CSI compression and decompression method, device and equipment based on AI model

Abstract

The invention discloses a CSI compression and decompression method, device and equipment based on an AI model, and belongs to the field of communication, and the CSI compression method based on the AI model in the embodiment of the invention comprises the steps that a terminal carries out downsampling or pre-compression processing on first CSI information obtained through measurement to obtain second CSI information; the terminal compresses the second CSI information through a first AI model to obtain target compressed CSI information; and the terminal sends first information to network side equipment, wherein the first information comprises the target compressed CSI information. In the embodiment of the invention, the CSI information obtained by measurement is subjected to down-sampling or pre-compression processing before being input into the AI model, and then is compressed through the AI model, so that the CSI compression performance can be improved.

Description

Technical Field

[0001] This application relates to the field of communications, and more particularly, to a CSI compression and decompression method, apparatus, and device based on an AI model. Background Art

[0002] At present, for the channel state information (CSI) based on an artificial intelligence (AI) model, the CSI is compressed by using the AI model. However, the compression performance of this CSI compression method is limited, and it is restricted by the model. How to further improve the CSI compression performance is a problem that needs to be solved. Summary of the Invention

[0003] Embodiments of this application provide a CSI compression and decompression method, apparatus, and device based on an AI model. The measured CSI information is subjected to downsampling or pre-compression processing before being input into the AI model, and then compressed by the AI model, thereby improving the CSI compression performance and solving the problem of limited compression performance when compressing CSI based on the AI model.

[0004] In a first aspect, a CSI compression method based on an AI model is provided, including:

[0005] The terminal performs downsampling or pre-compression processing on the measured first channel state information (CSI) to obtain second CSI information;

[0006] The terminal compresses the second CSI information through a first artificial intelligence (AI) model to obtain target compressed CSI information;

[0007] The terminal sends first information to the network-side device, where the first information includes the target compressed CSI information.

[0008] In a second aspect, a CSI decompression method based on an AI model is provided, including:

[0009] The network-side device receives the first information from the terminal; where the first information includes target compressed channel state information (CSI), the target compressed CSI is obtained by compressing the second CSI through a first AI model, and the second CSI is obtained by performing downsampling or pre-compression processing on the measured first CSI;

[0010] The network-side device decompresses the target compressed CSI through a second artificial intelligence (AI) model to obtain third CSI information.

[0011] In a third aspect, a CSI compression device based on an AI model is provided, including:

[0012] A processing unit configured to perform downsampling or pre-compression processing on the measured first channel state information (CSI) to obtain second CSI information;

[0013] The processing unit is further configured to perform compression processing on the second CSI information through a first artificial intelligence (AI) model to obtain target compressed CSI information;

[0014] A transceiver unit configured to send first information to a network-side device, where the first information includes the target compressed CSI information.

[0015] In a fourth aspect, a CSI decompression device based on an AI model is provided, including:

[0016] A transceiver unit configured to receive first information from a terminal; where the first information includes target compressed CSI information, the target compressed CSI information is obtained by performing compression processing on second CSI information through a first AI model, and the second CSI information is obtained by performing downsampling or pre-compression processing on the measured first CSI information;

[0017] A processing unit configured to perform decompression processing on the target compressed CSI information through a second artificial intelligence (AI) model to obtain third CSI information.

[0018] In a fifth aspect, a terminal is provided, the terminal includes a transceiver, a processor, and a memory, the memory stores a program or instruction that can run on the processor, and when the program or instruction is executed by the processor, the steps of the method described in the first aspect are implemented.

[0019] In a sixth aspect, a terminal is provided, including a processor and a communication interface;

[0020] Wherein, the processor is configured to perform downsampling or pre-compression processing on the measured first channel state information (CSI) to obtain second CSI information; the processor is further configured to perform compression processing on the second CSI information through a first artificial intelligence (AI) model to obtain target compressed CSI information;

[0021] Wherein, the communication interface is configured to send first information to a network-side device, where the first information includes the target compressed CSI information.

[0022] In a seventh aspect, a network-side device is provided. The network-side device includes a transceiver, a processor, and a memory. The memory stores a program or instructions that can run on the processor. When the program or instructions are executed by the processor, the steps of the method described in the second aspect are implemented.

[0023] In an eighth aspect, a network-side device is provided, including a processor and a communication interface;

[0024] wherein, the communication interface is used to receive first information from a terminal; wherein, the first information includes target compressed CSI information, and the target compressed CSI information is obtained by compressing second CSI information through a first AI model, and the second CSI information is obtained by downsampling or pre-compressing the measured first CSI information;

[0025] wherein, the processor is used to decompress the target compressed CSI information through a second artificial intelligence (AI) model to obtain third CSI information.

[0026] In a ninth aspect, a readable storage medium is provided. A program or instructions are stored on the readable storage medium. When the program or instructions are executed by a processor, the steps of the method described in the first aspect are implemented, or the steps of the method described in the second aspect are implemented.

[0027] In a tenth aspect, a wireless communication system is provided, including: a terminal and a network-side device. The terminal can be used to execute the steps of the method described in the first aspect, and the network-side device can be used to execute the steps of the method described in the second aspect.

[0028] In an eleventh aspect, a chip is provided. The chip includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run a program or instructions to implement the method described in the first aspect or the method described in the second aspect.

[0029] In a twelfth aspect, a computer program / program product is provided. The computer program / program product is stored in a storage medium. The program / program product is executed by at least one processor to implement the steps of the CSI compression method based on an AI model described in the first aspect or the steps of the CSI decompression method based on an AI model described in the second aspect.

[0030] In the embodiments of the present application, the terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information; and the terminal performs compression processing on the second CSI information through a first AI model to obtain target compressed CSI information. That is, in the embodiments of the present application, the measured CSI information is subjected to downsampling or pre-compression processing before being input into the AI model, and then compressed through the AI model, thereby improving the CSI compression performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for description in the embodiments of the present application. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0032] Figure 1 is a schematic diagram of a communication system architecture provided by an embodiment of the present application.

[0033] Figure 2 is a schematic diagram of a neural network provided by the present application.

[0034] Figure 3 is a schematic diagram of a neuron provided by the present application.

[0035] Figure 4 is a schematic diagram of CSI compression based on an AI model provided by the present application.

[0036] Figure 5 is a schematic flowchart of a method for CSI compression and decompression based on an AI model provided by an embodiment of the present application.

[0037] Figure 6 is a schematic diagram of CSI compression and decompression based on an AI model provided by an embodiment of the present application.

[0038] Figure 7 is another schematic diagram of CSI compression and decompression based on an AI model provided by an embodiment of the present application.

[0039] Figure 8 is a schematic block diagram of a CSI compression device based on an AI model provided by an embodiment of the present application.

[0040] Figure 9 is a schematic block diagram of a CSI decompression device based on an AI model provided by an embodiment of the present application.

[0041] Figure 10It is a schematic block diagram of a communication device provided according to an embodiment of the present application.

[0042] Figure 11 It is a schematic diagram of the hardware structure of a terminal provided according to an embodiment of the present application.

[0043] Figure 12 It is a schematic block diagram of a network-side device provided according to an embodiment of the present application. Specific implementation manners

[0044] Next, the technical solutions in the embodiments of the present application will be clearly described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0045] The terms "first", "second", etc. in the present application are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such terms can be interchanged under appropriate circumstances so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are usually of the same category, and do not limit the number of objects. For example, the first object can be one or multiple. In addition, "or" in the present application means at least one of the connected objects. For example, "A or B" covers three scenarios, namely, Scenario 1: including A and not including B; Scenario 2: including B and not including A; Scenario 3: including both A and B. The character " / " generally indicates an "or" relationship between the associated objects before and after.

[0046] The term "indicate" in the present application can be either a direct indication (or an explicit indication) or an indirect indication (or an implicit indication). Among them, a direct indication can be understood as that the sender clearly tells the receiver specific information, operations to be performed, or request results, etc. in the sent indication; an indirect indication can be understood as that the receiver determines the corresponding information according to the indication sent by the sender, or makes a judgment and determines the operations to be performed or request results, etc. according to the judgment result.

[0047] It should be noted that the technology described in the embodiments of the present application is not limited to the Ambient Internet of Things (IoT) system, but can also be used in other wireless communication systems, such as Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), Wireless Local Area Networks (WLAN), Wireless Fidelity (WiFi), Bluetooth systems, or other systems. The terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described technology can be used in the above-mentioned systems and radio technologies, as well as in other systems and radio technologies. The following description describes the New Radio (NR) system for example purposes and uses the NR term in most of the following descriptions, but these technologies can also be applied to systems other than the NR system, such as the 6th Generation (6 th Generation, 6G) communication system.

[0048] Figure 1The block diagram of a wireless communication system to which the embodiments of the present application can be applied is shown. The wireless communication system includes a terminal 11 and a network-side device 12. Among them, the terminal 11 can be a mobile phone, a tablet personal computer, a laptop computer, a notebook computer, a personal digital assistant (PDA), a handheld computer, a netbook, an ultra-mobile personal computer (UMPC), a mobile internet device (MID), an augmented reality (AR), a virtual reality (VR) device, a robot, a wearable device, a flight vehicle, a vehicle user equipment (VUE), a shipborne device, a pedestrian user equipment (PUE), a smart home (home appliances with wireless communication functions, such as refrigerators, TVs, washing machines, or furniture, etc.), a game console, a personal computer (PC), a teller machine, or a self-service machine, etc., which are terminal-side devices. Wearable devices include: smart watches, smart bracelets, smart earphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart ankle chains, etc.), smart wristbands, smart clothing, etc. Among them, vehicle user equipment can also be referred to as vehicle terminal, vehicle controller, vehicle module, vehicle component, vehicle chip, or vehicle unit, etc. It should be noted that the specific type of the terminal 11 is not limited in the embodiments of the present application.

[0049] The network-side device 12 may include an access network device or a core network device.

[0050] Among them, the access network device may also be referred to as a Radio Access Network (RAN) device, a radio access network function, or a radio access network unit. The access network device may include a base station, a Wireless Local Area Network (WLAN) Access Point (AP), or a Wireless Fidelity (WiFi) node, etc. Among them, the base station may be referred to as Node B (NB), Evolved Node B (eNB), the next generation Node B (gNB), New Radio Node B (NR Node B), access point, Relay Base Station (RBS), Serving Base Station (SBS), Base Transceiver Station (BTS), radio base station, radio transceiver, Basic Service Set (BSS), Extended Service Set (ESS), home Node B (HNB), home evolved Node B, Transmission Reception Point (TRP), or some other suitable term in the art. As long as the same technical effect is achieved, the base station is not limited to a specific technical term. It should be noted that in the embodiments of this application, only the base station in the NR system is taken as an example for introduction, and the specific type of the base station is not limited.

[0051] Among them, the core network devices may include but are not limited to at least one of the following: core network nodes, core network functions, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (L-NEF), Binding Support Function (BSF), Application Function (AF), Network Data Analytics Function (NWDAF), Location Management Function (LMF), etc. It should be noted that in the embodiments of this application, only the core network devices in the NR system are taken as examples for introduction, and the specific types of core network devices are not limited.

[0052] To facilitate a better understanding of the embodiments of this application, artificial intelligence (AI) is described.

[0053] Artificial intelligence (AI) has currently been widely applied in various fields. Incorporating artificial intelligence into wireless communication networks to significantly improve technical indicators such as throughput, latency, and user capacity is an important task for future wireless communication networks. There are various implementation methods for the AI module, such as neural networks, decision trees, support vector machines, Bayesian classifiers, etc. This application takes neural networks as an example for illustration, but does not limit the specific type of the AI module.

