Communication method, model training method and corresponding device

Abstract

Disclosed is a communication method, comprising: a first device determining a first adjustment model according to first information from a second device, the first adjustment model being associated with information of a first transmitter and information of a second receiver (602), or the first adjustment model being associated with information of a first receiver and information of a second transmitter (612), the first transmitter and the first receiver correspond to the first device, and the second receiver and the second transmitter correspond to the second device; the first device communicating with a second receiver of the second device based on the first transmitter and the first adjustment model (603); alternatively, the first device communicates with a second transmitter of the second device based on the first receiver and the first adjustment model (613). According to the technical scheme of the invention, even if the first transmitter and the second receiver or the first receiver and the second transmitter are from different manufacturers, and joint training is not carried out, the communication between the first device and the second device can be realized through the first adjustment model.

Description

A communication method, a model training method and a corresponding device Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a communication method, a model training method and corresponding devices. Background Art

[0002] Neural network-based transmitters and receivers can be used for physical layer signal processing, such as symbol modulation and demodulation, channel encoding and decoding, pilot and channel estimation, and compression and reconstruction of channel state information.

[0003] The neural networks of the transmitter and receiver are typically optimized jointly to achieve optimal performance. However, when communication is required between different vendors or devices, the transmitter / receiver trained on one end must be adapted to the receiver / transmitter trained on the other end. Independently trained transmitter / receiver models from different vendors cannot communicate directly with each other.

[0004] Therefore, how to achieve communication between transmitters and receivers from different manufacturers without disclosing the models of the transmitters / receivers of each manufacturer has become an urgent problem to be solved.

[0005] Summary of the Invention

[0006] The present application provides a communication method for enabling transmitters and receivers from different manufacturers to communicate by adjusting a model. The present application also provides a model training method, a corresponding communication device, a computer-readable storage medium, and a computer program product.

[0007] The first aspect of the present application provides a communication method, including: a first device determines a first adjustment model based on first information from a second device, the first adjustment model is associated with information of the first transmitter and information of the second receiver, or the first adjustment model is associated with information of the first receiver and information of the second transmitter, wherein the first transmitter and the first receiver correspond to the first device, and the second receiver and the second transmitter correspond to the second device; the first device communicates with the second receiver of the second device based on the first transmitter and the first adjustment model; or the first device communicates with the second transmitter of the second device based on the first receiver and the first adjustment model.

[0008] In this application, the first device and the second device can be devices or chips (systems) in devices. When the first device or the second device is a device, the first device can be a terminal device and the second device can be a network device or a terminal device; or, the first device can be a network device and the second device can be a terminal device or a network device. When the first device or the second device is a chip (system), the first device can be a chip (system) in a terminal device and the second device can be a chip (system) in a network device or a terminal device; or, the first device can be a chip (system) in a network device and the second device can be a chip (system) in a terminal device or a network device.

[0009] In this application, the first information may be an identifier of the first adjustment model or associated information that can be used to determine the first adjustment model. The first adjustment model may be provided in the first device in the form of a module / unit, or may be provided independently of the first device.

[0010] In the present application, the first adjustment model may correspond to the first transmitter or the first receiver, and the first adjustment model corresponding to the first transmitter may be different from the first adjustment model corresponding to the first receiver.

[0011] In this application, the information of the first transmitter may be an identifier of the first transmitter or other information that can indicate the first transmitter. The first adjustment model being associated with the information of the first transmitter and the information of the second receiver means that the first adjustment model can be determined (e.g., obtained by searching or calculating) through the information of the first transmitter and the information of the second receiver. Similarly, the first adjustment model being associated with the information of the first receiver and the information of the second transmitter means that the first adjustment model can be determined (e.g., obtained by searching or calculating) through the information of the first receiver and the information of the second transmitter.

[0012] In the present application, the functions of the first transmitter / second transmitter and the first receiver / second receiver may be relative, for example: the first transmitter / second transmitter may include an encoder, and the first receiver / second receiver may include a decoder; or, the first transmitter / second transmitter may include a modulator, and the first receiver / second receiver may include a demodulator; or, the first transmitter / second transmitter may include a compression unit, and the first receiver / second receiver may include a decompression unit.

