B, where L is a positive integer; Construct a diagonal block matrix W1 = [B, 0; 0, B] based on the matrix B; In the M task Randomly select M frequency domain basis vectors from the frequency domain basis vectors, and arrange the M frequency domain basis vectors in rows to obtain a matrix W f , where M is a positive integer; According to the diagonal block matrix W1, the matrix W f And the random number matrix W2 generates a CSI sample in the CSI sample set corresponding to a task.

15. The method according to claim 14, characterized in that L<L task , M<M task .

16. The method according to claim 14 or 15, characterized in that The elements in the random number matrix W2 obey the fourth distribution.

17. The method according to claim 16, characterized in that The fourth distribution is a complex Gaussian distribution or a complex uniform distribution.

18. The method according to any one of claims 14 to 17, characterized in that According to the diagonal block matrix W1, the matrix W f And the random number matrix W2, generate a CSI sample in the CSI sample set corresponding to a task, including: The diagonal block matrix W1, matrix W f Multiply it by the random number matrix W2 to obtain the first CSI sample; Normalize each column in the matrix of the first CSI samples to obtain target CSI samples.

19. The method according to claim 18, characterized in that The first CSI sample W'=W1W2W f =[w1,...,w Nsb ], The target CSI sample W=[w1 / norm(w1),...,w Nsb / norm(w Nsb )], where norm() represents the two-norm.

20. A device for generating a sample set, characterized in that: include: The processing unit is configured to construct a first basis vector space and a second basis vector space, wherein the first basis vector space includes O basis vector groups, each basis vector group includes N1*N2 spatial basis vectors, and the second basis vector space includes N sb frequency domain basis vectors, where O is a positive integer, N1 is the number of antenna ports in the first dimension of the transmitting end of the channel state information CSI, N2 is the number of antenna ports in the second dimension of the transmitting end of the CSI, and N sb represents the number of subbands; and Based on the first basis vector space and the second basis vector space, a CSI sample set corresponding to each task in a plurality of tasks is constructed, wherein the CSI sample set corresponding to each task is constructed based on some spatial domain basis vectors in a basis vector group in the first basis vector space and some frequency domain basis vectors in the second basis vector space. The CSI sample sets corresponding to the plurality of tasks are used to train a first model. The first model is used to train a second model based on the CSI sample set of a target scene, and the second model is adapted to the target scene.

21. The device according to claim 20, characterized in that The matrix composed of N1*N2 spatial basis vectors in each basis vector group of the O basis vector groups is a unitary matrix. sb The matrix composed of frequency domain basis vectors is a unitary matrix.

22. The device according to claim 20 or 21, characterized in that The processing unit is further configured to: Loop through the following steps O times to obtain the spatial basis vectors of each of the O basis vector groups: Performing random sampling based on a first distribution to obtain a first matrix, and performing random sampling based on a second distribution to obtain a second matrix, wherein the first matrix is ​​a matrix of N1*N1 dimension, and the second matrix is ​​a matrix of N2*N2 dimension; Determine an orthogonal matrix corresponding to the first matrix and an orthogonal matrix corresponding to the second matrix; Performing a Kronecker product on the orthogonal matrix corresponding to the first matrix and the orthogonal matrix corresponding to the second matrix to obtain a target matrix; A column of the target matrix is ​​used as a spatial basis vector in a basis vector group in the first basis vector space.

23. The device according to claim 22, characterized in that The processing unit is further configured to: Performing singular value decomposition (SVD) on the first matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the first matrix; Perform SVD on the second matrix, and use the obtained left singular matrix as the orthogonal matrix corresponding to the second matrix.

24. The device according to claim 22, characterized in that The processing unit is further configured to: Performing Schmidt orthogonalization on the first matrix, and using the obtained orthogonal matrix as the orthogonal matrix corresponding to the first matrix; Perform Schmidt orthogonalization on the second matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the second matrix.

25. The device according to any one of claims 22 to 24, characterized in that The first distribution is a complex Gaussian distribution or a complex uniform distribution; The second distribution is a complex Gaussian distribution or a complex uniform distribution.