[0054] A schematic diagram of a neural network can be as Figure 2 shown. Among them, a neural network is composed of neurons, and a schematic diagram of a neuron is as Figure 3 shown. Among them, a1, a2, …, aK are inputs, w is the weight (multiplicative coefficient), b is the bias (additive coefficient), and σ(.) is the activation function. Common activation functions include Sigmoid, tanh, Rectified Linear Unit (ReLU), and so on.

[0055] The parameters of the neural network are optimized through gradient optimization algorithms. Gradient optimization algorithms are a class of algorithms that minimize or maximize an objective function (sometimes also called a loss function), and the objective function is often a mathematical combination of model parameters and data. For example, given data X and its corresponding label Y, we construct a neural network model f(.). After having the model, according to the input x, we can obtain the predicted output f(x), and we can calculate the gap between the predicted value and the true value (f(x) - Y), which is the loss function. Our goal is to find the appropriate W and b to minimize the value of the above loss function. The smaller the loss value, the closer our model is to the real situation.

[0056] Currently, common optimization algorithms are basically based on the error Back Propagation (BP) algorithm. The basic idea of the BP algorithm is that the learning process consists of two processes: the forward propagation of signals and the backward propagation of errors. During forward propagation, the input samples are input from the input layer, and after being processed layer by layer through each hidden layer, they are transmitted to the output layer. If the actual output of the output layer does not match the expected output, it enters the stage of backward propagation of errors. The error backpropagation is to transmit the output error back layer by layer through the hidden layer in a certain form and allocate the error to all units of each layer, thereby obtaining the error signals of each layer of units. This error signal is used as the basis for correcting the weights of each unit. This process of adjusting the weights of each layer during the forward propagation of signals and the backward propagation of errors is carried out cyclically. The process of continuously adjusting the weights is also the learning and training process of the network. This process continues until the error of the network output is reduced to an acceptable level or until a pre-set number of learning times is reached.

[0057] Common optimization algorithms include Gradient Descent, Stochastic Gradient Descent (SGD), mini-batch gradient descent, Momentum, Nesterov (specifically Stochastic Gradient Descent with Momentum), ADAptive GRADient descent (Adagrad), Adadelta, root mean square prop (RMSprop), Adaptive Moment Estimation (Adam), etc. When these optimization algorithms perform backpropagation of errors, they all calculate the derivative / partial derivative of the current neuron based on the error / loss obtained from the loss function, and then add the learning rate, previous gradients / derivatives / partial derivatives, etc. to obtain the gradient, which is then passed to the previous layer.

[0058] To facilitate a better understanding of the embodiments of this application, the CSI compression for non-AI models will be described.

[0059] CSI compression for non-AI models includes type I CSI compression, type II CSI compression, and enhanced type II (etype 2) CSI compression.

[0060] 1. type I CSI compression

[0061] In the case where it is not possible to report the complete channel or precoder, type I CSI compression reports the precoding matrix indicator (PMI) of the wideband or subband, that is, a two-dimensional discrete Fourier transform (DFT) vector and its phase rotation amount on the wideband or subband. Among them, type I mainly needs to report the index of the two-dimensional DFT vector and its phase rotation amount.

[0062] The reporting format of the above type I CSI compression is as follows:

[0063] Wideband CSI: Rank Indicator (RI) - PMI - Channel Quality Indicator (CQI), where when RI > 4, two CQIs need to be reported for two transport blocks (TBs), otherwise one CQI.

[0064] Sub-band CSI:

[0065] Part 1 CSI: RI + CQI of the first TB;

[0066] Part 2 CSI: Wideband CQI - Wideband PMI - CQI + PMI of even sub-bands - CQI + PMI of odd sub-bands; The omission principle is: omit part2 based on priority, that is, odd sub-bands can be omitted first.

[0067] 2. type II CSI compression

[0068] type II CSI compression is relative to a simple two-dimensional DFT vector and its phase rotation amount. The precoding vector PMI is represented as a linear weighting of a set of basis vectors. Among them, type 2 needs to report the basis vector index and the projection (amplitude and phase) on the basis vector.

[0069] The above type II CSI compression reporting format is as follows:

[0070] Part 1: RI - CQI - Number of non-zero wideband amplitude coefficients per layer {separately encoded};

[0071] Part 2: Wideband PMI {L vector} - PMI - Layer Indicator (LI) {i 1,4,l (Wideband amplitude 1)i2,1,l (phase)i2,2,l (sub-band amplitude 2)}, where the above L vector is the basis vector.

[0072] Among them, amplitude 1: 3 bits (scalar); amplitude 2: 1 bit.

[0073] 3. etype II CSI compression

[0074] Since the overhead of type II CSI compression is up to several hundred or even several thousand bits, etype II CSI compression is a further compression of type II CSI compression, that is, the vector composed of weighted coefficients on different sub-bands is further compressed into a vector composed of a set of frequency-domain basis vectors.

[0075] The above etype II CSI compression reporting format is as follows:

[0076] Part 1: RI - CQI - Number of non-zero wideband amplitude coefficients per layer {separately encoded};

[0077] Part 2: Wideband PMI{vector}-PMI: i 2,4,l Amplitude i 2,5,l Phase and i 1,7,l , {reported bitmap};

[0078] Pri(l, i, f) = 2·L·υ·π(f) + υ·i + l,

[0079] where π(f) f is the frequency-domain basis vector, π(0) = 0, π(N3 - 1) = 1, π(1) = 2;

[0080]

[0081] where l = 1, 2, …, υ, i = 0, 1, …, 2L - 1, and f = 0, 1, …, M υ -1.

[0082] The above Part 2 adopts a feedback method of unified compression for all subbands, where:

[0083] 0: L spatial domain basis vectors i 1,1 , i 1,2 and the strongest coefficient information {log2 2L bit} i for each layer 1,8,l (l = 1, …, υ);

[0084] 1: M frequency-domain basis vectors (i 1,5 (if reported), i 1,6,l (if reported)), reference amplitude information i 2,3,l , the v2LM - [KNZ / 2] bit i with the highest priority among the non-zero coefficient positions 1,7,l , v is the rank, v2LM - [KNZ / 2] coefficients with the highest priority{i 2,4,l , i 2,5,l};

[0085] 2: The [KNZ / 2] coefficients with the lowest priority among the non-zero coefficient positions.

[0086] To facilitate better understanding of the embodiments of the present application, the CSI compression based on the AI model is described.

[0087] As Figure 4 shown, the target CSI (target CSI) or codebook (such as W N*B ) can be compressed through the AI model, such as compressed into an AI-based CSI feedback value, and then reported to the network-side device, and the network-side device performs decompression to obtain the decompressed codebook (such as: W′ N*B ).

[0088] To facilitate the 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 above related technologies can be arbitrarily combined with the technical solutions of the embodiments of the present application as optional solutions, and all of them fall within the protection scope of the embodiments of the present application. The embodiments of the present application include at least some of the following content.

[0089] Figure 5 is a schematic flowchart of a CSI compression and decompression method 200 based on an AI model according to an embodiment of the present application, as Figure 5 shown, the CSI compression and decompression method 200 based on an AI model may include at least some of the following content:

[0090] S210, the terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information;

[0091] S220, the terminal compresses the second CSI information through a first AI model to obtain target compressed CSI information;

[0092] S230, the terminal sends a first message to the network-side device, and the first message includes the target compressed CSI information;

[0093] S240, the network-side device receives the first message from the terminal;

[0094] S250, the network-side device decompresses the target compressed CSI information through a second AI model to obtain third CSI information.

[0095] It should be understood that Figure 5 shows the steps or operations of the CSI compression and decompression method 200 based on an AI model, but these steps or operations are only examples, and the present application can also perform other operations or Figure 5 variations of each operation in.

[0096] The CSI information described in the embodiments of the present application may also be referred to as channel information or other similar names, and the embodiments of the present application do not limit this.

[0097] In the embodiments of the present application, the terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information; and the terminal compresses the second CSI information through a first AI model to obtain target compressed CSI information. That is, in the embodiments of the present application, the measured CSI information is subjected to downsampling or pre-compression processing before being input into the AI model, and then compressed through the AI model, thereby improving the CSI compression performance.

[0098] The AI model described in the embodiments of the present application may also be referred to as an AI unit, an AI model / AI unit, a machine learning (ML) model, an ML unit, an AI structure, an AI function, an AI feature, a neural network, a neural network function, a neural network capability, etc. Alternatively, the AI model described in the present application may also refer to a processing unit capable of implementing specific algorithms, formulas, processing flows, capabilities, etc. related to AI. Alternatively, the AI model described in the present application may be a processing method, algorithm, function, module, or unit for a specific data set. Alternatively, the AI model described in the present application may be a processing method, algorithm, function, module, or unit running on AI / ML-related hardware such as a graphics processing unit (GPU), a neural network processing unit (NPU), a tensor processing unit (TPU), an application-specific integrated circuit (ASIC), etc. The present application does not make specific limitations thereto. Optionally, the specific data set includes the input or output of the AI model.

[0099] The identifier of the AI model described in the embodiments of the present application may be an AI unit identifier, an AI structure identifier, an AI algorithm identifier, or the identifier of a specific data set associated with the AI model described in the present application, or the identifier of a specific scenario, environment, channel feature, or device related to the AI model described in the present application, or the identifier of a function, feature, capability, or module related to the AI model described in the present application. The present application does not make specific limitations thereto.

[0100] The downsampling process described in the embodiments of the present application may also be referred to as a dimensionality reduction process or other similar names, and the embodiments of the present application do not limit this.

[0101] In some embodiments, the terminal performs a downsampling process on the measured first CSI information to obtain second CSI information. Specifically, it may include: the terminal performs a single downsampling process on the measured first CSI information, or the terminal performs at least two downsampling processes on the measured first CSI information to obtain second CSI information.

[0102] In some embodiments, the terminal performs a pre-compression process on the measured first CSI information to obtain second CSI information. Specifically, it may include: the terminal performs a single pre-compression process on the measured first CSI information, or the terminal performs at least two pre-compression processes on the measured first CSI information to obtain second CSI information.

[0103] In some embodiments, the terminal performs downsampling on the measured first CSI information, and then performs SVD decomposition on it to obtain second CSI information. It can be understood that the first CSI information is channel information. First, the channel information is downsampled, and then the sampled channel information is subjected to SVD decomposition to obtain codebook information.

[0104] In some embodiments, the terminal performs downsampling on the measured first CSI information, and then performs pre-compression processing on it to obtain second CSI information.

[0105] Exemplarily, the third CSI information obtained by the network-side device through decompressing the target compressed CSI information by the second AI model can be approximately regarded as the second CSI information. In other words, the closer the third CSI information is to the second CSI information, the better the CSI compression performance of the first AI model or the CSI decompression performance of the second AI model, or the better the performance of this pair of AI models (i.e., the first AI model and the second AI model).

[0106] In some embodiments, the CSI compression and decompression method 200 based on the AI model further includes:

[0107] The network-side device performs super-resolution processing on the third CSI information to obtain fourth CSI information.

[0108] In this embodiment, the network-side device performs super-resolution processing on the third CSI information to obtain fourth CSI information. The fourth CSI information can be equivalent to the first CSI information, that is, the network-side restores the CSI information before compression through decompression processing and super-resolution processing. Among them, the closer the fourth CSI information is to the first CSI information, the better the CSI compression performance of the first AI model or the CSI decompression performance of the second AI model.

[0109] The super-resolution processing described in the embodiments of the present application corresponds to downsampling or pre-compression processing and is the reverse processing process of downsampling or pre-compression processing. Specifically, it may refer to: performing interpolation processing on the third CSI information, or, performing downsampling or pre-compression processing on the third CSI information. Further, the super-resolution processing described in the embodiments of the present application can be understood as: inferring the entire sub-band from the sub-band subset, or, inferring all ports from the port subset, or, inferring the CSI information from the eigenvector.