[0013] In this application, if the first device is a terminal device or a network device, then the first receiver and the first transmitter corresponding to the first device means that the first device includes the first receiver and the first transmitter. If the first device is a chip (system) in a terminal device or a chip (system) in a network device, then the first receiver and the first transmitter corresponding to the first device means that the first receiver and the first transmitter are provided in the same device as the first device, such as both provided in the terminal device or the network device.

[0014] In the first aspect described above, the first device can determine a first adjustment model and then communicate with the second receiver of the second device based on the first transmitter and the first adjustment model, or communicate with the second transmitter of the second device based on the first receiver and the first adjustment model. In this way, even if the first transmitter and the second receiver, or the first receiver and the second transmitter, are from different manufacturers and have not undergone joint training, communication between the first and second devices can be achieved using the first adjustment model.

[0015] In one possible implementation, the first device communicates with the second receiver of the second device based on the first transmitter and the first adjustment model, including: the first device processes the signal to be transmitted based on the first transmitter, and then processes the signal processed by the first transmitter based on the first adjustment model; the first device sends the signal processed by the first adjustment model to the second device.

[0016] The signals appearing in this application can also be expressed as data, symbols or bits. The only difference is the way of expression. In essence, they are all data signals.

[0017] In this possible implementation, after the first transmitter processes the signal to be transmitted (x), the first adjustment model processes the signal processed by the first transmitter (f1(x)), and then sends the signal processed by the first adjustment model (t1(f1(x))), where f1 can represent the first transmitter and t1 represents the first adjustment model. Because the second device participates in the training of the first adjustment model, the signal processed by the first adjustment model can be correctly received by the second receiver, thereby realizing effective communication between the first transmitter and the second receiver.

[0018] It should be noted that the above-mentioned first transmitter first processes the signal to be sent (x), and the first adjustment model then processes the signal processed by the first transmitter (f1(x)). This is only a possible implementation method. Other processing orders can also be adopted, such as: the first adjustment model first processes the signal to be sent, and then the first transmitter processes the signal processed by the first adjustment model. This application does not limit this.

[0019] In one possible implementation, the first device communicates with the second transmitter of the second device based on the first receiver and the first adjustment model, including: the first device processes the signal from the second transmitter of the second device based on the first adjustment model; the first device processes the signal processed by the first adjustment model based on the first receiver.

[0020] In this possible implementation, the first adjustment model processes a signal (f2(x)) from a second transmitter of a second device to obtain t1(f2(x)), and the first receiver processes t1(f2(x)) to obtain g1(t1(f2(x))), where g1 represents the first receiver and f2 represents the second transmitter. Because the second device participated in the training of the first adjustment model, the first receiver can correctly receive the signal processed by the first adjustment model, thereby achieving effective communication between the second transmitter and the first receiver.

[0021] It should be noted that the above-mentioned first adjustment model first processes the signal sent by the second transmitter (f2(x)), and then the first receiver processes the signal processed by the first adjustment model (t1(f2(x))). This is only a possible implementation method. Other processing orders can also be adopted, such as: the first receiver first processes the signal sent by the second transmitter, and then the first adjustment model processes the signal processed by the first receiver. This application does not limit this.

[0022] In a possible implementation, the first information includes an identifier of the first adjustment model, and the identifier of the first adjustment model is used to determine the first adjustment model.

[0023] In this possible implementation, if the second device has determined the identifier of the first adjustment model communicating with the first device, the identifier of the first adjustment model may be fed back to the first device, thereby enabling the first device to quickly determine the first adjustment model.

[0024] In a possible implementation manner, the first information includes an identifier of the second receiver or an identifier of the second transmitter; the identifier of the second receiver or the identifier of the second transmitter is used to determine an identifier of the first adjustment model.

[0025] In this possible implementation, if the second device provides feedback to the first device regarding the identifier of the second receiver communicating with the first transmitter, the first device needs to search for the identifier of the first adjustment model based on the identifiers of the first transmitter and the second receiver, thereby determining the first adjustment model. If the second device provides feedback to the first device regarding the identifier of the second transmitter communicating with the first receiver, the first device needs to search for the identifier of the first adjustment model based on the identifiers of the first receiver and the second transmitter, thereby determining the first adjustment model. This approach increases the diversity of first information feedback forms.

