CN116982300A

CN116982300A - Signal processing method and receiver

Info

Publication number: CN116982300A
Application number: CN202180095411.1A
Authority: CN
Inventors: 肖寒
Original assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd
Current assignee: Guangdong Oppo Mobile Telecommunications Corp Ltd
Priority date: 2021-07-12
Filing date: 2021-07-12
Publication date: 2023-10-31
Also published as: WO2023283785A1

Abstract

A signal processing method and a receiver are provided. The method comprises the following steps: the receiver receives the wireless signal transmitted by the transmitter to obtain a received signal; the receiver inputs the received signal into an artificial intelligence AI decoder for decoding to obtain a decoded signal; the receiver generates a recovery signal of the received signal according to the decoded signal; the receiver performs on-line training on the AI decoder according to the difference between the recovery signal and the received signal, and the received signal and the recovery signal can be acquired by the receiver, so that the method can realize the on-line training process of the AI decoder so as to improve the generalization capability of the AI decoder.

Description

Signal processing method and receiver

Technical Field

The present application relates to the field of communications, and in particular, to a signal processing method and a receiver.

Background

A self encoder (AE) is an artificial intelligence (artificial intelligence, AI) model that targets an input signal, and that contains an architecture of the AI decoder that is naturally adapted to many architectures in a communication system. For example, the AI decoder may correspond to a receiver of a communication system.

Currently, in order to improve the accuracy of the decoding of the AI decoder, the AI decoder needs to be pre-trained based on a preset training set, also called "offline training", before the AI decoder is deployed to a receiver (i.e., online to receiver). However, the actual communication system situation is complex, the training data in the training set cannot cover all situations, and there is a large difference between the training data and the received signal actually received by the receiver, and if the AI decoder is trained offline based on the training set only, the generalization capability of the trained AI decoder is poor.

Disclosure of Invention

The application provides a signal processing method and a receiver, which are used for carrying out on-line training on an AI decoder and are beneficial to improving the generalization capability of the AI decoder.

In a first aspect, a method of signal processing is provided, comprising: the receiver receives the wireless signal transmitted by the transmitter to obtain a received signal; the receiver inputs the received signal into an artificial intelligence AI decoder for decoding to obtain a decoded signal; the receiver generates a recovery signal of the received signal according to the decoded signal; the receiver trains the AI decoder online according to the difference between the recovered signal and the received signal.

In a second aspect, there is provided a receiver comprising: the receiving unit is used for receiving the wireless signals transmitted by the transmitter to obtain received signals; the processing unit is used for inputting the received signal into an artificial intelligence AI decoder for decoding to obtain a decoded signal; the processing unit is used for generating a recovery signal of the received signal according to the decoded signal; the processing unit is used for carrying out online training on the AI decoder according to the difference between the recovery signal and the receiving signal.

In a third aspect, there is provided a receiver comprising a memory for storing a program and a processor for invoking the program in the memory to perform the method according to the first aspect.

In a fourth aspect, there is provided an apparatus comprising a processor for calling a program from a memory to perform the method of the first aspect.

In a fifth aspect, a chip is provided, comprising a processor for calling a program from a memory, so that a device on which the chip is mounted performs the method of the first aspect.

In a sixth aspect, there is provided a computer-readable storage medium having stored thereon a program that causes a computer to execute the method of the first aspect.

In a seventh aspect, there is provided a computer program product comprising a program for causing a computer to perform the method of the first aspect.

In an eighth aspect, there is provided a computer program for causing a computer to perform the method of the first aspect.

The receiver decodes the received signal by using the AI decoder to obtain a decoded signal, generates a recovery signal of the received signal according to the decoded signal, and performs on-line training on the AI decoder based on the difference between the received signal and the recovery signal.

Drawings

Fig. 1 is a flow chart of a transmission signal in a wireless communication system to which an embodiment of the present application is applied.

Fig. 2 is a schematic diagram of channel estimation and signal recovery applicable to an embodiment of the present application.

Fig. 3 is a block diagram of a neural network to which embodiments of the present application are applicable.

Fig. 4 is a block diagram of a CNN to which an embodiment of the present application is applied.

Fig. 5 is a schematic diagram of a process of channel estimation based on an AI decoder.

Fig. 6 is a schematic diagram of a self-encoder based CSI feedback system.

Fig. 7 is a schematic diagram of an AI decoder-based receiver.

Fig. 8 is a wireless communication system 800 to which embodiments of the present application are applicable.

Fig. 9 is a flowchart of a method of signal processing according to an embodiment of the present application.

Fig. 10 is a schematic diagram of a method of signal processing based on periodic online training according to an embodiment of the present application.

Fig. 11 is a schematic diagram of a method of signal processing based on non-persistent online training according to an embodiment of the present application.

Fig. 12 is a schematic diagram of a method of signal processing based on aperiodic online training according to an embodiment of the present application.

Fig. 13 is a schematic diagram of a signal processing method in a channel estimation process according to an embodiment of the present application.

Fig. 14 is a schematic diagram of a self-encoder deployed in a communication system.

Fig. 15 is a schematic diagram of a signal processing method in CSI feedback process according to an embodiment of the present application.

Fig. 16 is a schematic diagram of a signal processing method in a data transmission process according to an embodiment of the present application.

Fig. 17 is a schematic structural diagram of a receiver of an embodiment of the present application.

Fig. 18 is a schematic structural diagram of an apparatus for signal processing according to an embodiment of the present application.

Detailed Description

The technical scheme of the application will be described below with reference to the accompanying drawings. In order to facilitate understanding of the present application, terms and communication procedures related to embodiments of the present application are described below with reference to fig. 1 to 7.

1. Signal transmission procedure in a wireless communication system

Fig. 1 is a flow chart of a transmission signal in a wireless communication system to which an embodiment of the present application is applied. As shown in fig. 1, a signal transmission process in a wireless communication system can be roughly divided into various signal processing processes S111 to S118 shown in fig. 1. Some or all of the signal processing shown in fig. 1 may be implemented by a separate AI model, and a specific implementation thereof may be described with reference to fig. 5 to 7.

The transmitter performs channel coding on the information to be transmitted in the channel coding process S111, to obtain a coded code stream. Wherein the information to be transmitted may be in the form of a bit stream.

The code stream is modulated into modulation symbols in a modulation process S112.

In the pilot insertion process S113, pilot symbols are inserted into the modulation symbols to form a signal to be transmitted, where the pilot symbols may be used for channel estimation and symbol detection by a receiver.

In the transmission signal S114, the above signal is transmitted on a channel to a receiver. Where the signal typically adds noise during transmission through the channel.

In the channel estimation process S115, the receiver may perform channel estimation based on the pilot signal to obtain Channel State Information (CSI), and feed back the CSI to the transmitter through the feedback link, so that the transmitter may adjust modes such as channel coding, modulation, precoding, and the like.

In the symbol detection process S116, symbol detection is performed on the received modulation symbol, and a detection result is obtained.

In the demodulation process S117, the received modulation symbols are demodulated based on the detection result, resulting in a code stream.

In the channel decoding process S118, the code stream is decoded to obtain recovered information, where the recovered information may be in the form of a bit stream.

It should be understood that the signal processing procedures S111 to S118 shown in fig. 1 are only exemplary listed as signal processing procedures commonly found in a wireless communication system, and signal processing procedures such as resource mapping, precoding, interference cancellation, CSI measurement, etc. may be further included in the wireless communication system, and these signal processing procedures may also be implemented by a separate AI model. For brevity, the present application is not described in detail.