[0110] In one embodiment, the super-resolution processing and downsampling or pre-compression processing are also performed using an AI model.

[0111] Exemplarily, in the embodiments of the present application, the first CSI information may be downsampled or pre-compressed based on the payload indicated by the network side to obtain the second CSI information, so that the target compressed CSI information obtained by processing based on the first AI model meets the payload requirements indicated by the network side.

[0112] In some embodiments, the first CSI information includes, but is not limited to, at least one of the following:

[0113] Precoding matrix, PMI, channel information.

[0114] In some embodiments, the first CSI information may include CSI information at one or more time points. It can be understood that if the first CSI information includes CSI information at multiple time points, further compression processing in the time domain can be performed on the first CSI. It can be understood that each CSI information in the time domain is of the same type, either all precoding matrices or all channel information.

[0115] Optionally, the precoding matrix included in the first CSI information may be obtained by performing singular value decomposition (SVD) on the received channel information.

[0116] Optionally, the channel information included in the first CSI information may be the channel information obtained by receiving CSI-RS.

[0117] Optionally, the first CSI information includes CSI information corresponding to the measured CSI-RS, or CSI information predicted based on the CSI information corresponding to the measured CSI-RS.

[0118] Exemplarily, the first CSI information includes a precoding matrix such as: W N*B , where N represents the number of ports or CSI ports or CSI-RS ports, and B represents the number of subbands or CSI subbands or CSI-RS subbands. Specifically, for example, as Figure 6 shown, the terminal performs downsampling or pre-compression processing on the measured precoding matrix W N*B to obtain the second CSI information (such as: W N*B / 2 , W N*(B / 2) , W N / 2*B ); and the terminal performs compression processing on the second CSI information through the first AI model to obtain the target compressed CSI information; then, the terminal sends the target compressed CSI information to the network side device. Correspondingly, the network side device performs decompression processing on the target compressed CSI information through the second AI model to obtain the third CSI information (such as: W′ N*B / 2 , W′ N*(B / 2), W′ N / 2*B ); and the network-side device performs super-resolution processing on the third CSI information to obtain W N*B ′.

[0119] Exemplarily, the first CSI information includes channel information such as: H N*B , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, and B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands. Specifically, for example, as Figure 7 shown, the terminal performs downsampling or pre-compression processing on the measured channel information H N*B to obtain the second CSI information (such as: H N*B / 2 , H N*(B / 2) , H N / 2*B ); and the terminal compresses the second CSI information through a first AI model to obtain the target compressed CSI information; then, the terminal sends the target compressed CSI information to the network-side device. Correspondingly, the network-side device decompresses the target compressed CSI information through a second AI model to obtain the third CSI information (such as: H′ N*B / 2 , H′ N*(B / 2) , H′ N / 2*B ); and the network-side device performs super-resolution processing on the third CSI information to obtain H N*B ′.

[0120] In some embodiments, the first information further includes but is not limited to at least one of the following:

[0121] The payload of the target compressed CSI information, the first CSI information, channel quality indicator (CQI) information, rank indication (RI) information, the second CSI information, and the relevant information of downsampling or pre-compression corresponding to the second CSI information.

[0122] Exemplarily, the CQI information may correspond to the first CSI information. For example, the CQI of each subband before downsampling is reported. Optionally, the CQI information may correspond to the second CSI information. For example, the CQI of each subband after downsampling is reported.

[0123] Exemplarily, the first information includes the payload of the target compressed CSI information. In this case, the network-side device may judge the CSI compression performance of the first AI model based on the payload of the target compressed CSI information, or determine the second AI model for decompressing the CSI information.

[0124] Exemplarily, the first information includes CQI information or RI information. In this case, the network may judge the channel quality of the terminal based on the CQI information or RI information.

[0125] Exemplarily, the first information includes the CQI information or RI information associated with the target compressed CSI information. In this case, the network-side device may determine the CSI compression performance of the first AI model based on the CQI information or RI information associated with the target compressed CSI information.

[0126] In some embodiments, the CSI compression and decompression method 200 based on an AI model further includes:

[0127] The terminal sends second information to the network-side device;

[0128] Wherein, the second information includes at least one of the following:

[0129] The first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, and the association relationship between the second CSI information and the first information.

[0130] Exemplarily, the first information includes at least one of the following: the payload of the target compressed CSI information, the target compressed CSI information, CQI, and RI information. The terminal also reports second information, which includes at least one of the following: the first CSI information, the second CSI information, the indication information indicating the association between the first CSI information and the target compressed CSI information, the indication information indicating the association between the second CSI information and the target compressed CSI information, and timestamp information.

[0131] Exemplarily, the first information or the second information includes the first CSI information. In this case, the network-side device may compare the first CSI information and the fourth CSI information to determine at least one of the following: the CSI compression performance of the first AI model, the CSI decompression performance of the second AI model, the performance of the second CSI information obtained by downsampling or pre-compressing the first CSI information, and the performance of the fourth CSI information obtained by super-resolution processing of the third CSI information.

[0132] Exemplarily, the network-side device may compare the generalized cosine similarity (SGCS) between the first CSI information and the fourth CSI information to determine the CSI performance of the first AI model and / or the second AI model.

[0133] Exemplarily, the network-side device may compare the SGCS between the second CSI information and the fourth CSI information to determine the CSI performance of the first AI model and / or the second AI model.

[0134] Exemplarily, the network-side device compares the SGCS of the first CSI information and the fourth CSI information with the SGCS of the second CSI information and the fourth CSI information to determine the performance of downsampling or pre-compression. For example, by comparing the ratio of the SGCS of the first CSI information and the fourth CSI information to the SGCS of the second CSI information and the fourth CSI information, or by comparing the difference between the SGCS of the first CSI information and the fourth CSI information and the SGCS of the second CSI information and the fourth CSI information to determine the performance of downsampling or pre-compression. It should be noted that SGCS is an optional embodiment and can also be replaced by other parameters, such as the normalized mean square error (NMSE) in the time domain.

[0135] Exemplarily, the first information includes the CQI information or RI information associated with the first CSI information. In this case, the network-side device can determine the CSI compression performance of the first AI model or the channel quality of the downsampled first CSI based on the CQI information or RI information associated with the first CSI information.

[0136] Exemplarily, the first information or the second information includes the second CSI information. In this case, the network-side device can compare the second CSI information and the third CSI information to determine the CSI compression performance of the first AI model or the CSI decompression performance of the second AI model.

[0137] Exemplarily, the first information or the second information includes the relevant information of the downsampling or pre-compression corresponding to the second CSI information. In this case, the network-side device can determine the relevant information of the super-resolution processing based on the relevant information of the downsampling or pre-compression corresponding to the second CSI information.

[0138] In some embodiments, when the terminal performs downsampling processing on the first CSI information, the sub-bands corresponding to the second CSI information are a subset of the sub-bands corresponding to the first CSI information, or the ports corresponding to the second CSI information are a subset of the ports corresponding to the first CSI information.

[0139] For example, the first CSI information is the precoding matrix W N*B , and the second CSI information is W N*S(B) , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of sub-bands or the number of CSI sub-bands or the number of CSI-RS sub-bands, and S(B) is a subset of B.

[0140] For example, the first CSI information is the precoding matrix W N*B , and the second CSI information is W S(N)*B , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of sub-bands or the number of CSI sub-bands or the number of CSI-RS sub-bands, and S(N) is a subset of N.

[0141] For example, the first CSI information is the precoding matrix W N*B , and the second CSI information is W S(N)*S(B) , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands, S(N) is a subset of N, and S(B) is a subset of B.

[0142] For example, the first CSI information is the channel information H N*B , and the second CSI information is H N*B(B) , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands, and S(B) is a subset of B.

[0143] For example, the first CSI information is the channel information H N*B , and the second CSI information is H S(N)*B , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands, and S(N) is a subset of N.

[0144] For example, the first CSI information is the channel information H N*B , and the second CSI information is H S(N)*S(B) , where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands, S(N) is a subset of N, and S(B) is a subset of B.

[0145] In some embodiments, the subband corresponding to the second CSI information is the odd-numbered subband corresponding to the first CSI information, or the subband corresponding to the second CSI information is the even-numbered subband corresponding to the first CSI information. For example, the first CSI information includes the precoding matrix W N*B , and the second CSI information is W N*(B / 2) . Again, for example, the first CSI information includes the channel information H N*B , and the second CSI information is H N*(B / 2) . Where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, and B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands.

[0146] Exemplarily, the terminal can interact with the network-side device to determine whether the selected subband is an odd-numbered subband or an even-numbered subband.

[0147] In some embodiments, the port corresponding to the second CSI information is the odd-bit port corresponding to the first CSI information, or the port corresponding to the second CSI information is the even-bit port corresponding to the first CSI information. Alternatively, the port corresponding to the second CSI information is the vertically polarized port corresponding to the first CSI information, or the port corresponding to the second CSI information is the horizontally polarized port corresponding to the first CSI information. For example, the first CSI information includes a precoding matrix W N*B , and the second CSI information is W (N / 2)*B . Another example is that the first CSI information includes channel information H N*B , and the second CSI information is H (N / 2)*B . Where N represents the number of ports or the number of CSI ports or the number of CSI-RS ports, and B represents the number of subbands or the number of CSI subbands or the number of CSI-RS subbands.

[0148] Exemplarily, the terminal can interact with the network-side device to determine whether the selected port is an odd-bit port or an even-bit port. Alternatively, the terminal can interact with the network-side device to determine whether the selected polarization is vertical or horizontal.

[0149] In some embodiments, the subband corresponding to the second CSI information is determined based on at least one of the following: a bitmap associated with subband selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the subband selection interval, the correspondence between the RI and the subband selection interval, the correspondence between the output load of the first AI model and the subband selection interval, and the correspondence between the input load of the first AI model and the subband selection interval.

[0150] Optionally, the correspondence between the CSI-RS bandwidth and the subband selection interval can be specified by a protocol, or the correspondence between the CSI-RS bandwidth and the subband selection interval can be configured by the network side, or the correspondence between the CSI-RS bandwidth and the subband selection interval can be determined through negotiation between the terminal and the network-side device.

[0151] Exemplarily, the network side indicates to select one subband at an interval of M subbands.

[0152] Optionally, the correspondence between the RI and the subband selection interval can be specified by a protocol, or the correspondence between the RI and the subband selection interval can be configured by the network side, or the correspondence between the RI and the subband selection interval can be determined through negotiation between the terminal and the network-side device.

[0153] Exemplarily, the network indicates different subband selection intervals for different RIs, such as separately configuring the subband intervals for high Rank and low rank.

[0154] Exemplarily, the network indicates the sub-band selection for each layer, or all layers use the same sub-band selection rule.

[0155] Optionally, the correspondence between the output load of the first AI model and the sub-band selection interval can be agreed upon by a protocol, or the correspondence between the output load of the first AI model and the sub-band selection interval can be configured by the network side, or the correspondence between the output load of the first AI model and the sub-band selection interval can be determined through negotiation between the terminal and the network-side device.

[0156] Exemplarily, the network or a protocol agrees on a correspondence or a table for indicating the correspondence between each load and sub-band selection. In an optional embodiment, sub-band selection can be achieved by increasing the number of resource blocks (RBs) included in a sub-band. For example, for a channel with a bandwidth of 48 RBs, if the bandwidth of a sub-band is 4 RBs, the channel information of 12 sub-bands needs to be reported; if a sub-band includes 8 RBs, only 6 sub-bands need to be reported.

[0157] Optionally, sub-band selection can be understood as selecting M sub-bands from N sub-bands, where M is less than N, or sub-band selection can be understood as increasing the number of RBs included in a sub-band or indicating the number of RBs included in a sub-band.