[0026] In one possible implementation, the identifier of the first adjustment model is determined by the first device from a first identifier group based on the identifier of the first transmitter and the identifier of the second receiver, and the first identifier group includes the identifier of the first transmitter, the identifier of the second receiver, and the identifier of the first adjustment model; or, the identifier of the first adjustment model is determined by the first device from a second identifier group based on the identifier of the first receiver and the identifier of the second transmitter, and the second identifier group includes the identifier of the second transmitter, the identifier of the first receiver, and the identifier of the first adjustment model.

[0027] In this possible implementation, both the first and second devices can maintain a first identification group and a second identification group, and each of the first and second identification groups can maintain three identifications. For example, the identification group (Tx01, M01, Re02) indicates that the identification of the first adjustment model associated with communication between a transmitter identified as Tx01 and a receiver identified as Re02 is M01. Using the identification group, both the first and second devices can quickly locate the adjustment model to be used for communication between the transmitter and receiver corresponding to the two devices, thereby improving communication efficiency.

[0028] In one possible implementation, before the first device determines the first adjustment model based on the first information from the second device, the method also includes: the first device trains the neural network model according to the first training objective to obtain the first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal, the first signal is a signal obtained by the first transmitter processing the third signal and then processing it by the neural network model participating in the training, the second signal is a signal obtained by the second transmitter processing the third signal, and the third signal is the source signal.

[0029] In this possible implementation, the source signal (x) can be understood as the signal before being processed by the first transmitter and the first adjustment model. The source signal (x) can be provided by the first device or obtained from other channels. This application does not limit this. The first signal is the signal (t1(f1(x))) obtained after being processed by the first transmitter and then processed by the neural network model participating in the training (the first adjustment model before the training is completed), and the second signal is the signal (f2(x)) obtained by the second transmitter processing the third signal. The first training goal can be understood as making the first signal and the second signal as close as possible, or making the error between the first signal and the second signal as small as possible. That is, the signal (t1(f1(x))) after the joint processing of the first transmitter and the first adjustment model can be as close as possible to the signal (f2(x)) obtained after the second transmitter processes the same source signal. Because the second receiver and the second transmitter are usually jointly trained, the second receiver can correctly receive the data processed by the second transmitter. If t1(f1(x)) is as close as possible to f2(x), the second receiver can also correctly receive t1(f1(x)). Therefore, the training process can enable the first device and the second device to communicate normally.

[0030] In one possible implementation, before the first device determines the first adjustment model based on the first information from the second device, the method also includes: the first device trains the neural network model according to the second training objective to obtain the first adjustment model, wherein the second training objective is used to indicate that the loss function of the neural network model participating in the training is reduced according to the fourth signal, the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is the source signal.

[0031] In this possible implementation, the source signal (x) can be understood as the signal before being processed by the first transmitter and the first adjustment model. The source signal (x) can be provided by the first device or obtained from other channels. This application does not limit this. The fifth signal is the signal (t1(f1(x))) after the first transmitter processes the source signal (x) and then the signal is processed by the first adjustment model. The fourth signal is the signal obtained by the second receiver processing the fifth signal (t1(f1(x))). The processing process may include first decoding, demodulating, etc. the fifth signal (t1(f1(x))) to obtain g2(t1(f1(x))), and then calculating the loss function L of g2(t1(f1(x))), such as mean square error (mse)(x,g2(t1(f1(x)))) or cross-entropy (CE)(x,g2(t1(f1(x)))), and performing differentiation to obtain the gradient Here, g2 can represent the second receiver. The second training objective can be understood as the first device adjusting the weights of the neural network model involved in training based on the gradient feedback from the second device, minimizing the loss function. The trained first adjusted model can then enable normal communication between the first transmitter and the second receiver.

[0032] The above-mentioned first training target and second training target are both used to train the neural network model and the scenario corresponding to the first transmitter, and the first adjustment model is represented by t1.

[0033] In one possible implementation, before the first device determines the first adjustment model based on the first information from the second device, the method also includes: the first device trains the neural network model according to the third training objective to obtain the first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal, the seventh signal is a signal obtained by the first receiver processing the ninth signal with the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal.