2. Channel estimation

Due to the complexity and time-varying nature of the wireless channel environment, in a wireless communication system (e.g., the wireless communication system described above), a receiver needs to recover the received signal based on the estimation of the channel. Fig. 2 is a schematic diagram of channel estimation and signal recovery applicable to an embodiment of the present application.

As shown in fig. 2, in step S210, the transmitter transmits a series of pilot signals known to the receiver, such as a channel state information reference signal (CSI-RS), a demodulation reference signal (demodulation reference signal, DMRS), etc., in addition to the data signals on the time-frequency resource.

In step S211, the transmitter transmits the above data signal and pilot signal to the transmitter through a channel.

In step S212, the receiver may perform channel estimation after receiving the pilot signal. In one possible implementation, the receiver may estimate channel information for the channel transmitting the pilot signal by a channel estimation algorithm (e.g., least squares (least squares method, LS) channel estimation) based on the pre-stored pilot sequence and the received pilot sequence.

In step S213, the receiver may recover the channel information on the full time-frequency resource by using an interpolation algorithm according to the channel information of the channel transmitting the pilot sequence, for subsequent CSI feedback or data recovery, etc.

3. CSI feedback

In a wireless communication system, a codebook-based scheme is mainly utilized to realize extraction and feedback of channel characteristics, namely, after a receiver performs channel estimation, a precoding matrix which is most matched with a current channel is selected from a preset precoding codebook according to a certain optimization criterion according to a channel estimation result, and precoding matrix index (precoding matrix index, PMI) information is fed back to a transmitter through an air-interface feedback link so as to realize precoding by the transmitter. In some implementations, the receiver may also feed back the measured channel quality indication (channel quality indication, CQI) to the transmitter for the transmitter to implement adaptive modulation coding, etc.

4. Self-encoder

A self-encoder is a neural network that targets an input signal to be trained, and that contains an architecture of an AI encoder and/or AI decoder that is naturally adapted to many architectures in a communication system. For example, the AI encoder and AI decoder may correspond to a transmitter and receiver, respectively, of a wireless communication system. For another example, the AI encoder and the AI decoder may also correspond to a channel compression module and a decompression module, respectively, in the CSI feedback process. For another example, the AI decoder in the self-encoder may also be used separately in the channel estimation process for the recovery of channel information by the receiver. The following description will be made with reference to fig. 5 to 7, and will not be repeated here for the sake of brevity.

Typically, the self-encoder may be trained based on a training set prior to deployment of the self-encoder into the communication system. For example, when the self-encoder contains only AI decoders, the AI decoders may be trained based on the training set. When the self-encoder includes an AI encoder, the AI encoder may be trained based on the training set. When the self-encoder includes an AI encoder and an AI decoder, the AI encoder and the AI decoder may be jointly trained.

In some implementations, since AE is a neural network model that takes the input signal as a training target, when the difference between the input and output of the self-encoder is represented by a loss function, the training target of the self-encoder can be understood as optimizing the weights of the AI encoder and AI decoder with the loss function minimized.

For example, one self-encoder that includes an AI encoder f (·) and an AI decoder g (·) is denoted as g (f (·)). After the original signal s is first encoded by the AI encoder f (·), the AI encoder f (·) outputs an encoded signal denoted q=f(s). When the encoded signal is input to the AI decoder g (·) and decoded, the decoded signal output from the AI decoder g (·) is expressed as s' =g (q) =g (f (s)). In the joint training stage, min/u can be calculated _{g,f} l (s, g (f (s))) as training targets, the AI encoder f (·) and the AI decoder g (·) are jointly trained, where l (·) represents the loss function.

5. Neural network

In recent years, artificial intelligence research represented by neural networks has achieved very great results in many fields, and will also play an important role in the production and life of people for a long time in the future. Common neural networks are convolutional neural networks (convolutional neural network, CNN), cyclic neural networks (recurrent neural network, RNN), deep neural networks (deep neural network, DNN), and the like.

A neural network to which embodiments of the present application are applicable is described below with reference to fig. 3. The neural networks shown in fig. 3 can be divided into three categories according to the location division of the different layers: an input layer 310, a hidden layer 320, and an output layer 330. In general, the first layer is the input layer 310, the last layer is the output layer 330, and the intermediate layers between the first layer and the last layer are both hidden layers 320.

The input layer 310 is used for inputting data, wherein the input data may be, for example, a received signal received by a receiver. The concealment layer 320 is configured to process input data, for example, decompress a received signal. The output layer 330 is configured to output the processed output data, for example, output the decompressed signal.

As shown in fig. 3, the neural network includes a plurality of layers, each layer includes a plurality of neurons, and the neurons between layers may be fully connected or partially connected. For connected neurons, the output of the upper layer of neurons may serve as the input to the lower layer of neurons.

With the continuous development of the neural network research, a neural network deep learning algorithm is proposed in recent years, more hidden layers are introduced into the neural network to form DNN, and the DNN can be more descriptive of complex situations in the real world by more hidden layers. Theoretically, the more parameters the higher the model complexity, the greater the "capacity", meaning that it can accomplish more complex learning tasks. The neural network model is widely applied to aspects such as pattern recognition, signal processing, optimization combination, anomaly detection and the like.

CNN is a deep neural network with a convolutional structure, which may include an input layer 410, a convolutional layer 420, a pooling layer 430, a full connection layer 440, and an output layer 450, as shown in fig. 4.

Each convolution layer 420 may comprise a number of convolution operators, also known as kernels, which act as a filter to extract specific information from the input signal, which may be essentially a weight matrix, which is typically predefined.

The weight values in the weight matrixes are required to be obtained through a large amount of training in practical application, and each weight matrix formed by the weight values obtained through training can extract information from the input signals, so that correct prediction of the CNN is facilitated.

When CNN has multiple convolutional layers, the initial convolutional layer tends to extract more general features, which may also be referred to as low-level features; as the depth of CNNs deepens, features extracted by the later convolutional layers become more and more complex.

The pooling layer 430, because it is often desirable to reduce the number of training parameters, often requires periodic introduction of the pooling layer after the convolutional layers, for example, one convolutional layer followed by one pooling layer as shown in fig. 4, or multiple convolutional layers followed by one or more pooling layers. The only purpose of the pooling layer during signal processing is to reduce the spatial size of the extracted information.

The full connection layer 440, after being processed by the convolution layer 420 and the pooling layer 430, is not enough for CNN to output the required output information. Because, as previously described, convolutional layer 420 and pooling layer 430 will only extract features and reduce the parameters imposed by the input data. However, in order to generate the final output information (e.g., a bit stream of the original information transmitted by the transmitting end), the CNN also needs to utilize the full connection layer 440. In general, the fully-connected layer 440 may include a plurality of hidden layers, where parameters included in the plurality of hidden layers may be pre-trained according to training data associated with a specific task type, e.g., the task type may include decoding a data signal received by the receiver, and e.g., the task type may further include channel estimation based on a pilot signal received by the receiver.

After the multi-hidden layer in the fully connected layer 440, i.e., the final layer of the entire CNN is the output layer 450 for outputting the result. Typically, the output layer 450 is provided with a loss function (e.g., a class-cross entropy like loss function) for calculating a prediction error, or for evaluating the degree of difference between the result (also called a predicted value) output by the CNN model and an ideal result (also called a true value).