26. The device according to claim 25, characterized in that The processing unit is further configured to: Random sampling is performed based on the third distribution to obtain a third matrix, and the third matrix is ​​N sb *N sb Matrix of dimensions; Determine an orthogonal matrix corresponding to the third matrix; A column of the orthogonal matrix corresponding to the third matrix is ​​used as a frequency domain basis vector in the second basis vector space.

27. The device according to claim 26, characterized in that The processing unit is further configured to: Performing SVD on the third matrix, and using the obtained left singular matrix as the orthogonal matrix corresponding to the third matrix; or Perform Schmidt orthogonalization on the third matrix, and use the obtained orthogonal matrix as the orthogonal matrix corresponding to the third matrix.

28. The apparatus according to claim 26 or 27, characterized in that The third distribution is a complex Gaussian distribution or a complex uniform distribution.

29. The apparatus according to any one of claims 20 to 28, characterized in that The processing unit is further configured to: Loop through the following steps d times to obtain a CSI sample set corresponding to each of the multiple tasks, where d is the number of tasks: Selecting a basis vector group from the O basis vector groups in the first basis vector space; According to L in the one basis vector group task spatial basis vectors and the second basis vector space M task frequency domain basis vectors to construct a CSI sample set corresponding to one of the multiple tasks, where L task , M task is a positive integer greater than 1.

30. The apparatus according to claim 29, wherein The one basis vector group is randomly selected from the O basis vector groups.

31. The apparatus according to claim 29 or 30, characterized in that The L task The spatial basis vectors are randomly selected from the one basis vector group; The M task The frequency domain basis vectors are N in the second basis vector space. sb The frequency domain basis vectors are randomly selected.

32. The apparatus according to any one of claims 29 to 31, characterized in that The L task Less than N1*N2, the M task Less than N sb .

33. The apparatus according to any one of claims 29 to 32, characterized in that The processing unit is further configured to: The following steps are executed k times in a loop to obtain a CSI sample set corresponding to a task, including k CSI samples, where k is greater than 1: In the L task Randomly select L spatial basis vectors from the spatial basis vectors, and arrange the L spatial basis vectors in columns to obtain a matrix B, where L is a positive integer; Construct a diagonal block matrix W1 = [B, 0; 0, B] based on the matrix B; In the M task Randomly select M frequency domain basis vectors from the frequency domain basis vectors, and arrange the M frequency domain basis vectors in rows to obtain a matrix W f , where M is a positive integer; According to the diagonal block matrix W1, the matrix W f And the random number matrix W2 generates a CSI sample in the CSI sample set corresponding to a task.

34. The device according to claim 33, characterized in that L<L task ,M<M task 。 35. The apparatus according to claim 33 or 34, characterized in that The elements in the random number matrix W2 obey the fourth distribution.

36. The apparatus according to claim 35, wherein The fourth distribution is a complex Gaussian distribution or a complex uniform distribution.

37. The apparatus according to any one of claims 34 to 36, characterized in that The processing unit is further configured to: The diagonal block matrix W1, matrix W f Multiply it by the random number matrix W2 to obtain the first CSI sample; Normalize each column in the matrix of the first CSI samples to obtain target CSI samples.

38. The apparatus according to claim 37, wherein The first CSI sample W'=W1W2W f =[w1,...,w Nsb ], The target CSI sample W=[w1 / norm(w1),...,w Nsb / norm(w Nsb )], where norm() represents the two-norm.

39. A device for generating a sample set, characterized in that include: A processor and a memory, the memory being used to store a computer program, the processor being used to call and run the computer program stored in the memory, and execute the device according to any one of claims 1 to 18.

40. A chip, characterized in that: include: A processor, configured to call and run a computer program from a memory, so that a device equipped with the chip executes the method according to any one of claims 1 to 18.

41. A computer-readable storage medium, characterized in that Used to store a computer program, the computer program causing a computer to execute the method according to any one of claims 1 to 18.

42. A computer program product, characterized in that The method comprises computer program instructions for causing a computer to execute the method according to any one of claims 1 to 18.

43. A computer program, characterized in that The computer program causes a computer to execute the method according to any one of claims 1 to 18.

Images