[0158] Optionally, the correspondence between the input load of the first AI model and the sub-band selection interval can be agreed upon by a protocol, or the correspondence between the input load of the first AI model and the sub-band selection interval can be configured by the network side, or the correspondence between the input load of the first AI model and the sub-band selection interval can be determined through negotiation between the terminal and the network-side device.

[0159] In some embodiments, the port corresponding to the second CSI information is determined based on at least one of the following: a bitmap associated with port selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the port selection interval, the correspondence between the RI and the port selection interval, the correspondence between the output load of the first AI model and the port selection interval, and the correspondence between the input load of the first AI model and the port selection interval.

[0160] Optionally, the correspondence between the CSI-RS bandwidth and the port selection interval can be agreed upon by a protocol, or the correspondence between the CSI-RS bandwidth and the port selection interval can be configured by the network side, or the correspondence between the CSI-RS bandwidth and the port selection interval can be determined through negotiation between the terminal and the network-side device.

[0161] Optionally, the correspondence between the RI and the port selection interval can be agreed upon by a protocol, or the correspondence between the RI and the port selection interval can be configured by the network side, or the correspondence between the RI and the port selection interval can be determined through negotiation between the terminal and the network-side device.

[0162] Optionally, the correspondence between the output payload of the first AI model and the port selection interval may be agreed upon by a protocol, or the correspondence between the output payload of the first AI model and the port selection interval may be configured by the network side, or the correspondence between the output payload of the first AI model and the port selection interval may be determined through negotiation between the terminal and the network side device.

[0163] Optionally, the correspondence between the input payload of the first AI model and the port selection interval may be agreed upon by a protocol, or the correspondence between the input payload of the first AI model and the port selection interval may be configured by the network side, or the correspondence between the input payload of the first AI model and the port selection interval may be determined through negotiation between the terminal and the network side device.

[0164] In some embodiments, the subbands corresponding to the second CSI information are determined based on at least one of the following:

[0165] The first measurement, a specific AI model;

[0166] Wherein, the first measurement includes but is not limited to at least one of the following: the received power of the CSI-RS on the subband corresponding to the first CSI (such as the Reference Signal Received Power (RSRP)), the signal-to-noise ratio (SNR) or the signal-to-interference plus noise ratio (SINR) on the subband corresponding to the first CSI, the channel or channel characteristics on the subband corresponding to the first CSI.

[0167] Exemplarily, the specific AI model is used for downsampling.

[0168] Exemplarily, the subbands corresponding to the second CSI information may also be determined based on the selection result of the subbands corresponding to the first CSI information output by the specific AI model.

[0169] In some embodiments, the second CSI information is determined based on a specific AI model. The terminal inputs the first CSI information into the specific AI model, and the obtained output is the second CSI information.

[0170] Exemplarily, the sub-band selection is performed by the terminal itself. Optionally, the terminal determines the target sub-band based on the first measurement. For example, the terminal determines whether to select the first sub-band based on whether the first measurement difference between the first sub-band and the first reference sub-band is greater than the first threshold. Alternatively, the terminal determines the target sub-band based on a specific AI model. Optionally, the terminal inputs the first CSI information into the specific AI model, and the output obtained is the target sub-band. Or, the terminal inputs the first CSI information into the specific AI model, and the output obtained is the second CSI information.

[0171] In some embodiments, the port corresponding to the second CSI information is determined based on at least one of the following:

[0172] The first measurement, the specific AI model;

[0173] Wherein, the first measurement includes but is not limited to at least one of the following: the received power of the CSI-RS on the sub-band corresponding to the first CSI (such as the Reference Signal Received Power (RSRP)), the signal-to-noise ratio (SNR) or the signal-to-interference plus noise ratio (SINR) on the sub-band corresponding to the first CSI, the channel or channel characteristics on the sub-band corresponding to the first CSI.

[0174] Exemplarily, the port selection is performed by the terminal itself. Optionally, the terminal determines the target port based on the first measurement. For example, the terminal determines whether to select the first port based on whether the first measurement difference between the first port and the first reference port is greater than the first threshold. Alternatively, the terminal determines the target port based on a specific AI model. Optionally, the terminal inputs the first CSI information into the specific model, and the output obtained is the target port, or the second CSI information.

[0175] In some embodiments, the first information further includes at least one of the following:

[0176] The sub-band information corresponding to the second CSI information, the port information corresponding to the second CSI information, the information of the specific AI model.

[0177] In this embodiment, the network-side device may determine the relevant information for super-resolution processing based on the sub-band information corresponding to the second CSI information or the port information corresponding to the second CSI information. Or, the network-side device may determine the corresponding decoding AI model based on a specific AI model to recover the fourth CSI information approximated to the first CSI information.

[0178] Exemplarily, the first information includes but is not limited to at least one of the following:

[0179] The target compressed CSI information, the payload of the target compressed CSI information, the first CSI information, the CQI information, the RI information, the second CSI information, the related information of downsampling or pre-compression corresponding to the second CSI information, the sub-band information corresponding to the second CSI information, and the port information corresponding to the second CSI information.

[0180] In some embodiments, when the terminal performs pre-compression processing on the first CSI information, the second CSI information is represented by K eigenvectors, and the value of K satisfies at least one of the following: K < N, K < B;

[0181] Wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of sub-bands corresponding to the first CSI information, and K, N, and B are all positive integers.

[0182] In some embodiments, the second CSI information includes but is not limited to at least one of the following:

[0183] K discrete Fourier transform (DFT) vectors, K eigenvectors in type II CSI compression, K eigenvectors in enhanced type II CSI compression, the amplitude or phase coefficients of K DFT vectors, and the amplitude or phase coefficients of K - 1 DFT vectors.

[0184] It should be noted that when the second CSI information includes the amplitude or phase coefficients of K - 1 DFT vectors, it can be understood that the amplitude or phase coefficients of one of them (such as the first one) DFT vector are normalized.

[0185] In some embodiments, the K eigenvectors are determined based on at least one of the following:

[0186] A bit map associated with eigenvector selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the eigenvectors, the correspondence between the port and the eigenvectors, the correspondence between the RI and the eigenvectors, the correspondence between the output payload of the first AI model and the eigenvectors, and the correspondence between the input payload of the first AI model and the eigenvectors.

[0187] Optionally, the correspondence between the CSI-RS bandwidth and the eigenvectors can be agreed by the protocol, or the correspondence between the CSI-RS bandwidth and the eigenvectors can be configured by the network side, or the correspondence between the CSI-RS bandwidth and the eigenvectors can be determined by negotiation between the terminal and the network side device.

[0188] Optionally, the correspondence between the RI and the feature vector can be agreed upon by a protocol, or the correspondence between the RI and the feature vector can be configured by the network side, or the correspondence between the RI and the feature vector can be determined through negotiation between the terminal and the network side device.

[0189] Optionally, the correspondence between the output payload of the first AI model and the feature vector can be agreed upon by a protocol, or the correspondence between the output payload of the first AI model and the feature vector can be configured by the network side, or the correspondence between the output payload of the first AI model and the feature vector can be determined through negotiation between the terminal and the network side device.

[0190] Optionally, the correspondence between the input payload of the first AI model and the feature vector can be agreed upon by a protocol, or the correspondence between the input payload of the first AI model and the feature vector can be configured by the network side, or the correspondence between the input payload of the first AI model and the feature vector can be determined through negotiation between the terminal and the network side device.

[0191] In some embodiments, the first information further includes the K feature vectors.

[0192] In this embodiment, the network side device can determine information related to super-resolution processing based on the K feature vectors.

[0193] Exemplarily, the first information includes but is not limited to at least one of the following:

[0194] The target compressed CSI information, the payload of the target compressed CSI information, the first CSI information, CQI information, RI information, the second CSI information, information related to downsampling or pre-compression corresponding to the second CSI information, the K feature vectors.

[0195] In some embodiments, the CSI compression and decompression method 200 based on an AI model further includes:

[0196] The network side device performs at least one of the following according to the first information:

[0197] Monitor the performance of the first AI model;

[0198] Monitor the performance of the second AI model;

[0199] Monitor the performance of the fourth CSI information obtained after super-resolution processing of the third CSI information;

[0200] Monitor the performance of the second CSI information obtained after downsampling or pre-compression processing of the first CSI information.

[0201] Exemplarily, the network-side device monitors the performance of the first AI model and / or the second AI model. In such a case, the terminal needs to report the CSI information after downsampling or pre-compression processing (i.e., the second CSI information) to assist the network-side device in monitoring the CSI information (i.e., the third CSI information) decompressed from the second AI model (e.g., comparing the second CSI information and the third CSI information to determine the performance of the second AI model).

[0202] Exemplarily, the network-side monitors the performance of the fourth CSI information obtained after super-resolution processing of the third CSI information. In such a case, the terminal needs to report the CSI information before downsampling or pre-compression processing (i.e., the first CSI information) to assist the network-side in monitoring the CSI information after super-resolution processing (i.e., the fourth CSI information).

[0203] Optionally, in an alternative embodiment, the network-side needs to segmentally determine the effectiveness of the second AI model and the effectiveness of super-resolution. In such a case, the terminal needs to report the CSI information before downsampling or pre-compression processing (i.e., the first CSI information) and the relevant information on downsampling or pre-compression.

[0204] In some embodiments, after downsampling or pre-compression processing, the dimensions of the input information (such as sub-bands or ports) of different AI models are different, or the input loads of different AI models are different, or the output loads of different AI models are different.

[0205] Exemplarily, different AI models can be identified by at least one of the following:

[0206] Model ID;

[0207] Model input information;

[0208] Model output information;

[0209] Model function information.

[0210] In some embodiments, before the terminal performs downsampling or pre-compression processing on the first CSI information, the CSI compression and decompression method 200 based on the AI model further includes:

[0211] The terminal receives the third information from the network-side device; or,

[0212] The terminal sends the third information to the network-side device;

[0213] Among them, the third information includes, but is not limited to, at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression related information respectively associated with the at least one AI model, and dataset information respectively associated with the at least one AI model for training or updating or finetuning. Among them, the at least one AI model includes the first AI model.

[0214] Optionally, the input-related information of the i-th AI model among the at least one AI model includes, but is not limited to, at least one of the following: the input payload size of the i-th AI model, the input payload type of the i-th AI model (such as a precoding matrix or channel information), and the input dimension of the i-th AI model (such as a subband or a port).

[0215] Optionally, the output-related information of the i-th AI model among the at least one AI model includes, but is not limited to, at least one of the following: the output payload size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output payload type of the i-th AI model (such as a precoding matrix or channel information), and the output dimension of the i-th AI model (such as a subband or a port).

[0216] Exemplarily, different output payloads or input payloads may be associated with different dimensions.

[0217] Exemplarily, different AI models may use the same dataset, but when the UE trains different AI models, different downsampling or pre-compression processes need to be performed on the same dataset.

[0218] In some embodiments, the dataset information is associated with at least one of the following:

[0219] The payload of the target compressed CSI information, the dimension information of the second CSI information, the number of subbands of the second CSI information, and the number of ports of the second CSI information.

[0220] In some embodiments, all the CSI information included in a dataset corresponds to the same number of subbands. If the model is trained based on this dataset, during the inference process, the number of subbands of the second CSI information or the first CSI information is the same as the number of subbands of the CSI information in this dataset.

[0221] In some embodiments, all the CSI information included in a dataset corresponds to the same dimension. If the model is trained based on this dataset, during the inference process, the dimension of the second CSI information or the first CSI information is the same as the dimension of the CSI information in this dataset, where the dimension includes one or more of the following: time domain dimension, frequency domain dimension (such as the number of subbands, the number of RBs), spatial domain dimension (such as: the number of ports), and Doppler domain dimension.