In order to minimize the loss function, the CNN model needs to be trained. In some implementations, the CNN model may be trained using a back propagation algorithm (backpropagation algorithm, BP). The training process of BP consists of a forward propagation process and a backward propagation process. In the forward propagation (e.g., propagation from 410 to 450 in fig. 4) the input data is input to the layers of the CNN model, processed layer by layer, and passed to the output layer. If the difference between the output result at the output layer and the ideal result is large, the minimization of the loss function is used as an optimization target, the reverse propagation (such as the propagation from 450 to 410 in fig. 4 is the reverse propagation), the partial derivative of the optimization target to each neuron weight is obtained layer by layer, the gradient of the optimization target to the weight vector is formed, the basis of modifying the model weight is used, and the training process of the CNN is completed in the weight modification process. When the error reaches the expected value, the training process of the CNN is ended.

It should be noted that the CNN shown in fig. 4 is only an example of a convolutional neural network, and in a specific application, the convolutional neural network may also exist in the form of other network models, which is not limited in the embodiment of the present application.

The wireless communication system signal transmission process, channel estimation, CSI feedback, and self-encoder are described above in connection with fig. 1 to 4, and the self-encoder based signal transmission process, channel estimation, CSI feedback are described below in connection with fig. 5 to 7.

6. Channel estimation based on AI decoder

Channel estimation based on AI decoders is intended to be achieved by processing pilot signals received by the receiver with AI decoders. Fig. 5 shows a process of channel estimation based on an AI decoder. Referring to fig. 5, a pilot signal received by the receiver 500 is taken as an input to the AI decoder 510, and accordingly, the AI decoder 510 processes the input pilot signal to output channel information. In addition, in some implementations, other side information may be added in addition to the pilot signal to improve the accuracy of the AI decoder output channel information. For example, the original sequence of the pilot signal pre-stored in the receiver 500, the energy level of the received pilot signal by the receiver 500, the transmission delay when transmitting the pilot signal, or the noise when transmitting the pilot signal, etc. may also be input to the AI decoder 510.

7. CSI feedback based on self-encoder

The AI encoder in the self-encoder can perform feature extraction on the received signal carrying the CSI, and the AI decoder in the self-encoder can restore the CSI fed back by the transmitter compression as far as possible at the receiver, so that the communication overhead for feeding back the CSI can be saved while the accuracy of the CSI transmission is not affected.

Fig. 6 shows a self-encoder based CSI feedback system. As shown in fig. 6, the entire feedback system includes a self-encoder including an AI encoder 611 and an AI decoder 621 portion, wherein the AI encoder 611 is disposed at the transmitter 610 and the AI decoder 621 is disposed at the receiver 620. The transmitter 610 performs compression coding on the CSI to be transmitted through the AI encoder 611 to obtain compressed CSI. The compressed CSI is fed back to the receiver 620 through a feedback link, and the receiver 620 decodes the compressed CSI through the AI decoder 621 to obtain recovered CSI.

8. Receiver based on AI decoder

The performance of the receiver is improved by incorporating an AI decoder in the receiver design and utilizing the AI decoder to effect processing (e.g., demodulation, decompression, etc.) of the signal within the receiver. Fig. 7 is a schematic diagram of an AI decoder-based receiver. In the receiver 700 shown in fig. 7, the input of the AI decoder 710 is a received signal received by the receiver and output as a decoded signal.

From the above description, it can be seen that the self-encoder based modular communication system design is a trend of communication system development, which can make good use of the a priori structure of the conventional communication system model, while also being flexibly adapted and trained for AI encoders and/or AI decoders in the self-encoder.

Currently, before the AI decoder is on-line (i.e., deployed to a receiver), the AI decoder is trained by using an off-line training method based on a preset training set. However, the training effect is not ideal, and the actual communication system condition is complex, and the training data in the training set cannot cover all conditions, so that the trained AI decoder can output more accurate decoded signals only when processing signals with similar characteristics to the training data in the training set. When the difference between the received signal actually received by the receiver and the training data is large, the AI decoder cannot decode the received signal more accurately, that is, the generalization capability of the AI decoder after offline training is poor, where the generalization capability of the AI decoder is used to describe the accuracy of the AI decoder in decoding other signals with large differences from the training data.

The application provides a signal processing scheme to realize an on-line training process of an AI decoder. In order to facilitate understanding of the present application, a wireless communication system to which an embodiment of the present application is applied will be described with reference to fig. 8, and a method according to an embodiment of the present application will be described with reference to fig. 9.

Fig. 8 is a wireless communication system 800 to which embodiments of the present application are applicable. The wireless communication system 800 may include a network device 810. Network device 810 may be a device in communication with terminal device 820. Network device 810 may provide communication coverage for a particular geographic area and may communicate with terminal devices 820 located within that coverage area.

Fig. 8 illustrates one network device 810 and two terminal devices 820, alternatively, the wireless communication system 800 may include multiple network devices and each network device may include other numbers of terminal devices within its coverage area, which is not limited by the embodiments of the present application.

Optionally, the wireless communication system 800 may further include a network controller, a mobility management entity, and other network entities, which are not limited by the embodiments of the present application.

Alternatively, the terminal devices 820 may communicate directly with each other, for example, two terminal devices 820 may communicate with each other via a device-to-device (D2D) link.

It should be understood that the technical solution of the embodiment of the present application may be applied to various communication systems, for example: fifth generation (5th generation,5G) systems or New Radio (NR), long term evolution (long term evolution, LTE) systems, LTE frequency division duplex (frequency division duplex, FDD) systems, LTE time division duplex (time division duplex, TDD), and the like. The technical scheme provided by the application can also be applied to future communication systems, such as a sixth generation mobile communication system, a satellite communication system and the like.

The Terminal device in the embodiments of the present application may also be referred to as a User Equipment (UE), an access Terminal, a subscriber unit, a subscriber station, a Mobile Station (MS), a Mobile Terminal (MT), a remote station, a remote Terminal, a mobile device, a user Terminal, a wireless communication device, a user agent, or a user equipment. The terminal device in the embodiment of the application can be a device for providing voice and/or data connectivity for a user, and can be used for connecting people, things and machines, such as a handheld device with a wireless connection function, a vehicle-mounted device and the like. The terminal device in the embodiment of the present application may be a mobile phone (mobile phone), a tablet (Pad), a notebook, a palm, a mobile internet device (mobile internet device, MID), a wearable device, a Virtual Reality (VR) device, an augmented reality (augmented reality, AR) device, a wireless terminal in industrial control (industrial control), a wireless terminal in self driving (self driving), a wireless terminal in teleoperation (remote medical surgery), a wireless terminal in smart grid (smart grid), a wireless terminal in transportation security (transportation safety), a wireless terminal in smart city (smart city), a wireless terminal in smart home (smart home), or the like. Alternatively, the UE may be used to act as a base station. For example, the UEs may act as scheduling entities that provide side-uplink signals between UEs in V2X or D2D, etc. For example, a cellular telephone and a car communicate with each other using side-link signals. Communication between the cellular telephone and the smart home device is accomplished without relaying communication signals through the base station.