[0034] In this possible implementation, the source signal (x) can be understood as the signal before being processed by the second transmitter. The source signal (x) can be provided by the second device or obtained from other means, and this application does not limit this. The ninth signal is the signal (f2(x)) obtained by the second transmitter processing the source signal (x), the seventh signal is the signal (g1(t1(f2(x)))) obtained by the first receiver processing the ninth signal by the first adjustment model (t1(f2(x))), and the eighth signal is the signal obtained by the source signal or the second receiver processing the ninth signal. The third training goal can be understood as making the seventh signal as close as possible to the eighth signal, or making the error between the seventh signal and the eighth signal as small as possible, that is, the signal (g1(t1(f2(x)))) after the first adjustment model and the first receiver jointly processed can be as close as possible to the source signal (x) or the signal (g2(f2(x))) after the second receiver processes the signal (f2(x)). In this way, the result of the joint processing of the signal (f2(x)) encoded or modulated by the second transmitter by the first adjustment model and the first receiver is substantially consistent with the signal (x) before encoding or modulation by the second transmitter or the result of the second receiver's processing of the signal (f2(x)) (g2(f2(x))). Because the second receiver and the second transmitter are typically jointly trained, the second receiver can correctly receive the data processed by the second transmitter. If g1(t1(f2(x))) is as close as possible to x or g2(f2(x)), the first receiver can also correctly receive the signal (f2(x)) sent by the second transmitter. Therefore, this training process enables the first and second devices to communicate normally.

[0035] It should be noted that the first adjustment model in this possible implementation corresponds to the first receiver, which is represented by t2.

[0036] In a possible implementation, the method further includes: the first device sending the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

[0037] In this possible implementation, the first device repeatedly executes the training process for the first or second training target to obtain a first adjustment model. This first adjustment model can then be sent to the second device, which then uses it to derive a second adjustment model suitable for the second transmitter. This eliminates the need for the second device to retrain an adjustment model suitable for the second transmitter, reducing computational complexity and improving communication efficiency.

[0038] In one possible implementation, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0039] In this possible implementation, the second adjustment model and the first adjustment model are mutually inverse neural networks, which can reduce the overhead of acquiring the second adjustment model.

[0040] In a possible implementation, the method further includes: the first device sending a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0041] In this possible implementation, after the first device trains a first adjustment model for the first receiver using the third training target, it can send the first adjustment model to the second device, allowing the second device to derive a third adjustment model for the second receiver using the first adjustment model. This eliminates the need for the second device to train an adjustment model for the second receiver, reducing computational complexity and improving communication efficiency.

[0042] In one possible implementation, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0043] In this possible implementation, the third adjustment model and the first adjustment model are mutually inverse neural networks, which can reduce the cost of obtaining the third adjustment model.

[0044] In one possible implementation, if there are multiple first adjustment models, and the multiple first adjustment models are connected in parallel; each first adjustment model in parallel is used for the first transmitter to communicate with a different second receiver; or, each first adjustment model in parallel is used for the first receiver to communicate with a different second transmitter.

[0045] In this possible implementation, there may be multiple first adjustment models connected in parallel, so that the first transmitter or the first receiver can use different first adjustment models to communicate with different second receivers or second transmitters, thereby improving the flexibility of communication.

[0046] In one possible implementation, if there are multiple first adjustment models, and the multiple first adjustment models are connected in series; at least two directly connected first adjustment models in series are used for the first transmitter to communicate with different second receivers; or, at least two directly connected first adjustment models in series are used for the first receiver to communicate with different second transmitters.

[0047] In this possible implementation, multiple first adjustment models may be connected in series, so that the first transmitter or the first receiver may use one or more first adjustment models to communicate with different second receivers or second transmitters, thereby improving the flexibility of communication.

[0048] A second aspect of the present application provides a method for model training, comprising: a first device obtains a signal for model training, the signal for model training includes a first signal and a second signal; wherein the first signal is a signal obtained by a first transmitter after processing a third signal and then processed by a neural network model participating in the training, the second signal is a signal obtained by a second transmitter after processing the third signal, and the third signal is a source signal; the first device trains the neural network model according to a first training objective to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal.