[0222] In some embodiments, the CSI information included in a data set includes different numbers of sub-bands. Before training the model, the data set needs to be preprocessed, and the preprocessing includes at least one of the following:

[0223] Classify the data set into N data sets, and each data set includes the same number of sub-bands;

[0224] Downsample a part of the data in the data set; for example, for CSI information including 4 sub-bands and CSI information including 8 sub-bands, sample the 8 sub-bands to obtain 4 sub-band data).

[0225] In some embodiments, the CSI information included in a data set includes different dimensionality information. Before training the model, the data set needs to be preprocessed, and the preprocessing includes at least one of the following:

[0226] Classify the data set into N data sets, and each data set includes the same dimensionality information;

[0227] Downsample a part of the data in the data set; for example, downsample the high-dimensional CSI information so that the dimensionality information of part of the data or all of the data is the same.

[0228] In some embodiments, before the terminal performs downsampling or pre-compression processing on the first CSI information, the CSI compression and decompression method 200 based on the AI model further includes:

[0229] The terminal sends fourth information to the network-side device;

[0230] Wherein, the fourth information includes at least one of the following: whether the terminal supports downsampling or pre-compression, the relevant information of the downsampling or pre-compression supported by the terminal, the candidate values of the output payload of the AI model used for CSI compression supported by the terminal, the candidate values of the input payload of the AI model used for CSI compression supported by the terminal, the relevant information of the AI model used for CSI compression supported by the terminal, the candidate values of the input dimension of the AI model used for CSI compression supported by the terminal.

[0231] In this embodiment, the network-side device can obtain the relevant capabilities of the terminal based on the fourth information, so as to configure the relevant parameters for CSI compression based on the AI model.

[0232] In some embodiments, the relevant information of the downsampling or pre-compression includes but is not limited to at least one of the following:

[0233] The granularity of downsampling or pre-compression, the method of downsampling or pre-compression, the dimension of downsampling or pre-compression and the corresponding super-resolution dimension, and the relationship between downsampling or pre-compression and the first CSI information.

[0234] Optionally, the granularity of the downsampling or pre-compression includes but is not limited to at least one of the following:

[0235] Per-layer granularity, per-rank granularity.

[0236] Exemplarily, the method of downsampling or pre-compression can be a sub-band-based method, or the method of downsampling or pre-compression can be a port-based method, or the method of downsampling or pre-compression can be an AI model-based method, or the method of downsampling or pre-compression can be a DFT vector-based method.

[0237] Therefore, in the embodiments of the present application, the terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information; and the terminal performs compression processing on the second CSI information through a first AI model to obtain target compressed CSI information. That is, in the embodiments of the present application, the measured CSI information is subjected to downsampling or pre-compression processing before being input into the AI model, and then compressed through the AI model, thereby improving the CSI compression performance.

[0238] For the CSI compression and decompression method based on an AI model provided in the embodiments of the present application, the execution subject can be a CSI compression device based on an AI model or a CSI decompression device based on an AI model, or a processing unit in a CSI compression device based on an AI model or a CSI decompression device based on an AI model for executing the CSI compression and decompression method based on an AI model. In the embodiments of the present application, taking a CSI compression device based on an AI model or a CSI decompression device based on an AI model executing the CSI compression and decompression method based on an AI model as an example, the CSI compression device based on an AI model or the CSI decompression device based on an AI model provided in the embodiments of the present application is described.

[0239] Figure 8 Shows a schematic block diagram of a CSI compression device 300 based on an AI model according to an embodiment of the present application. As Figure 8 shown, the CSI compression device 300 based on an AI model includes:

[0240] A processing unit 310, configured to perform downsampling or pre-compression processing on the measured first channel state information CSI information to obtain second CSI information;

[0241] The processing unit 310 is further configured to perform compression processing on the second CSI information through a first artificial intelligence (AI) model to obtain target compressed CSI information;

[0242] The transceiver unit 320 is configured to send first information to a network-side device, where the first information includes the target compressed CSI information.

[0243] In some embodiments, the first information further includes at least one of the following: the payload of the target compressed CSI information, the first CSI information, channel quality indicator (CQI) information, rank indicator (RI) information, the second CSI information, and related information of downsampling or pre-compression corresponding to the second CSI information.

[0244] In some embodiments, when the CSI compression device 300 based on the AI model performs downsampling processing on the first CSI information, the subbands corresponding to the second CSI information are a subset of the subbands corresponding to the first CSI information, or the ports corresponding to the second CSI information are a subset of the ports corresponding to the first CSI information.

[0245] In some embodiments, the subbands corresponding to the second CSI information are the odd-numbered subbands corresponding to the first CSI information, or the subbands corresponding to the second CSI information are the even-numbered subbands corresponding to the first CSI information; or,

[0246] The ports corresponding to the second CSI information are the odd-numbered ports corresponding to the first CSI information, or the ports corresponding to the second CSI information are the even-numbered ports corresponding to the first CSI information.

[0247] In some embodiments, the subbands corresponding to the second CSI information are determined based on at least one of the following: a bit map associated with subband selection indicated by the network side, the correspondence between the channel state information reference signal (CSI-RS) bandwidth and the subband selection interval, the correspondence between RI and the subband selection interval, the correspondence between the output payload of the first AI model and the subband selection interval, and the correspondence between the input payload of the first AI model and the subband selection interval; or,

[0248] The ports corresponding to the second CSI information are determined based on at least one of the following: a bit map associated with port selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the port selection interval, the correspondence between RI and the port selection interval, the correspondence between the output payload of the first AI model and the port selection interval, and the correspondence between the input payload of the first AI model and the port selection interval.

[0249] In some embodiments, the subbands corresponding to the second CSI information are determined based on at least one of the following: a first measurement, a specific AI model;

[0250] Wherein, the first measurement includes at least one of the following: the received power of the CSI-RS on the subband corresponding to the first CSI, the signal-to-noise ratio or signal-to-interference-plus-noise ratio on the subband corresponding to the first CSI, the channel or channel characteristics on the subband corresponding to the first CSI.

[0251] In some embodiments, the first information further includes at least one of the following:

[0252] The subband information corresponding to the second CSI information, the port information corresponding to the second CSI information.

[0253] In some embodiments, when the CSI compression device 300 based on the AI model performs pre-compression processing on the first CSI information, the second CSI information is characterized by K eigenvectors, and the value of K satisfies at least one of the following: K < N, K < B;

[0254] Wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of subbands corresponding to the first CSI information, and K, N, and B are all positive integers.

[0255] In some embodiments, the second CSI information includes at least one of the following: K discrete Fourier transform (DFT) vectors, K eigenvectors in type II CSI compression, K eigenvectors in enhanced type II CSI compression, the amplitude or phase coefficients of K DFT vectors, the amplitude or phase coefficients of K-1 DFT vectors.

[0256] In some embodiments, the K eigenvectors are determined based on at least one of the following:

[0257] A bit map associated with eigenvector selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the eigenvectors, the correspondence between the ports and the eigenvectors, the correspondence between the RI and the eigenvectors, the correspondence between the output load of the first AI model and the eigenvectors, the correspondence between the input load of the first AI model and the eigenvectors.

[0258] In some embodiments, the first information further includes the K eigenvectors.

[0259] In some embodiments, the transceiver unit 320 is further configured to send second information to the network side device;

[0260] Wherein, the second information includes at least one of the following:

[0261] The first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, and the association relationship between the second CSI information and the first information.

[0262] In some embodiments, after downsampling or pre-compression processing, the dimensions of the input information of different AI models are different, or the input loads of different AI models are different, or the output loads of different AI models are different.

[0263] In some embodiments, before the AI model-based CSI compression device 300 performs downsampling or pre-compression processing on the first CSI information, the transceiver unit 320 is further configured to receive third information from the network-side device; or the transceiver unit 320 is further configured to send third information to the network-side device;

[0264] Wherein, the third information includes at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression-related information respectively associated with the at least one AI model, and dataset information respectively associated with the at least one AI model for training or updating or adjusting;

[0265] Wherein, the input-related information of the i-th AI model in the at least one AI model includes at least one of the following: the input load size of the i-th AI model, the input load type of the i-th AI model, and the input dimension of the i-th AI model;

[0266] Wherein, the output-related information of the i-th AI model in the at least one AI model includes at least one of the following: the output load size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output load type of the i-th AI model, and the output dimension of the i-th AI model;

[0267] Wherein, the at least one AI model includes the first AI model.

[0268] In some embodiments, the dataset information is associated with at least one of the following:

[0269] The load of the target compressed CSI information, the dimension information of the second CSI information, the number of sub-bands of the second CSI information, and the number of ports of the second CSI information.

[0270] In some embodiments, before the AI model-based CSI compression device 300 performs downsampling or pre-compression processing on the first CSI information, the transceiver unit 320 is further configured to send fourth information to the network-side device;

[0271] Wherein, the fourth information includes at least one of the following: whether the AI model-based CSI compression device 300 supports downsampling or pre-compression, relevant information on the downsampling or pre-compression supported by the AI model-based CSI compression device 300, candidate values of the output load of the AI model for CSI compression supported by the AI model-based CSI compression device 300, candidate values of the input load of the AI model for CSI compression supported by the AI model-based CSI compression device 300, relevant information on the AI model for CSI compression supported by the AI model-based CSI compression device 300, and candidate values of the input dimension of the AI model for CSI compression supported by the AI model-based CSI compression device 300.

[0272] In some embodiments, the relevant information on the downsampling or pre-compression includes at least one of the following: the granularity of the downsampling or pre-compression, the method of the downsampling or pre-compression, the dimension of the downsampling or pre-compression and the corresponding super-resolution dimension, and the relationship between the downsampling or pre-compression and the first CSI information.

[0273] In some embodiments, the granularity of the downsampling or pre-compression includes at least one of the following: layer granularity, rank granularity.

[0274] In some embodiments, the processing unit 310 is specifically configured to:

[0275] Perform downsampling or pre-compression processing on the first CSI information according to a specific AI model to obtain the second CSI information;

[0276] Wherein, the dimension information of the second CSI information is smaller than the dimension information of the first CSI information.

[0277] In some embodiments, the first CSI information includes at least one of the following:

[0278] Precoding matrix, precoding matrix indicating PMI, channel information.

[0279] In some embodiments, the above transceiver unit 320 may be a communication interface or transceiver, or an input / output interface of a communication chip or system-on-chip. The processing unit 310 may be embedded in or independent of the processor of the terminal in the form of hardware.

[0280] It should be understood that the AI model-based CSI compression device 300 according to the embodiments of the present application may correspond to the terminal in the method embodiments of the present application, and each unit in the AI model-based CSI compression device 300 respectively Figure 5 To implement the corresponding processes of the terminal in the method 200 shown, for the sake of brevity, it will not be elaborated here.

[0281] Therefore, in the embodiments of the present application, the terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information; and the terminal performs compression processing on the second CSI information through a first AI model to obtain target compressed CSI information. That is, in the embodiments of the present application, the measured CSI information is subjected to downsampling or pre-compression processing before being input into the AI model, and then compressed through the AI model, thereby improving the CSI compression performance.

[0282] Figure 9 FIG. 4 shows a schematic block diagram of a CSI decompression device 400 based on an AI model according to an embodiment of the present application.

[0283] As Figure 9 shown, the CSI decompression device 400 based on an AI model includes:

[0284] A transceiver unit 410, configured to receive first information from a terminal; wherein, the first information includes target compressed channel state information CSI information, and the target compressed CSI information is obtained by compressing second CSI information through a first AI model, and the second CSI information is obtained by performing downsampling or pre-compression processing on the measured first CSI information;

[0285] A processing unit 420, configured to perform decompression processing on the target compressed CSI information through a second artificial intelligence AI model to obtain third CSI information.