The network device in the embodiment of the present application may be a device for communicating with a terminal device, and the network device may also be referred to as an access network device or a radio access network device, for example, the network device may be a base station. The network device in the embodiments of the present application may refer to a radio access network (radio access network, RAN) node (or device) that accesses the terminal device to the wireless network. The base station may broadly cover or replace various names in the following, such as: a node B (NodeB), an evolved NodeB (eNB), a next generation NodeB (gNB), a relay station, an access point, a transmission point (transmitting and receiving point, TRP), a transmission point (transmitting point, TP), a master MeNB, a secondary SeNB, a multi-mode wireless (MSR) node, a home base station, a network controller, an access node, a wireless node, an Access Point (AP), a transmission node, a transceiving node, a baseband unit (BBU), a remote radio unit (Remote Radio Unit, RRU), an active antenna unit (active antenna unit, AAU), a radio head (remote radio head, RRH), a Central Unit (CU), a Distributed Unit (DU), a positioning node, and the like. The base station may be a macro base station, a micro base station, a relay node, a donor node, or the like, or a combination thereof. A base station may also refer to a communication module, modem, or chip for placement within the aforementioned device or apparatus. The base station may also be a mobile switching center, a device-to-device (D2D), a vehicle-to-device (V2X), a device that assumes a base station function in machine-to-machine (M2M) communication, a network-side device in a 6G network, a device that assumes a base station function in a future communication system, or the like. The base stations may support networks of the same or different access technologies. The embodiment of the application does not limit the specific technology and the specific equipment form adopted by the network equipment.

The base station may be fixed or mobile. For example, a helicopter or drone may be configured to act as a mobile base station, and one or more cells may move according to the location of the mobile base station. In other examples, a helicopter or drone may be configured to function as a device to communicate with another base station.

In some deployments, the network device in embodiments of the application may refer to a CU or a DU, or the network device may include a CU and a DU. The gNB may also include an AAU.

Network devices and terminal devices may be deployed on land, including indoors or outdoors, hand-held or vehicle-mounted; the device can be deployed on the water surface; but also on aerial planes, balloons and satellites. In the embodiment of the application, the scene where the network equipment and the terminal equipment are located is not limited.

It should be understood that the communication device referred to in the present application may be a network device or may also be a terminal device. For example, the first communication device is a network device, and the second communication device is a terminal device. As another example, the first communication device is a terminal device and the second communication device is a network device. As another example, the first communication device and the second communication device are both network devices, or are both terminal devices.

It should also be understood that all or part of the functionality of the communication device in the present application may also be implemented by software functions running on hardware or by virtualized functions instantiated on a platform, such as a cloud platform.

Fig. 9 is a flowchart of a method of signal processing according to an embodiment of the present application. The method shown in fig. 9 includes steps S910 to S940. It will be appreciated that the receiver shown in fig. 9 may be the terminal device described above and that the transmitter may be the terminal device described above, accordingly, in which case communication between the receiver and the transmitter may be via a D2D link. Of course, when the receiver is a terminal device, the transmitter may be a network device.

In step S910, the receiver receives the wireless signal transmitted by the transmitter, and obtains a received signal.

In step S920, the receiver inputs the received signal to the AI decoder to decode, thereby obtaining a decoded signal.

In step S930, the receiver generates a recovery signal of the received signal from the decoded signal.

In the embodiment of the present application, the manner of generating the recovery signal is not limited. In some embodiments, the recovery signal may be obtained by simple signal processing of the decoded signal. In other embodiments, the decoded signal may be input to an AI encoder to generate a recovered signal of the received signal. The following will be described in detail with reference to fig. 14 to 16, and will not be described here again for brevity.

In step S940, the receiver trains the AI decoder online according to the difference between the recovered signal and the received signal. Wherein the difference between the recovered signal and the received signal may be represented by an error between the recovered signal and the received signal.

In some implementations, the step S940 may include the receiver updating the weight of the AI decoder or optimizing the weight of the AI decoder to train the AI decoder online with the minimized difference between the recovered signal and the received signal as an optimization target.

When the AI decoder is the neural network model described above, the weight of the AI decoder can be understood as a weight used when each layer model in the neural network model processes the input data. For example, referring to fig. 4, the weights of the ai decoder may include weights used by the input layer 410, the convolution layer 420, the pooling layer 430, the full connection layer 440, and the output layer 450 to process inputs of the respective layers.

The above updating of the weights in the AI decoder can be understood as updating the weights of all of the AI decoder or updating the weights of a part of the AI decoder.

The difference between the recovered signal and the received signal may be represented by an error between the recovered signal and the received signal. In some implementations, the error between the recovered signal and the received signal may be calculated by a loss function, or the loss function may represent the error between the recovered signal and the received signal.

In general, in order to minimize the loss function, an on-line training of the AI decoder is required. For example, the BP may be used to train the AI decoder. The training process of BP may consist of a forward propagation process and a backward propagation process. Taking the AI decoder as an example of the model shown in fig. 4, in the forward propagation (such as propagation from 410 to 450 in fig. 4 is forward propagation), the received signal is input into each layer of the AI decoder model, is processed layer by layer and is transmitted to the output layer, and the decoded signal is output by the output layer accordingly, and the receiver obtains a recovery signal of the received signal based on the decoded signal. If the difference between the received signal and the recovered signal is large, the minimization of the loss function is used as an optimization target, the reverse propagation (such as the propagation from 450 to 410 in fig. 4 is the reverse propagation), the partial derivative of the optimization target to the weight of each neuron is obtained layer by layer, the gradient of the optimization target to the weight vector is formed, the gradient of the optimization target to the weight vector is used as the basis for modifying the weight of the AI decoder, and the AI decoder online training process is completed in the weight modification process. When the error reaches a desired value, the online training process of the AI decoder ends. Of course, other known training principles may be used in the online training process of the AI decoder according to the present application, which is not limited in this embodiment of the present application.

In the embodiment of the application, the receiver decodes the received signal by using the AI decoder to obtain the decoded signal, then generates the recovery signal of the received signal according to the decoded signal, and carries out on-line training on the AI decoder based on the difference between the received signal and the recovery signal. Or after the AI decoder is on line, the AI decoder is trained on line by utilizing the received signal actually received by the receiver, so that even if the received signal has a large difference from training data, the AI decoder after on-line training has higher accuracy when decoding the received signal, thereby being beneficial to improving the generalization capability of the AI decoder.

Compared with the scheme for training the AI decoder based on the difference between the original signal to be transmitted by the transmitter and the received signal received by the receiver, the scheme of the embodiment of the application avoids the problem that the AI decoder cannot train on line in the receiver because the original signal cannot be accurately known by the receiver due to the distributed arrangement of the receiver and the transmitter.

The time of the online training may be configured for the receiver by the network device. In some implementations, the network device may send online training indication information to the receiver, where the online training indication information is used to indicate when the AI decoder is online training (for ease of understanding, the time when the AI decoder is online training is hereinafter referred to as an "online training time"). Of course, the receiver may also perform online training according to the online training time specified by the communication protocol. The embodiment of the present application is not particularly limited thereto.

The indication information of the online training may indicate the time of the online training in various ways. For example, the indication of online training may indicate that the receiver is periodically online training the AI decoder by indicating an online training period. For another example, the indication information of the online training may also directly indicate the start time and/or the end time of the online training.

In some implementations, the indication information of online training may be transmitted through radio resource control (radio resource control, RRC) signaling, and in other implementations, the indication information of online training may also be transmitted through a control channel, which is not limited in the embodiments of the present application.

Based on the difference of the indication modes of the online training, the online training configuration modes of the AI decoder can be divided into periodic online training, non-continuous online training and non-periodic online training. The above three online training configurations are described below in conjunction with fig. 10-12.

The online training configuration mode is first and periodic online training is carried out. I.e., the receiver periodically trains the AI decoder online based on the configured online training period.