[0049] In the second aspect, the source signal (x) can be understood as the signal before being processed by the first transmitter and the trained neural network model (the first adjustment model before training is complete). The first signal is the signal (t1(f1(x))) obtained after processing by the first transmitter and then by the trained neural network model. The second signal is the signal (f2(x)) obtained by the second transmitter processing the third signal. The first training objective can be understood as making the first and second signals as close as possible, or minimizing the error between the first and second signals. In other words, the signal (t1(f1(x))) jointly processed by the first transmitter and the first adjustment model should be as close as possible to the signal (f2(x)) obtained by processing the same source signal by the second transmitter. Because the second receiver and the second transmitter are typically jointly trained, the second receiver can correctly receive data processed by the second transmitter. If t1(f1(x)) is as close as possible to f2(x), the second receiver can also correctly receive t1(f1(x)). Therefore, this training process enables normal communication between the first and second devices.

[0050] In a possible implementation, the method further includes: the first device sending a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

[0051] In this possible implementation, the first device repeatedly executes the training process for the first or second training target to obtain a first adjustment model. This first adjustment model can then be sent to the second device, which then uses it to derive a second adjustment model suitable for the second transmitter. This eliminates the need for the second device to retrain an adjustment model suitable for the second transmitter, reducing computational complexity and improving communication efficiency.

[0052] In a possible implementation, the method further includes: the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0053] In this possible implementation, the second adjustment model and the first adjustment model are mutually inverse neural networks, which can reduce the overhead of acquiring the second adjustment model.

[0054] A third aspect of the present application provides a method for model training, including: a first device obtains a signal for model training, the signal for model training includes a fourth signal and a fifth signal; wherein the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal; the first device trains the neural network model according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate reducing the loss function of the neural network model participating in the training according to the fourth signal.

[0055] In the above-mentioned third aspect, the source signal (x) can be understood as the signal before being processed by the first transmitter and the first adjustment model, the fifth signal is the signal (t1(f1(x))) after the first transmitter processes the source signal (x) and is processed by the first adjustment model, and the fourth signal is the signal obtained by the second receiver processing the fifth signal (t1(f1(x))). The processing process may include first decoding, demodulating, etc. the fifth signal (t1(f1(x))) to obtain g2(t1(f1(x))), and then calculating the loss function L of g2(t1(f1(x))), such as mse(x,g2(t1(f1(x)))) or cross entropy CE(x,g2(t1(f1(x)))), and performing differentiation to obtain the gradient Here, g2 can represent the second receiver. The second training objective can be understood as the first device adjusting the weights of the neural network model involved in training based on the gradient feedback from the second device, minimizing the loss function. The trained first adjusted model can then enable normal communication between the first transmitter and the second receiver.

[0056] In a possible implementation, the method further includes: the first device sending a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

[0057] In this possible implementation, the first device repeatedly executes the training process for the first or second training target to obtain a first adjustment model. This first adjustment model can then be sent to the second device, which then uses it to derive a second adjustment model suitable for the second transmitter. This eliminates the need for the second device to retrain an adjustment model suitable for the second transmitter, reducing computational complexity and improving communication efficiency.

[0058] In a possible implementation, the method further includes: the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0059] In this possible implementation, the second adjustment model and the first adjustment model are mutually inverse neural networks, which can reduce the overhead of acquiring the second adjustment model.

[0060] The fourth aspect of the present application provides a method for model training, including: a first device obtains a signal for model training, the signal for model training includes a seventh signal and an eighth signal; wherein the seventh signal is a signal obtained by the first receiver processing the ninth signal of the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal; the first device trains the neural network model according to the third training target to obtain a first adjustment model, wherein the third training target is used to indicate reducing the error between the seventh signal and the eighth signal.

[0061] In the fourth aspect described above, the source signal (x) can be understood as the signal before being processed by the second transmitter. The source signal (x) can be provided by the second device or obtained from other means, and this application does not limit this. The ninth signal is the signal (f2(x)) obtained by the second transmitter processing the source signal (x), the eighth signal is the signal (g2(f2(x))) obtained by the source signal (x) or the second receiver processing the ninth signal, and the seventh signal is the signal (g1(t1(f2(x)))) obtained by the first receiver processing the ninth signal by the first adjustment model. The third training goal can be understood as making the seventh signal as close as possible to the eighth signal, or making the error between the seventh signal and the eighth signal as small as possible, that is, the signal (g1(t1(f2(x)))) after the first adjustment model and the first receiver jointly processing can be as close as possible to the source signal (x) or the signal (g2(f2(x))) after the second receiver processes the signal (f2(x)). In this way, the result of the first adjustment model and the first receiver jointly processing the signal (f2(x)) encoded or modulated by the second transmitter is substantially consistent with the signal (x) before encoding or modulation by the second transmitter or the result (g2(f2(x))) of the second receiver processing the signal (f2(x)). Because the second receiver and the second transmitter are typically jointly trained, the second receiver can correctly receive the data processed by the second transmitter. If g1(t1(f2(x))) is as close as possible to x or g2(f2(x)), the first receiver can also correctly receive the signal (f2(x)) sent by the second transmitter. Therefore, this training process enables the first device and the second device to communicate normally.