[0286] In some embodiments, the processing unit 420 is further configured to perform super-resolution processing on the third CSI information to obtain fourth CSI information.

[0287] In some embodiments, the first information includes at least one of the following: the payload of the target compressed CSI information, the first CSI information, channel quality indicator CQI information, rank indicator RI information, the second CSI information, and the relevant information of downsampling or pre-compression corresponding to the second CSI information.

[0288] In some embodiments, when the second CSI information is obtained by downsampling the first CSI information, the subband corresponding to the second CSI information is a subset of the subband corresponding to the first CSI information, or the port corresponding to the second CSI information is a subset of the port corresponding to the first CSI information.

[0289] In some embodiments, the subband corresponding to the second CSI information is the odd subband corresponding to the first CSI information, or the subband corresponding to the second CSI information is the even subband corresponding to the first CSI information; or,

[0290] The port corresponding to the second CSI information is the odd - numbered port corresponding to the first CSI information, or the port corresponding to the second CSI information is the even - numbered port corresponding to the first CSI information.

[0291] In some embodiments, the sub - band corresponding to the second CSI information is determined based on at least one of the following: the bit - map associated with sub - band selection indicated by the AI - model - based CSI decompression device 400, the correspondence between the channel state information reference signal CSI - RS bandwidth and the sub - band selection interval, the correspondence between RI and the sub - band selection interval, the correspondence between the output load of the first AI model and the sub - band selection interval, the correspondence between the input load of the first AI model and the sub - band selection interval; or,

[0292] The port corresponding to the second CSI information is determined based on at least one of the following: the bit - map associated with port selection indicated by the AI - model - based CSI decompression device 400, the correspondence between the CSI - RS bandwidth and the port selection interval, the correspondence between RI and the port selection interval, the correspondence between the output load of the first AI model and the port selection interval, the correspondence between the input load of the first AI model and the port selection interval.

[0293] In some embodiments, the sub - band corresponding to the second CSI information is determined based on at least one of the following: the first measurement, the selection result of the sub - band corresponding to the first CSI information by a specific AI model;

[0294] Wherein, the first measurement includes at least one of the following: the received power of the CSI - RS on the sub - band corresponding to the first CSI, the signal - to - noise ratio or signal - to - interference - plus - noise ratio on the sub - band corresponding to the first CSI, the channel or channel characteristics on the sub - band corresponding to the first CSI.

[0295] In some embodiments, the first information further includes at least one of the following:

[0296] The sub - band information corresponding to the second CSI information, the port information corresponding to the second CSI information.

[0297] In some embodiments, when the second CSI information is obtained by pre - compressing the first CSI information, the second CSI information is characterized by K eigenvectors, and the value of K satisfies at least one of the following: K < N, K < B;

[0298] Wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of sub - bands corresponding to the first CSI information, and K, N, and B are all positive integers.

[0299] In some embodiments, the second CSI information includes at least one of the following: K discrete Fourier transform (DFT) vectors, K eigenvectors in type II CSI compression, K eigenvectors in enhanced type II CSI compression, the amplitude or phase coefficients of K DFT vectors, and the amplitude or phase coefficients of K-1 DFT vectors.

[0300] In some embodiments, the K eigenvectors are determined based on at least one of the following:

[0301] The bitmap associated with eigenvector selection indicated by the CSI decompression device 400 based on the AI model, the correspondence between the CSI-RS bandwidth and the eigenvectors, the correspondence between the ports and the eigenvectors, the correspondence between the RI and the eigenvectors, the correspondence between the output load of the first AI model and the eigenvectors, and the correspondence between the input load of the first AI model and the eigenvectors.

[0302] In some embodiments, the first information further includes the K eigenvectors.

[0303] In some embodiments, the transceiver unit 410 is further configured to receive second information from the terminal;

[0304] Wherein, the second information includes at least one of the following:

[0305] The first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, and the association relationship between the second CSI information and the first information.

[0306] In some embodiments, the processing unit 420 is further configured to perform at least one of the following according to the first information:

[0307] Monitor the performance of the first AI model;

[0308] Monitor the performance of the second AI model;

[0309] Monitor the performance of the fourth CSI information obtained after the super-resolution processing of the third CSI information;

[0310] Monitor the performance of the second CSI information obtained after the downsampling or pre-compression processing of the first CSI information.

[0311] In some embodiments, after the downsampling or pre-compression processing, the dimensions of the input information of different AI models are different, or the input loads of different AI models are different, or the output loads of different AI models are different.

[0312] In some embodiments, before the AI model-based CSI decompression device 400 receives the first information from the terminal, the transceiver unit 410 is further configured to send third information to the terminal; or, the transceiver unit 410 is further configured to receive third information from the terminal;

[0313] Wherein, the third information includes at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression related information respectively associated with the at least one AI model, and dataset information respectively associated with the at least one AI model for training or updating or adjusting;

[0314] Wherein, the input-related information of the i-th AI model among the at least one AI models includes at least one of the following: the input payload size of the i-th AI model, the input payload type of the i-th AI model, and the input dimension of the i-th AI model;

[0315] Wherein, the output-related information of the i-th AI model among the at least one AI models includes at least one of the following: the output payload size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output payload type of the i-th AI model, and the output dimension of the i-th AI model;

[0316] Wherein, the at least one AI model includes the first AI model.

[0317] In some embodiments, the dataset information is associated with at least one of the following:

[0318] The payload of the target compressed CSI information, the dimension information of the second CSI information, the number of subbands of the second CSI information, and the number of ports of the second CSI information.

[0319] In some embodiments, before the AI model-based CSI decompression device 400 receives the first information from the terminal, the transceiver unit 410 is further configured to receive fourth information from the terminal;

[0320] Wherein, the fourth information includes at least one of the following:

[0321] Whether the terminal supports downsampling or pre-compression, the related information of the downsampling or pre-compression supported by the terminal, the candidate values of the output payload of the AI model for CSI compression supported by the terminal, the candidate values of the input payload of the AI model for CSI compression supported by the terminal, the related information of the AI model for CSI compression supported by the terminal, and the candidate values of the input dimension of the AI model for CSI compression supported by the terminal.

[0322] In some embodiments, the relevant information of the downsampling or pre - compression includes at least one of the following:

[0323] The granularity of the downsampling or pre - compression, the way of the downsampling or pre - compression, the dimension of the downsampling or pre - compression and the corresponding super - resolution dimension, and the relationship between the downsampling or pre - compression and the first CSI information.

[0324] In some embodiments, the granularity of the downsampling or pre - compression includes at least one of the following: layer granularity, rank granularity.

[0325] In some embodiments, the first CSI information includes at least one of the following:

[0326] Precoding matrix, precoding matrix indicator PMI, channel information.

[0327] In some embodiments, the above - mentioned transceiver unit 410 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 420 may be embedded in or independent of the processor of the network - side device in hardware form.

[0328] It should be understood that the CSI decompression device 400 based on the AI model according to the embodiments of the present application may correspond to the network - side device in the method embodiments of the present application, and each unit in the CSI decompression device 400 based on the AI model respectively Figure 5 to implement the corresponding processes of the network - side device in the method 200 shown. For the sake of brevity, details are not described here again.

[0329] Therefore, in the embodiments of the present application, the terminal performs downsampling or pre - compression processing on the measured first CSI information to obtain second CSI information; and the terminal performs compression processing on the second CSI information through the first AI model to obtain the target compressed CSI information. That is, in the embodiments of the present application, the measured CSI information is subjected to downsampling or pre - compression processing before being input into the AI model, and then compressed through the AI model, thereby improving the CSI compression performance.

[0330] The CSI compression device or CSI decompression device based on an AI model in the embodiments of the present application may be an electronic device, such as an electronic device with an operating system, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal or a network-side device, or other devices other than terminals or network-side devices. Exemplarily, the terminal may include, but is not limited to, the types of the above-listed terminal 11, the network-side device may include, but is not limited to, the types of the above-listed network-side device 12, and other devices may be servers, Network Attached Storage (NAS), etc., which are not specifically limited in the embodiments of the present application.

[0331] The CSI compression device or CSI decompression device based on an AI model provided in the embodiments of the present application can implement Figure 5 each process implemented in the method embodiments and achieve the same technical effects. To avoid repetition, details are not described here again.

[0332] As Figure 10 shown, the embodiments of the present application further provide a communication device 500, including a processor 501 and a memory 502, and a program or instruction that can run on the processor 501 is stored on the memory 502.

[0333] For example, when the communication device 500 is a terminal, when the program or instruction is executed by the processor 501, it implements each step executed by the terminal in the above-mentioned CSI compression and decompression method embodiments based on an AI model, and can achieve the same technical effects. To avoid repetition, details are not described here again.

[0334] Again, for example, when the communication device 500 is a network-side device, when the program or instruction is executed by the processor 501, it implements each step executed by the network-side device in the above-mentioned CSI compression and decompression method embodiments based on an AI model, and can achieve the same technical effects. To avoid repetition, details are not described here again.

[0335] The embodiments of the present application further provide a terminal, including a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps executed by the terminal in the method embodiments as Figure 5 shown. This terminal embodiment corresponds to the above-mentioned terminal-side method embodiments. Each implementation process and implementation manner of the above method embodiments can be applied to this terminal embodiment, and the same technical effects can be achieved.

[0336] Specifically, Figure 11 is a schematic diagram of the hardware structure of a terminal for implementing the embodiments of the present application.

[0337] The terminal 600 includes, but is not limited to, at least some components such as a radio frequency unit 601, a network module 602, an audio output unit 603, an input unit 604, a sensor 605, a display unit 606, a user input unit 607, an interface unit 608, a memory 609, and a processor 610.

[0338] Those skilled in the art can understand that the terminal 600 may further include a power supply (such as a battery) for powering each component. The power supply may be logically connected to the processor 610 through a power management system, so as to implement functions such as management of charging, discharging, and power consumption management through the power management system. Figure 11 The terminal structure shown does not limit the terminal. The terminal may include more or fewer components than shown, or combine some components, or have different component arrangements, which will not be elaborated here.

[0339] It should be understood that in the embodiments of the present application, the input unit 604 may include a graphics processing unit (GPU) 6041 and a microphone 6042. The graphics processor 6041 processes the image data of a still picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 606 may include a display panel 6061, and the display panel 6061 may be configured in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit 607 includes at least one of a touch panel 6071 and other input devices 6072. The touch panel 6071 is also called a touch screen. The touch panel 6071 may include two parts: a touch detection device and a touch controller. The other input devices 6072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be elaborated here.

[0340] In the embodiments of the present application, after receiving downlink data from a network-side device, the radio frequency unit 601 may transmit it to the processor 610 for processing; in addition, the radio frequency unit 601 may send uplink data to the network-side device. Generally, the radio frequency unit 601 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low-noise amplifier, a duplexer, etc.

[0341] The memory 609 can be used to store software programs or instructions as well as various data. The memory 609 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data. Among them, the first storage area may store an operating system, application programs or instructions required for at least one function (such as a sound playback function, an image playback function, etc.). In addition, the memory 609 may include volatile memory or non-volatile memory. 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), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDR SDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM), and a direct rambus random access memory (DRRAM). The memory 609 in the embodiments of the present application includes but is not limited to these and any other suitable types of memory.

[0342] The processor 610 may include at least one processing unit; optionally, the processor 610 integrates an application processor and a modulation and demodulation processor. Among them, the application processor mainly processes operations related to the operating system, user interface, and application programs, etc., and the modulation and demodulation processor mainly processes transmission signals in a multi-connection scenario, such as a baseband processor. It can be understood that the above modulation and demodulation processor may not be integrated into the processor 610 either.