Since the on-line training needs to be performed based on the signal actually transmitted between the transmitter and the receiver (i.e., the received signal above), in order to simplify the configuration of the on-line training period, in some implementations, the on-line training period may include a plurality of transmission periods of the actually transmitted signal. Accordingly, the receiver may train online during any of the plurality of transmission periods. In general, in order to ensure the accuracy of the AI decoder decoding, the on-line training may be performed in a first transmission period of a plurality of transmission periods, and thus the trained AI decoder may be used for decoding in several transmission periods after the first transmission period.

As described above, the receiver may perform online training in any of a plurality of transmission periods, that is, the online training period includes a first type of transmission period and a second type of transmission period, wherein the receiver may perform online training on the AI decoder in the first type of transmission period and decode a received signal received in the current transmission period using the trained AI decoder. In the second type of transmission period, the receiver may directly employ the AI decoder to decode the received signal without on-line training of the AI decoder. In general, to improve the accuracy of the on-line training, the second type of transmission period may also be used to collect the received signal as training data for the on-line training.

It should be noted that, the arrangement manner of the first type of transmission period and the second type of transmission period in the online training period may be preconfigured based on the communication protocol. Of course, the network device may also be indicated by signaling, which is not limited by the embodiment of the present application.

In addition, the number of transmission periods included in the online training period is not particularly limited in the embodiment of the present application. In some implementations, the online training period may include fewer transmission periods, and accordingly, the frequency of online training may be higher, and the configuration of such online training period may be suitable for more frequent communications scenarios, such as high-speed rail, etc. In other embodiments, the online training period may comprise more transmission periods, and accordingly, the frequency of online training may be lower, which may be advantageous in reducing the computational resources required for online training. This configuration of online training periods is applicable to less frequent changing communication scenarios, such as office buildings and the like.

The method of signal processing according to the embodiment of the present application will be described below by taking the periodic on-line training shown in fig. 10 as an example. As shown in fig. 10, the online training period r ₁ Comprising 4 transmission periods t ₁ And 4 transmission periods t ₁ The first transmission period of the (c) is a first type of transmission period and the other 3 transmission periods are a second type of transmission period.

In the first type of transmission period, the receiver needs to train the AI decoder online based on the training data, and then decode the received signal received in the current transmission period by using the AI decoder after the online training.

In the second type of transmission period, the receiver may directly use the above-described AI decoder after on-line training to decode the received signal received in the current transmission period. Meanwhile, the receiver can collect the received signal received in the current transmission period as training data of subsequent online training.

And the online training configuration mode is two, and online training is non-continuous. That is, on the basis of the periodic online training, the start time and the end time of the online training can be indicated by the indication information of the online training, so that the flexibility of configuring the online training is improved.

In this training mode, the periodic online training may be referred to above in the description of the first online training configuration mode, and in the interest of brevity, the mode of indicating the start time and the end time of the online training will be described with emphasis.

The indication information of the online training can indicate the starting time of the online training in a display or hidden mode. For example, the indication of online training may directly carry the start time. For another example, the receiver may take a time domain unit in which the indication information of the online training is transmitted as a starting point, and offset the preset time domain unit to obtain an offset time domain unit, where the offset time domain unit is the starting time, and the time domain unit may be a time slot, a subframe, or the like, for example.

Accordingly, the indication information of the online training can also indicate the ending time of the online training in a display or hidden mode. For example, the indication of online training may directly carry the end time. For another example, the indication of online training may indicate the end time by indicating a total number of cycles of the online training period.

Hereinafter, as shown in FIG. 11For example, non-continuous on-line training, a method of signal processing according to an embodiment of the present application is described. As shown in fig. 11, it is assumed that the online training instruction information indicates that the start time of online training is the ith transmission period t _i The total cycle number of the online training is 3, and the online training period r ₂ Including 2 transmission periods, then the on-line training period r ₂ The start time of (a) is the ith transmission period t _i Start time of on-line training period r ₂ Ending time of (i+6) th transmission period t _i+6 End time of (2).

Every online training period r ₂ The first transmission period in (a) is a first type of transmission period in which the receiver needs to train the AI decoder online based on the training data, and decodes the received signal received in the current transmission period using the AI decoder after the online training.

Every online training period r ₂ The second transmission period in (a) is a second type of transmission period, and in the second type of transmission period, the receiver can directly use the AI decoder after online training to decode the received signal received in the current transmission period. Meanwhile, the receiver can collect the received signal received in the current transmission period as training data of subsequent online training.

And the online training configuration mode III is aperiodic online training. I.e. each online training needs to be performed by means of the indication information of the online training.

In some implementations, the start time of the online training and the end time of the online training may be carried in the indication information of the online training, and accordingly, the receiver performs the online training in a time period indicated by the start time and the end time. In other implementations, the start time of the online training may be carried in the indication information of the online training, and accordingly, the receiver performs the online training at the start time after receiving the indication information of the online training.

The indication information of the online training can indicate the starting time of the online training in a display or hidden mode. For example, the indication of online training may directly carry the start time. For another example, the receiver may obtain an offset time slot after offsetting a preset time slot from a time slot where the indication information for transmitting the online training is located, where the offset time slot is the start time.

The indication information of the online training can indicate the ending time of the online training in a display or hidden mode. For example, the indication of online training may directly carry the end time. For another example, the receiver may obtain an offset time slot after offsetting a preset time slot with a time slot in which the indication information for online training is transmitted as a start point, where the offset time slot is an end time.

The method of signal processing according to the embodiment of the present application will be described below by taking aperiodic on-line training as shown in fig. 12. As shown in FIG. 12, assume that the indication information of the online training ₁ Indicating that in transmission period t _i Performing online training, and indicating information of online training ₂ Indicating that in transmission period t _i+4 And performing online training.

When the receiver receives the indication information of the online training ₁ After that, in the transmission period t _i The receiver needs to train the AI decoder online based on the training data, and decodes the received signal received in the current transmission period by using the AI decoder after the online training.

In transmission period t _i+1 By transmission period t _i+3 The receiver directly uses the AI decoder after the online training to decode the received signal received in the current transmission period.

When the receiver receives the indication information of the online training ₂ After that, in the transmission period t _i+4 The receiver continues to train the AI decoder online based on the training data, and decodes the received signal received in the current transmission period using the online trained AI decoder.

In the above-described online training process, the receiver may collect the received signals as training data for online training, but in some cases, the number of received signals collected by the receiver is insufficient to perform online training on the AI decoder, and at this time, part or all of the training data in the training set pre-stored in the receiver may be additionally selected to perform online training. In some implementations, the pre-stored training set may be a data set used by an offline training AI decoder.

The fewer training data in the training set are selected, the greater the number of received signals collected by the receiver, the higher the generalization performance of the AI decoder after online training.

In some implementations, the size of the training data used for online training may be predefined. In other embodiments, the size of the training data used for online training may also be indicated by the network device. For example, the network device may indicate the size of training data used for the first type of transmission period through the online training indication information. For another example, the network device may indicate the size of the training data used by the first type of transmission period through other information.

The size of the training data used in the above-mentioned online training can be understood as the total amount of all the training data used in each online training. In some implementations, if the number of received signals collected by the receiver is insufficient to train the AI decoder online, the size of the training data used for the online training includes a sum of the number of received signals collected by the receiver and the number of training data selected in the training set. In other implementations, the size of the training data used for the online training may include only the number of received signals as training data if the number of received signals collected by the receiver is sufficient.