[0062] In a possible implementation, the first device sends a first adjustment model to the second device, and the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0063] In this possible implementation, after the first device trains a first adjustment model for the first receiver using the third training target, it can send the first adjustment model to the second device, allowing the second device to derive a third adjustment model for the second receiver using the first adjustment model. This eliminates the need for the second device to train an adjustment model for the second receiver, reducing computational complexity and improving communication efficiency.

[0064] In one possible implementation, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0065] In this possible implementation, the third adjustment model and the first adjustment model are mutually inverse neural networks, which can reduce the cost of obtaining the third adjustment model.

[0066] A fifth aspect of the present application provides a communication device, including:

[0067] A processing module is used to determine a first adjustment model based on first information from a second device, wherein the first adjustment model is associated with information of the first transmitter and information of the second receiver, or the first adjustment model is associated with information of the first receiver and information of the second transmitter, wherein the first transmitter and the first receiver correspond to the first device, and the second receiver and the second transmitter correspond to the second device.

[0068] The processing module is further configured to communicate with the second receiver of the second device based on the first transmitter and the first adjustment model; or the first device communicates with the second transmitter of the second device based on the first receiver and the first adjustment model.

[0069] In a possible implementation, the processing module is specifically configured to process the signal to be transmitted based on the first transmitter, and then process the signal processed by the first transmitter based on the first adjustment model.

[0070] The communication device further includes a transceiver module configured to transmit a signal processed by the first adjustment model to the second device. The transceiver module refers to a module having transceiver functions, having the functions of a transmitter and a receiver, and may include a first transmitter and a first receiver, or the transceiver module includes modules on the first transmitter and the first receiver.

[0071] In a possible implementation, the processing module is specifically configured to process a signal from a second transmitter of the second device based on the first adjustment model, and process the signal processed by the first adjustment model based on the first receiver.

[0072] In a possible implementation, the first information includes an identifier of the first adjustment model, and the identifier of the first adjustment model is used to determine the first adjustment model.

[0073] In a possible implementation, the first information includes an identifier of the second receiver or an identifier of the second transmitter; the identifier of the second receiver or the identifier of the second transmitter is used to determine an identifier of the first adjustment model.

[0074] In one possible implementation, the identifier of the first adjustment model is determined by the first device from a first identifier group based on the identifier of the first transmitter and the identifier of the second receiver, and the first identifier group includes the identifier of the first transmitter, the identifier of the second receiver, and the identifier of the first adjustment model; or, the identifier of the first adjustment model is determined by the first device from a second identifier group based on the identifier of the first receiver and the identifier of the second transmitter, and the second identifier group includes the identifier of the second transmitter, the identifier of the first receiver, and the identifier of the first adjustment model.

[0075] In one possible implementation, the processing module is also used to train the neural network model according to the first training objective to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal, the first signal is a signal obtained by the first transmitter processing the third signal and then processing it by the neural network model participating in the training, the second signal is a signal obtained by the second transmitter processing the third signal, and the third signal is the source signal.

[0076] In one possible implementation, the processing module is also used to train the neural network model according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate reducing the loss function of the neural network model participating in the training according to the fourth signal, the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal.

[0077] In one possible implementation, the processing module is also used to train the neural network model according to the third training objective to obtain a first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal, the seventh signal is a signal obtained by the first receiver processing the ninth signal by the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal.