[0343] Among them, the processor 610 is used to perform downsampling or pre-compression processing on the measured first channel state information (CSI) to obtain second CSI information; the processor 610 is also used to perform compression processing on the second CSI information through a first artificial intelligence (AI) model to obtain target compressed CSI information;

[0344] Among them, the radio frequency unit 601 is used to send first information to a network-side device, where the first information includes the target compressed CSI information.

[0345] It can be understood that the implementation processes of the various implementation manners mentioned in this embodiment may refer to the relevant descriptions of the method embodiment and achieve the same or corresponding technical effects. To avoid repetition, they will not be elaborated here.

[0346] The embodiment of the present application further provides a network-side device, including a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps performed by the network-side device in the method embodiment as Figure 5 shown. This network-side device embodiment corresponds to the above network-side device method embodiment. Each implementation process and implementation manner of the above method embodiment can be applied to this network-side device embodiment and can achieve the same technical effect. For the sake of brevity, they will not be elaborated here.

[0347] Specifically, the embodiment of the present application further provides a network-side device. As Figure 12 shown, the network-side device 700 includes: an antenna 71, a radio frequency device 72, a baseband device 73, a processor 74, and a memory 75. The antenna 71 is connected to the radio frequency device 72. In the uplink direction, the radio frequency device 72 receives information through the antenna 71 and sends the received information to the baseband device 73 for processing. In the downlink direction, the baseband device 73 processes the information to be sent and sends it to the radio frequency device 72. After processing the received information, the radio frequency device 72 sends it out through the antenna 71.

[0348] The method executed by the network-side device in the above embodiment can be implemented in the baseband device 73, and the baseband device 73 includes a baseband processor.

[0349] The baseband device 73 may include, for example, at least one baseband board, and at least two chips are arranged on the baseband board. As Figure 12 shown, one of the chips is, for example, a baseband processor, which is connected to the memory 75 through a bus interface to call the program in the memory 75 and execute the operations of the network device shown in the above method embodiment.

[0350] The network-side device may further include a network interface 76, and this interface is, for example, a Common Public Radio Interface (CPRI).

[0351] Specifically, the network-side device 700 in the embodiment of the present application further includes: instructions or programs stored on the memory 75 and executable on the processor 74. The processor 74 calls the instructions or programs in the memory 75 to execute the method performed by each unit shown in Figure 9 shown and achieve the same technical effect. To avoid repetition, they will not be elaborated here.

[0352] An embodiment of the present application further provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements each process of the above embodiment of the CSI compression and decompression method based on the AI model and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0353] Among them, the processor is the processor in the terminal described in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk, or an optical disc, etc. In some examples, the readable storage medium may be a non-transitory readable storage medium.

[0354] Another embodiment of the present application provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run a program or instruction to implement each process of the above embodiment of the CSI compression and decompression method based on the AI model and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0355] It should be understood that the chip mentioned in the embodiment of the present application may also be referred to as a system-on-chip, a system chip, a chip system, or a system-on-chip.

[0356] Another embodiment of the present application provides a computer program / program product, which is stored in a storage medium. The computer program / program product is executed by at least one processor to implement each process of the above embodiment of the CSI compression and decompression method based on the AI model and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0357] An embodiment of the present application further provides a communication system, including: a terminal and a network-side device. Among them, the terminal can be used to execute the steps executed by the terminal in the above CSI compression and decompression method based on the AI model, and the network-side device can be used to execute the steps executed by the network-side device in the above CSI compression and decompression method based on the AI model.

[0358] It should be noted that in this text, the term "including", "comprising", or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a series of elements not only includes those elements but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, method, article, or apparatus that includes such an element. In addition, it should be pointed out that the scope of the methods and apparatuses in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted, or combined. Additionally, the features described with reference to certain examples may be combined in other examples.

[0359] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of a computer software product plus a necessary general hardware platform, and of course, they can also be implemented by hardware. This computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions for causing a terminal or a network-side device to execute the methods described in various embodiments of the present application.

[0360] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms of embodiments without departing from the purpose of the present application and the scope protected by the claims. These embodiments are all within the protection scope of the present application.

Claims

1. A CSI compression method based on an AI model, characterized in that, Including: The terminal performs downsampling or pre-compression processing on the measured first Channel State Information (CSI) to obtain second CSI; The terminal compresses the second CSI through a first Artificial Intelligence (AI) model to obtain target compressed CSI; The terminal sends first information to a network-side device, where the first information includes the target compressed CSI.

2. The method according to claim 1, wherein: The first information further includes at least one of the following: the payload of the target compressed CSI, the first CSI, Channel Quality Indicator (CQI) information, Rank Indicator (RI) information, the second CSI, and the relevant information of downsampling or pre-compression corresponding to the second CSI.

3. The method according to claim 1 or 2, wherein: When the terminal performs downsampling on the first CSI, the sub-bands corresponding to the second CSI are a subset of the sub-bands corresponding to the first CSI, or the ports corresponding to the second CSI are a subset of the ports corresponding to the first CSI.

4. The method according to claim 3, wherein: The sub-bands corresponding to the second CSI are the odd-numbered sub-bands corresponding to the first CSI, or the sub-bands corresponding to the second CSI are the even-numbered sub-bands corresponding to the first CSI; or The ports corresponding to the second CSI are the odd-numbered ports corresponding to the first CSI, or the ports corresponding to the second CSI are the even-numbered ports corresponding to the first CSI.

5. The method according to claim 3 or 4, wherein: The sub-bands corresponding to the second CSI are determined based on at least one of the following: a bitmap associated with sub-band selection indicated by the network side, the correspondence between the CSI Reference Signal (CSI-RS) bandwidth and the sub-band selection interval, the correspondence between RI and the sub-band selection interval, the correspondence between the output payload of the first AI model and the sub-band selection interval, and the correspondence between the input payload of the first AI model and the sub-band selection interval; or The ports corresponding to the second CSI are determined based on at least one of the following: a bitmap associated with port selection indicated by the network side, the correspondence between the CSI-RS bandwidth and the port selection interval, the correspondence between RI and the port selection interval, the correspondence between the output payload of the first AI model and the port selection interval, and the correspondence between the input payload of the first AI model and the port selection interval.

6. The method according to claim 3 or 4, wherein: The sub-bands corresponding to the second CSI are determined based on at least one of the following: first measurement, a specific AI model; wherein the first measurement includes at least one of the following: the received power of the CSI-RS on the sub-bands corresponding to the first CSI, the signal-to-noise ratio or signal-to-interference-plus-noise ratio on the sub-bands corresponding to the first CSI, and the channel or channel characteristics on the sub-bands corresponding to the first CSI.

7. The method according to any one of claims 3 to 6, wherein: The first information further includes at least one of the following: subband information corresponding to the second CSI information, port information corresponding to the second CSI information.

8. The method according to claim 1 or 2, wherein when the terminal performs pre-compression processing on the first CSI information, the second CSI information is represented by K eigenvectors, and the value of K satisfies at least one of the following: K < N, K < B; wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of subbands corresponding to the first CSI information, and K, N, and B are all positive integers.

9. The method according to claim 8, wherein the second CSI information includes at least one of the following: K discrete Fourier transform (DFT) vectors, K eigenvectors in type II CSI compression, K eigenvectors in enhanced type II CSI compression, amplitude or phase coefficients of K DFT vectors, amplitude or phase coefficients of K - 1 DFT vectors.

10. The method according to claim 8 or 9, wherein the K eigenvectors are determined based on at least one of the following: a bit map associated with eigenvector selection indicated by the network side, the correspondence between the CSI-RS bandwidth and eigenvectors, the correspondence between ports and eigenvectors, the correspondence between RI and eigenvectors, the correspondence between the output load of the first AI model and eigenvectors, the correspondence between the input load of the first AI model and eigenvectors.

11. The method according to any one of claims 8 to 10, wherein the first information further includes the K eigenvectors.

12. The method according to any one of claims 1 to 11, characterized in that, The method further includes: the terminal sends second information to the network side device; wherein, the second information includes at least one of the following: the first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, the association relationship between the second CSI information and the first information.

13. The method according to any one of claims 1 to 12, wherein after downsampling or pre-compression processing, the dimensions of the input information of different AI models are different, or the input loads of different AI models are different, or the output loads of different AI models are different.

14. The method according to any one of claims 1 to 13, wherein before the terminal performs downsampling or pre-compression processing on the first CSI information, the method further includes: the terminal receives third information from the network side device; or, the terminal sends third information to the network side device; wherein, the third information includes at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression related information respectively associated with the at least one AI model, dataset information respectively associated with the at least one AI model for training or updating or adjustment; Among them, the input-related information of the i-th AI model among the at least one AI model includes at least one of the following: the input payload size of the i-th AI model, the input payload type of the i-th AI model, the input dimension of the i-th AI model; Among them, the output-related information of the i-th AI model among the at least one AI model includes at least one of the following: the output payload size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output payload type of the i-th AI model, the output dimension of the i-th AI model; Among them, the at least one AI model includes the first AI model.

15. The method according to claim 14, wherein The dataset information is associated with at least one of the following: The payload of the target compressed CSI information, the dimension information of the second CSI information, the number of subbands of the second CSI information, the number of ports of the second CSI information.

16. The method according to any one of claims 1 to 15, wherein Before the terminal performs downsampling or pre-compression processing on the first CSI information, the method further includes: The terminal sends fourth information to the network-side device; Among them, the fourth information includes at least one of the following: whether the terminal supports downsampling or pre-compression, the relevant information of the downsampling or pre-compression supported by the terminal, the candidate values of the output payload of the AI model for CSI compression supported by the terminal, the candidate values of the input payload of the AI model for CSI compression supported by the terminal, the relevant information of the AI model for CSI compression supported by the terminal, the candidate values of the input dimension of the AI model for CSI compression supported by the terminal.

17. The method according to any one of claims 1 to 16, wherein The relevant information of the downsampling or pre-compression includes at least one of the following: the granularity of the downsampling or pre-compression, the method of the downsampling or pre-compression, the downsampling or pre-compression dimension and the corresponding super-resolution dimension, the relationship between the downsampling or pre-compression and the first CSI information.

18. The method according to claim 17, wherein The granularity of the downsampling or pre-compression includes at least one of the following: Layer granularity, rank granularity.

19. The method according to any one of claims 1 to 18, wherein The terminal performs downsampling or pre-compression processing on the measured first CSI information to obtain second CSI information, including: The terminal performs downsampling or pre-compression processing on the first CSI information according to a specific AI model to obtain the second CSI information; Among them, the dimension information of the second CSI information is smaller than the dimension information of the first CSI information.

20. The method according to any one of claims 1 to 19, wherein The first CSI information includes at least one of the following: Precoding matrix, precoding matrix indicator PMI, channel information.

21. A CSI decompression method based on an AI model, characterized in that, Including: The network-side device receives first information from the terminal; wherein, the first information includes target compressed channel state information (CSI) information, and the target compressed CSI information is obtained by compressing second CSI information through a first AI model, and the second CSI information is obtained by downsampling or pre-compressing the measured first CSI information; The network-side device decompresses the target compressed CSI information through a second artificial intelligence (AI) model to obtain third CSI information.

22. The method according to claim 21, wherein The method further includes: The network-side device performs super-resolution processing on the third CSI information to obtain fourth CSI information.

23. The method according to claim 21 or 22, wherein The first information includes at least one of the following: the payload of the target compressed CSI information, the first CSI information, channel quality indicator (CQI) information, rank indicator (RI) information, the second CSI information, and the relevant information of the downsampling or pre-compression corresponding to the second CSI information.