It should be noted that the above-described scheme in which the network device indicates the size of the training data may be used in combination with any of the three online configurations described above. For example, when used in conjunction with the online training configuration, the online training indication information may include a first indication information indicating an online training period for the AI decoder to perform online training, and the online training indication information may further include a second indication information indicating a size of training data used for the first type of transmission period.

The method of signal processing according to the embodiment of the present application is described above with reference to fig. 10 to 12, and the method of signal processing is described below with reference to fig. 13 by taking the application of the AI decoder in the channel estimation process as an example. It should be noted that the online training process described below may be combined with any of the online training configurations described above. For brevity, the following description is omitted.

In the channel estimation process, the received signal includes a pilot signal transmitted by the transmitter, the step S920 includes the receiver inputting the received signal into the AI decoder for decoding to perform channel estimation, so as to obtain a decoded signal, the decoded signal includes estimated first channel information, and the step S930 includes the receiver processing the pilot signal stored in the receiver according to the first channel information and the first random noise, so as to obtain a recovered signal.

In general, the pilot signal transmitted by the transmitter may be configured to be identical to the pilot signal stored by the receiver so that the receiver may perform channel estimation based on the received signal and the stored pilot signal to obtain estimated channel information.

The first random noise may be pre-stored by the receiver or may be pre-generated by the receiver, which is not limited in the embodiment of the present application.

In order to improve the accuracy of the channel information estimated by the AI decoder, in some implementations, the pilot signal stored by the receiver may also be input to the AI decoder along with the received signal. Of course, the received signal may be input only to the AI decoder, which is not limited by the embodiment of the present application.

In addition, because of the high difficulty in acquiring the noise in the actual communication environment, in the process of acquiring the recovery signals, the receiver can process the pilot signals based on a plurality of pre-stored different random noises to acquire a plurality of recovery signals. And selecting a recovery signal most similar to the received signal from the plurality of recovery signals, and training the AI decoder online according to the difference between the similar recovery signal and the received signal.

A method for signal processing in a channel estimation process according to an embodiment of the present application is described below with reference to fig. 13.

As shown in fig. 13, the transmitter 1310 transmits the pilot signal p through the channel H, and during the transmission, the receiver 1320 collects the received signal Y due to the presence of the noise N in the actual channel _p Can be expressed as Y _p ＝H*p+N。

Receiver 1320 will collect received signal Y _p The AI decoder 1321 is input to obtain the channel information H'. Then, the receiver 1320 processes the pilot signal p stored in the receiver 1320 based on the channel information H 'and the random noise N' to obtain the recovery signal Y _p ', i.e. Y _p ' H ' ×p+n '. Finally, receiver 1320 is based on recovery signal Y _p ' and received signal Y _p The difference between them, the AI decoder 1321 is trained online.

As introduced previously, the self-encoder may also include an AI encoder. For self-encoders that include an AI encoder and an AI decoder, the AI encoder and AI decoder are typically trained in a joint training manner. That is, the self-encoder formed by the AI encoder and the AI decoder is jointly trained based on the difference between the input signal of the AI encoder and the output signal of the AI decoder.

Also, the AI encoder and AI decoder after the online are deployed in the receiver and transmitter, respectively, and the inputs of the AI encoder are not known at the receiver due to the distributed deployment of the receiver and transmitter, resulting in the above joint training being only performed on-line based on the training set. As shown in fig. 14, an AI encoder 1411 is disposed at the transmitter 1410 for encoding the original signal s to obtain an encoded signal q. The transmitter 1410 transmits the encoded signal q to the receiver 1420 over a wireless link. Accordingly, the receiver 1420 inputs the received encoded signal q to the AI decoder 1421 to be decoded, resulting in a decoded signal s'. In the on-line training process, the difference between the original signal s and the decoded signal s' needs to be performed, and the original signal s cannot be accurately transmitted through the communication link between the transmitter 1410 and the receiver 1420, so the AI encoder 1411 and the AI decoder 1421 cannot be jointly trained in an on-line training manner, where the joint training may be understood as updating weights in the AI encoder and the AI decoder as a whole.

In order to improve the generalization capability of the self-encoder. The embodiment of the application also provides a signal processing method which is used for carrying out joint training on the AI encoder and the AI decoder in an on-line training mode. By configuring an AI encoder in the receiver, the AI encoder may be configured to be jointly trained with the receiver's original AI decoder, i.e., based on the difference between the AI decoder's input and the AI encoder's output. That is, the step S930 includes: the receiver inputs the decoded signal to an AI encoder to generate a recovery signal.

In some implementations, to improve accuracy of joint training, the AI encoder configured at the receiver described above may have the same model structure and/or the same model parameters, e.g., the same model weights, etc., as the AI encoder in the transmitter. Of course, if the goal of the joint training is to focus on improving the generalization performance of the AI decoder, a simpler AI encoder may be deployed in the receiver. Alternatively, the parameters of the AI encoder may be frozen during the co-training process, with only the AI decoder parameters being trained.

If the joint training mode is hoped to be adopted, the generalization performance of the AI encoder in the transmitter and the AI decoder in the receiver is integrally improved, the joint training of the AI encoder and the AI decoder in the receiver can be carried out in an on-line training mode, and after the joint training is finished, model parameters of the AI encoder trained in the receiver are fed back to the transmitter.

In some implementations, the training objectives (or optimization objectives) of the receiver's self-encoder are represented by a loss function as: min/u _{g’} l '(q, f' (g '(q))), wherein l' (-) represents a loss function; q represents the received reception by the receiverReceiving a signal; g' (·) denotes an AI decoder in the receiver; f' (·) denotes an AI encoder in the receiver.

The on-line training process of the self-encoder by the joint training will be described below with reference to fig. 15 and 16 by taking the application of the self-encoder formed by the AI encoder and the AI decoder in the CSI feedback process and the data transmission process as an example. It should be noted that the online training process described below may be combined with any of the online training configurations described above. For brevity, the following description is omitted.

In the CSI feedback process, the received signal includes CSI transmitted by the transmitter, where the CSI is compressed CSI, and step S920 includes the receiver inputting the CSI into the AI decoder to decode, to obtain a decoded signal, where the decoded signal includes recovered channel information; the step S930 includes the receiver inputting the decoded CSI into the AI encoder for encoding to recover the CSI, and generating a recovery signal, where the recovery signal includes the recovered CSI.

Fig. 15 is a schematic diagram of a signal processing method in CSI feedback process according to an embodiment of the present application. As shown in fig. 15, a transmitter 1510 encodes CSI to be transmitted into an AI encoder 1511 to obtain encoded CSI (also referred to as "compressed CSI"), and the transmitter 1510 transmits the compressed CSI to a receiver 1520. Accordingly, the receiver 1520 decodes the received signal containing the compressed CSI in the AI decoder 1521 to obtain decoded CSI. The receiver 1520 then encodes the decoded CSI into an AI encoder 1522 to obtain recovered CSI. The receiver 1520 jointly trains the AI encoder 1522 and the AI decoder 1521 in the receiver 1520 based on the difference between the recovered CSI and the compressed CSI.

In the data transmission process, the received signal includes a first data signal transmitted by the transmitter, and the step S920 includes the receiver inputting the first data signal into the AI decoder for decoding, so as to recover the data transmitted by the transmitter, and obtain a decoded signal, where the decoded signal includes the recovered data; the receiver inputs the decoded signal to an AI encoder to generate a recovery signal, comprising: the receiver inputs the recovered data into an AI encoder to obtain a second data signal; the receiver processes the second data signal according to the estimated second channel information and the second random noise to obtain a recovered signal.