24. The method according to any one of claims 21 to 23, wherein When the second CSI information is obtained by downsampling the first CSI information, the subband corresponding to the second CSI information is a subset of the subband corresponding to the first CSI information, or the port corresponding to the second CSI information is a subset of the port corresponding to the first CSI information.

25. The method according to claim 24, wherein The subband corresponding to the second CSI information is the odd-numbered subband corresponding to the first CSI information, or the subband corresponding to the second CSI information is the even-numbered subband corresponding to the first CSI information; or The port corresponding to the second CSI information is the odd-numbered port corresponding to the first CSI information, or the port corresponding to the second CSI information is the even-numbered port corresponding to the first CSI information.

26. The method according to claim 24 or 25, wherein The subband corresponding to the second CSI information is determined based on at least one of the following: the bitmap associated with subband selection indicated by the network-side device, the correspondence between the channel state information reference signal (CSI-RS) bandwidth and the subband selection interval, the correspondence between RI and the subband selection interval, the correspondence between the output payload of the first AI model and the subband selection interval, and the correspondence between the input payload of the first AI model and the subband selection interval; or The port corresponding to the second CSI information is determined based on at least one of the following: the bitmap associated with port selection indicated by the network-side device, the correspondence between the CSI-RS bandwidth and the port selection interval, the correspondence between RI and the port selection interval, the correspondence between the output payload of the first AI model and the port selection interval, and the correspondence between the input payload of the first AI model and the port selection interval.

27. The method according to claim 24 or 25, wherein The sub - band corresponding to the second CSI information is determined based on at least one of the following: the first measurement, a specific AI model; Wherein, the first measurement includes at least one of the following: the received power of the CSI - RS on the sub - band corresponding to the first CSI, the signal - to - noise ratio or signal - to - interference - plus - noise ratio on the sub - band corresponding to the first CSI, the channel or channel characteristics on the sub - band corresponding to the first CSI.

28. The method according to any one of claims 24 to 27, characterized in that The first information further includes at least one of the following: The sub - band information corresponding to the second CSI information, the port information corresponding to the second CSI information.

29. The method according to any one of claims 21 to 23, characterized in that When the second CSI information is obtained by pre - compressing the first CSI information, the second CSI information is characterized by K eigen - vectors, and the value of K satisfies at least one of the following: K < N, K < B; Wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of sub - bands corresponding to the first CSI information, and K, N, and B are all positive integers.

30. The method according to claim 29, characterized in that The second CSI information includes at least one of the following: K discrete Fourier transform (DFT) vectors, K eigen - vectors in type II CSI compression, K eigen - vectors in enhanced type II CSI compression, the amplitude or phase coefficients of K DFT vectors, the amplitude or phase coefficients of K - 1 DFT vectors.

31. The method according to claim 29 or 30, characterized in that The K eigen - vectors are determined based on at least one of the following: The bit - map associated with eigen - vector selection indicated by the network - side device, the correspondence between the CSI - RS bandwidth and eigen - vectors, the correspondence between ports and eigen - vectors, the correspondence between RI and eigen - vectors, the correspondence between the output load of the first AI model and eigen - vectors, the correspondence between the input load of the first AI model and eigen - vectors.

32. The method according to any one of claims 29 to 31, characterized in that The first information further includes the K eigen - vectors.

33. The method according to any one of claims 21 to 32, characterized in that, The method further includes: The network - side device receives second information from the terminal; Wherein, the second information includes at least one of the following: The first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, the association relationship between the second CSI information and the first information.

34. The method according to any one of claims 21 to 33, characterized in that, The method further includes: The network - side device performs at least one of the following according to the first information: Monitor the performance of the first AI model; Monitor the performance of the second AI model; Monitor the performance of the fourth CSI information obtained after super - resolution processing of the third CSI information; Monitor the performance of the second CSI information obtained after down - sampling or pre - compressing the first CSI information.

35. The method according to any one of claims 21 to 34, characterized in that After downsampling or pre-compression processing, the dimensions of the input information of different AI models are different, or the input loads of different AI models are different, or the output loads of different AI models are different.

36. The method according to any one of claims 21 to 35, characterized in that Before the network-side device receives the first information from the terminal, the method further includes: The network-side device sends third information to the terminal; or The network-side device receives third information from the terminal; Wherein, the third information includes at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression related information respectively associated with the at least one AI model, dataset information respectively associated with the at least one AI model for training or updating or adjusting; Wherein, the input-related information of the i-th AI model in the at least one AI model includes at least one of the following: the input load size of the i-th AI model, the input load type of the i-th AI model, the input dimension of the i-th AI model; Wherein, the output-related information of the i-th AI model in the at least one AI model includes at least one of the following: the output load size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output load type of the i-th AI model, the output dimension of the i-th AI model; Wherein, the at least one AI model includes the first AI model.

37. The method according to claim 36, characterized in that The dataset information is associated with at least one of the following: The load of the target compressed CSI information, the dimension information of the second CSI information, the number of sub-bands of the second CSI information, the number of ports of the second CSI information.

38. The method according to any one of claims 21 to 37, characterized in that Before the network-side device receives the first information from the terminal, the method further includes: The network-side device receives fourth information from the terminal; Wherein, the fourth information includes at least one of the following: Whether the terminal supports downsampling or pre-compression, the relevant information of the downsampling or pre-compression supported by the terminal, the candidate values of the output load of the AI model for CSI compression supported by the terminal, the candidate values of the input load of the AI model for CSI compression supported by the terminal, the relevant information of the AI model for CSI compression supported by the terminal, the candidate values of the input dimension of the AI model for CSI compression supported by the terminal.

39. The method according to any one of claims 21 to 38, characterized in that The relevant information of the downsampling or pre-compression includes at least one of the following: The granularity of downsampling or pre-compression, the method of downsampling or pre-compression, the downsampling or pre-compression dimension and the corresponding super-resolution dimension, the relationship between downsampling or pre-compression and the first CSI information.

40. The method according to claim 39, characterized in that The granularity of the downsampling or pre-compression includes at least one of the following: Layer granularity, rank granularity.

41. The method according to any one of claims 21 to 40, characterized in that The first CSI information includes at least one of the following: Precoding matrix, precoding matrix indicator PMI, channel information.

42. A CSI compression device based on an AI model, characterized in that, Including: A processing unit, configured to perform downsampling or pre-compression processing on the measured first channel state information CSI information to obtain second CSI information; The processing unit is further configured to perform compression processing on the second CSI information through a first artificial intelligence AI model to obtain target compressed CSI information; A transceiver unit, configured to send first information to a network-side device, where the first information includes the target compressed CSI information.

43. The apparatus according to claim 42, characterized in that The first information further includes at least one of the following: the payload of the target compressed CSI information, the first CSI information, channel quality indicator CQI information, rank indicator RI information, the second CSI information, and the relevant information of the downsampling or pre-compression corresponding to the second CSI information.

44. The apparatus according to claim 42 or 43, characterized in that When the CSI compression apparatus based on the AI model performs downsampling processing on the first CSI information, the subband corresponding to the second CSI information is a subset of the subbands corresponding to the first CSI information, or the port corresponding to the second CSI information is a subset of the ports corresponding to the first CSI information.

45. The apparatus according to claim 44, characterized in that The first information further includes at least one of the following: The subband information corresponding to the second CSI information, the port information corresponding to the second CSI information.

46. The apparatus according to claim 42 or 43, characterized in that When the CSI compression apparatus based on the AI model performs pre-compression processing on the first CSI information, the second CSI information is characterized by K eigenvectors, and the value of K satisfies at least one of the following: K < N, K < B; Wherein, N represents the number of ports corresponding to the first CSI information, B represents the number of subbands corresponding to the first CSI information, and K, N, and B are all positive integers.

47. The apparatus according to claim 46, characterized in that The first information further includes the K eigenvectors.

48. The apparatus according to any one of claims 42 to 47, characterized in that The transceiver unit is further configured to send second information to the network-side device; Wherein, the second information includes at least one of the following: The first CSI information, the second CSI information, the association relationship between the first CSI information and the first information, and the association relationship between the second CSI information and the first information.

49. The apparatus according to any one of claims 42 to 48, characterized in that Before the CSI compression device based on the AI model performs downsampling or pre-compression processing on the first CSI information, the transceiver unit is further configured to receive third information from the network-side device; or, the transceiver unit is further configured to send third information to the network-side device; Wherein, the third information includes at least one of the following: input-related information of at least one AI model, output-related information of the at least one AI model, downsampling or pre-compression-related information respectively associated with the at least one AI model, and dataset information respectively associated with the at least one AI model for training or updating or adjusting; Wherein, the input-related information of the i-th AI model among the at least one AI models includes at least one of the following: the input payload size of the i-th AI model, the input payload type of the i-th AI model, and the input dimension of the i-th AI model; Wherein, the output-related information of the i-th AI model among the at least one AI models includes at least one of the following: the output payload size of the i-th AI model, the quantization method of the output information of the i-th AI model, the output payload type of the i-th AI model, and the output dimension of the i-th AI model; Wherein, the at least one AI model includes the first AI model.

50. The device according to any one of claims 42 to 49, wherein Before the CSI compression device based on the AI model performs downsampling or pre-compression processing on the first CSI information, the transceiver unit is further configured to send fourth information to the network-side device; Wherein, the fourth information includes at least one of the following: whether the CSI compression device based on the AI model supports downsampling or pre-compression, downsampling or pre-compression-related information supported by the CSI compression device based on the AI model, candidate values of the output payload of the AI model for CSI compression supported by the CSI compression device based on the AI model, candidate values of the input payload of the AI model for CSI compression supported by the CSI compression device based on the AI model, information about the AI model for CSI compression supported by the CSI compression device based on the AI model, and candidate values of the input dimension of the AI model for CSI compression supported by the CSI compression device based on the AI model.

51. A CSI decompression device based on an AI model, characterized in that, Comprising: A transceiver unit, configured to receive first information from a terminal; wherein, the first information includes target compressed channel state information CSI information, and the target compressed CSI information is obtained by compressing second CSI information through a first AI model, and the second CSI information is obtained by performing downsampling or pre-compression processing on the measured first CSI information; A processing unit, configured to decompress the target compressed CSI information through a second artificial intelligence AI model to obtain third CSI information.

52. The device according to claim 51, wherein The processing unit is further configured to perform super-resolution processing on the third CSI information to obtain fourth CSI information.

53. The device according to claim 51 or 52, wherein the first information further includes at least one of the following: the payload of the target compressed CSI information, the first CSI information, channel quality indicator (CQI) information, rank indicator (RI) information, the second CSI information, and the relevant information corresponding to the downsampling or pre-compression of the second CSI information.

54. The device according to any one of claims 51 to 53, wherein the processing unit is further configured to perform at least one of the following according to the first information: monitor the performance of the first AI model; monitor the performance of the second AI model; monitor the performance of the fourth CSI information obtained after the super-resolution processing of the third CSI information; monitor the performance of the second CSI information obtained after the downsampling or pre-compression processing of the first CSI information.

55. A terminal, characterized in that, comprising a transceiver, a processor, and a memory, the memory storing a program or instructions executable on the processor, and when the program or instructions are executed by the processor, the steps of the AI model-based CSI compression method according to any one of claims 1 to 20 are implemented.

56. A network-side device, characterized in that, comprising a transceiver, a processor, and a memory, the memory storing a program or instructions executable on the processor, and when the program or instructions are executed by the processor, the steps of the AI model-based CSI decompression method according to any one of claims 21 to 41 are implemented.

57. A readable storage medium, characterized in that, A program or instructions are stored on the readable storage medium, and when the program or instructions are executed by a processor, the steps of the AI model-based CSI compression method according to any one of claims 1-20 are implemented, or the steps of the AI model-based CSI decompression method according to any one of claims 21 to 41 are implemented.

Images