The second random noise may be pre-stored by the receiver or may be pre-generated by the receiver, which is not limited in the embodiment of the present application.

In some implementations, a pilot signal may be inserted into the first data signal, such that the receiver may perform channel estimation based on the pilot signal to increase the similarity of the recovered signal to the received signal. Of course, the second data signal may be directly input into the AI model to obtain the recovery signal without performing channel estimation, which is not limited in the embodiment of the present application.

Fig. 16 is a schematic diagram of a signal processing method in a data transmission process according to an embodiment of the present application. As shown in fig. 16, the transmitter 1610 encodes data to be transmitted into an AI encoder 1611 to obtain a first data signal. The transmitter 1620 transmits the first data signal to the receiver 1620 through the wireless channel H, and during the transmission, noise N is superimposed on the first data signal to form the reception signal Y. Accordingly, the receiver 1620 receives the received signal Y, and then inputs the received signal Y to the AI decoder 1621 to decode, thereby obtaining recovered data. The receiver 1620 then encodes the recovered data in an AI encoder 1622 to obtain a second data signal. The receiver 1620 processes the second data signal according to the random noise N ' and the estimated channel information H ' to obtain a recovery signal Y '. The receiver 1620 jointly trains the AI encoder 1622 and the AI decoder 1621 in the receiver 1620 based on the difference between the recovered signal Y' and the received signal Y.

Of course, only the AI decoder may be used instead of the AI encoder during data transmission, and the signal processing performed by the transmitter may be similar to a conventional signal processing, for example, encoding, modulating, etc. a signal to be transmitted. Accordingly, the receiver may also perform signal processing on the decoded signal output by the AI decoder in a manner similar to that of the conventional signal processing flow, for example, encoding, modulating, etc. the signal to be transmitted. In this case, it can be understood that only the AI decoder is trained online, and in particular, reference may be made to the online training process of the AI decoder, which is not described herein for brevity.

The method embodiments of the present application are described above in detail with reference to fig. 1 to 16, and the apparatus embodiments of the present application are described below in detail with reference to fig. 17 to 18. It is to be understood that the description of the method embodiments corresponds to the description of the device embodiments, and that parts not described in detail can therefore be seen in the preceding method embodiments.

Fig. 17 is a schematic structural diagram of a receiver of an embodiment of the present application. The receiver 1700 shown in fig. 17 includes: a receiving unit 1710 and a processing unit 1720.

The receiving unit 1710 may be configured to receive a wireless signal transmitted by a transmitter, to obtain a received signal.

The processing unit 1720 may be configured to input the received signal to an AI decoder to decode the received signal to obtain a decoded signal; generating a recovery signal of the received signal from the decoded signal; and performing on-line training on the AI decoder according to the difference between the recovery signal and the received signal.

Optionally, the receiving unit 1710 may be further configured to receive indication information of online training sent by a network device, where the indication information of online training is used to indicate a time when the AI decoder performs the online training.

Optionally, the indication information of the online training includes first indication information, where the first indication information is used to indicate an online training period of the online training performed by the AI decoder, the online training period includes a plurality of transmission periods, the online training period includes a first type of transmission period and a second type of transmission period, the first type of transmission period is used to perform online training on the AI decoder, and the second type of transmission period is used to collect training data of the online training.

Optionally, the indication information of the online training further includes second indication information, where the second indication information is used to indicate a size of training data used in the transmission period of the first type.

Optionally, the time of the online training includes a start time and/or an end time of the online training.

Optionally, the received signal is a received signal of a pilot signal transmitted by the transmitter, and the processing unit 1720 is further configured to input the received signal to the AI decoder for decoding, perform channel estimation, and obtain the decoded signal, where the decoded signal includes estimated first channel information; and processing the pilot signal stored by the receiver according to the first channel information and the first random noise to obtain the recovery signal, wherein the pilot signal transmitted by the transmitter is identical to the pilot signal stored by the receiver.

Optionally, the processing unit may be further configured to input the pilot signal stored by the receiver and the received signal into the AI decoder, and perform channel estimation to obtain the decoded signal.

Alternatively, the processing unit 1720 may be specifically configured to input the decoded signal to an AI encoder to generate the recovery signal.

Optionally, the received signal includes compressed CSI transmitted by the transmitter, where the CSI is compressed CSI, and the processing unit 1720 is further configured to input the CSI to the AI decoder for decoding, to obtain the decoded signal, where the decoded signal includes recovered channel information; and inputting the recovered channel information into the AI encoder for encoding, recovering the CSI, and generating the recovery signal, wherein the recovery signal contains the recovered CSI.

Optionally, the received signal includes a first data signal transmitted by the transmitter, and the processing unit 1720 is further configured to input the first data signal to the AI decoder for decoding, recover data transmitted by the transmitter, and obtain the decoded signal, where the decoded signal includes the recovered data; inputting the recovered data into the AI encoder to obtain a second data signal; and processing the second data signal according to the estimated second channel information and the second random noise to obtain the recovery signal.

Optionally, the processing unit 1720 may be further configured to update the weights of the AI decoders to perform the online training on the AI decoders by minimizing the difference as an optimization objective.

Optionally, the difference between the recovered signal and the received signal is represented by an error between the recovered signal and the received signal.

Fig. 18 is a schematic structural diagram of an apparatus for signal processing according to an embodiment of the present application. The dashed lines in fig. 18 indicate that the unit or module is optional. The apparatus 1800 may be used to implement the methods described in the method embodiments above. The apparatus 1800 may be a chip, a terminal device, or a network device.

The apparatus 1800 may include one or more processors 1810. The processor 1810 may support the apparatus 1800 to implement the methods described in the method embodiments above. The processor 1810 may be a general purpose processor or a special purpose processor. For example, the processor may be a central processing unit (central processing unit, CPU). Alternatively, the processor may be another general purpose processor, a digital signal processor (digital signal processor, DSP), an application specific integrated circuit (application specific integrated circuit, ASIC), an off-the-shelf programmable gate array (field programmable gate array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The apparatus 1800 may also include one or more memories 1820. The memory 1820 has stored thereon a program that can be executed by the processor 1810 to cause the processor 1810 to perform the methods described in the method embodiments above. The memory 1820 may be separate from the processor 1810 or may be integrated within the processor 1810.

The apparatus 1800 may also include a transceiver 1830. The processor 1810 may communicate with other devices or chips through a transceiver 1830. For example, the processor 1810 may transmit and receive data to and from other devices or chips through the transceiver 1830.

The embodiment of the application also provides a computer readable storage medium for storing a program. The computer-readable storage medium may be applied to a terminal or a network device provided in an embodiment of the present application, and the program causes a computer to execute the method performed by the terminal or the network device in the respective embodiments of the present application.

The embodiment of the application also provides a computer program product. The computer program product includes a program. The computer program product may be applied to a terminal or a network device provided in an embodiment of the present application, and the program causes a computer to execute a method executed by the terminal or the network device in each embodiment of the present application.

The embodiment of the application also provides a computer program. The computer program can be applied to a terminal or a network device provided in an embodiment of the present application, and cause a computer to perform a method performed by the terminal or the network device in each embodiment of the present application.

It should be understood that the term "and/or" is merely an association relationship describing the associated object, and means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.

It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.

In the several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present application, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by a wired (e.g., coaxial cable, fiber optic, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be read by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a digital versatile disk (digital video disc, DVD)), or a semiconductor medium (e.g., a Solid State Disk (SSD)), or the like.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

A method of signal processing, comprising:

the receiver receives the wireless signal transmitted by the transmitter to obtain a received signal;

the receiver inputs the received signal into an artificial intelligence AI decoder for decoding to obtain a decoded signal;

the receiver generates a recovery signal of the received signal according to the decoded signal;

the receiver trains the AI decoder online according to the difference between the recovered signal and the received signal.
The method of claim 1, wherein the method further comprises:

the receiver receives indication information of online training sent by network equipment, wherein the indication information of online training is used for indicating the time of the AI decoder for online training.
The method of claim 2, wherein the indication information of the online training comprises first indication information for indicating an online training period for the AI decoder to conduct the online training, the online training period comprising a plurality of transmission periods, the online training period comprising a first type of transmission period for online training the AI decoder and a second type of transmission period for collecting training data of the online training.
The method of claim 3, wherein the indication information of the online training further comprises second indication information indicating a size of training data used for the first type of transmission period.
The method according to any of claims 2-4, wherein the time of online training comprises a start time and/or an end time of the online training.
The method according to any one of claims 1 to 5, wherein the received signal is a received signal of a pilot signal transmitted by the transmitter,

The receiver inputs the received signal to an AI decoder for decoding to obtain a decoded signal, including:

the receiver inputs the received signal into the AI decoder for decoding, and performs channel estimation to obtain the decoded signal, wherein the decoded signal contains estimated first channel information;

the receiver generates a recovery signal of the received signal from the decoded signal, comprising:

and the receiver processes the pilot signal stored by the receiver according to the first channel information and the first random noise to obtain the recovery signal, wherein the pilot signal transmitted by the transmitter is identical to the pilot signal stored by the receiver.
The method of claim 6, wherein the receiver inputting the received signal to the AI decoder for decoding for channel estimation to obtain the decoded signal comprises:

the receiver inputs the pilot signal stored by the receiver and the received signal into the AI decoder to perform channel estimation, so as to obtain the decoded signal.
The method of any of claims 1-5, wherein the receiver generating a recovery signal for the received signal from the decoded signal comprises:

The receiver inputs the decoded signal to an AI encoder to generate the recovery signal.
The method of claim 8, wherein the received signal comprises channel state information, CSI, transmitted by the transmitter, the CSI being compressed CSI,

the receiver inputs the received signal to an AI decoder for decoding to obtain a decoded signal, including:

the receiver inputs the CSI into the AI decoder for decoding to obtain the decoding signal, wherein the decoding signal contains the restored channel information;

the receiver inputs the decoded signal to an AI encoder, generating the recovery signal, comprising:

the receiver inputs the decoded CSI into the AI encoder for encoding, recovers the CSI, and generates the recovery signal, wherein the recovery signal contains the recovered CSI.
The method of claim 8, wherein the received signal comprises a first data signal transmitted by the transmitter,

the receiver inputs the received signal to an AI decoder for decoding to obtain a decoded signal, including:

the receiver inputs the first data signal into the AI decoder for decoding, and recovers the data transmitted by the transmitter to obtain the decoded signal, wherein the decoded signal contains the recovered data;

The receiver inputs the decoded signal to an AI encoder, generating the recovery signal, comprising:

the receiver inputs the recovered data into the AI encoder to obtain a second data signal;

and the receiver processes the second data signal according to the estimated second channel information and the second random noise to obtain the recovery signal.
The method of any of claims 1-10, wherein the receiver training the AI decoder online based on a difference between the recovered signal and the received signal, comprising:

the receiver takes the minimization of the difference as an optimization target, updates the weight of the AI decoder, and carries out the online training on the AI decoder.
The method of any of claims 1-11, wherein the difference between the recovered signal and the received signal is represented by an error between the recovered signal and the received signal.
A receiver, comprising:

the receiving unit is used for receiving the wireless signals transmitted by the transmitter to obtain received signals;

the processing unit is used for inputting the received signal into an artificial intelligence AI decoder for decoding to obtain a decoded signal;

The processing unit is used for generating a recovery signal of the received signal according to the decoded signal;

the processing unit is used for carrying out online training on the AI decoder according to the difference between the recovery signal and the receiving signal.
The receiver of claim 13, wherein the receiving unit is further configured to:

and receiving indication information of online training sent by network equipment, wherein the indication information of online training is used for indicating the time of the AI decoder for carrying out the online training.
The receiver of claim 14, wherein the indication information of the on-line training comprises first indication information for indicating an on-line training period for the AI decoder to perform the on-line training, the on-line training period comprising a plurality of transmission periods, the on-line training period comprising a first type of transmission period for on-line training the AI decoder and a second type of transmission period for collecting training data for the on-line training.
The receiver of claim 15, wherein the indication information of the online training further comprises second indication information indicating a size of training data used for the first type of transmission period.
The receiver according to any of claims 14-16, wherein the time of online training comprises a start time and/or an end time of the online training.
The receiver according to any of claims 13-17, wherein the received signal is a received signal of a pilot signal transmitted by the transmitter, the processing unit being further configured to:

inputting the received signal into the AI decoder for decoding, and performing channel estimation to obtain the decoded signal, wherein the decoded signal contains estimated channel information; and

and processing the pilot signal stored by the receiver according to the first channel information and the first random noise to obtain the recovery signal, wherein the pilot signal transmitted by the transmitter is identical to the pilot signal stored by the receiver.
The receiver of claim 18, wherein the processing unit is further configured to:

and inputting the pilot signal stored by the receiver and the received signal into the AI decoder for channel estimation to obtain the decoded signal.
The receiver according to any one of claims 13-17, wherein,

The processing unit is further configured to input the decoded signal to an AI encoder, and generate the recovery signal.
The receiver of claim 20, wherein the received signal comprises channel state information, CSI, transmitted by the transmitter, the CSI being compressed CSI, the processing unit further configured to:

inputting the CSI into the AI decoder for decoding to obtain the decoding signal, wherein the decoding signal contains the recovered channel information; and

inputting the recovered channel information into the AI encoder for encoding, recovering the CSI, and generating the recovery signal, wherein the recovery signal contains the recovered CSI.
The receiver of claim 20, wherein the received signal comprises a first data signal transmitted by the transmitter, the processing unit further to:

inputting the first data signal into the AI decoder for decoding, and recovering the data transmitted by the transmitter to obtain the decoded signal, wherein the decoded signal contains the recovered data;

inputting the recovered data into the AI encoder to obtain a second data signal; and

and processing the second data signal according to the estimated second channel information and the second random noise to obtain the recovery signal.
The receiver according to any of claims 13-22, wherein the processing unit is further configured to:

and taking the minimization of the difference as an optimization target, updating the weight of the AI decoder, and carrying out the online training on the AI decoder.
The receiver of any of claims 13-23, wherein the difference between the recovered signal and the received signal is represented by an error between the recovered signal and the received signal.
A receiver comprising a memory for storing a program and a processor for invoking the program in the memory to perform the method of any of claims 1-12.
An apparatus comprising a processor configured to invoke a program from memory to perform the method of any of claims 1-12.
A chip comprising a processor for calling a program from a memory, causing a device on which the chip is mounted to perform the method of any one of claims 1-12.
A computer-readable storage medium, characterized in that a program is stored thereon, which program causes a computer to perform the method according to any of claims 1-12.
A computer program product comprising a program for causing a computer to perform the method of any one of claims 1-12.
A computer program, characterized in that the computer program causes a computer to perform the method according to any one of claims 1-12.