[0078] In a possible implementation, the transceiver module is further configured to send the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

[0079] In one possible implementation, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0080] In a possible implementation, the transceiver module is further configured to send the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0081] In one possible implementation, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0082] In one possible implementation, if there are multiple first adjustment models, and the multiple first adjustment models are connected in parallel; each first adjustment model in parallel is used for the first transmitter to communicate with a different second receiver; or, each first adjustment model in parallel is used for the first receiver to communicate with a different second transmitter.

[0083] In one possible implementation, if there are multiple first adjustment models, and the multiple first adjustment models are connected in series; at least two directly connected first adjustment models in series are used for the first transmitter to communicate with different second receivers; or, at least two directly connected first adjustment models in series are used for the first receiver to communicate with different second transmitters.

[0084] A sixth aspect of the present application provides a communication device, including:

[0085] A processing module is used to obtain signals for model training, and the signals for model training include a first signal and a second signal; wherein the first signal is a signal obtained by processing a third signal by a first transmitter and then processed by a neural network model participating in the training, and the second signal is a signal obtained by processing the third signal by a second transmitter, and the third signal is a source signal; according to a first training objective, the neural network model is trained to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal.

[0086] In one possible implementation, the communication device further includes a transceiver module configured to transmit the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter. The transceiver module refers to a module having transceiver functionality, including both transmitter and receiver functions. The transceiver module may include the first transmitter and the first receiver, or the transceiver module may include modules on the first transmitter and the first receiver.

[0087] In one possible implementation, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0088] A seventh aspect of the present application provides a communication device, including:

[0089] A processing module is used to obtain signals for model training, and the signals for model training include a fourth signal and a fifth signal; wherein the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal; the neural network model is trained according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate reducing the loss function of the neural network model participating in the training according to the fourth signal.

[0090] In one possible implementation, the communication device further includes a transceiver module configured to transmit the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter. The transceiver module refers to a module having transceiver functionality, including both transmitter and receiver functions. The transceiver module may include the first transmitter and the first receiver, or the transceiver module may include modules on the first transmitter and the first receiver.

[0091] In one possible implementation, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0092] An eighth aspect of the present application provides a communication device, including:

[0093] A processing module is used to obtain signals for model training, and the signals for model training include a seventh signal and an eighth signal; wherein the seventh signal is a signal obtained by reprocessing the ninth signal obtained by the first receiver on the neural network model participating in the training, the ninth signal is obtained by processing the source signal by the second transmitter, and the eighth signal is obtained by processing the ninth signal by the source signal or the second receiver; the neural network model is trained according to the third training objective to obtain a first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal.

[0094] In one possible implementation, the communication device further includes a transceiver module configured to transmit the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver. The transceiver module refers to a module having transceiver functionality, including both transmitter and receiver functions. The transceiver module may include a first transmitter and a first receiver, or may include modules on both the first transmitter and the first receiver.

[0095] In one possible implementation, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0096] The communication device described in the fifth to eighth aspects above may be a device or a chip (system) in the device. As an example, the processing module may be a processor, and the transceiver module may be a transceiver or a communication interface. It is understandable that the transceiver module may be a transceiver in the communication device, for example, implemented by an antenna, feeder, and / or codec in the communication device, or, if the communication device is a chip (system) provided in the device, the transceiver module may be an input / output interface of the chip (system), such as an input / output circuit, a pin, etc.

[0097] In a ninth aspect, the present application provides a communication device, comprising a processor configured to call and execute a computer program stored in a memory, so that the processor implements the first aspect or any one of the implementations of the first aspect.

[0098] Optionally, the communication device further includes a transceiver; the processor is further configured to control the transceiver to transmit and receive signals.

[0099] Optionally, the communication device includes a memory in which a computer program is stored.

[0100] In a tenth aspect, the present application provides a communication device, comprising a processor, wherein the processor is configured to call and execute a computer program stored in a memory, so that the processor implements the second aspect or any one of the implementations of the second aspect.

[0101] Optionally, the communication device further includes a transceiver; the processor is further configured to control the transceiver to transmit and receive signals.

[0102] Optionally, the communication device includes a memory in which a computer program is stored.

[0103] In an eleventh aspect of the present application, a communication device is provided, comprising a processor configured to call and execute a computer program stored in a memory, so that the processor implements the third aspect or any one of the implementations of the third aspect.

[0104] Optionally, the communication device further includes a transceiver; the processor is further configured to control the transceiver to transmit and receive signals.

[0105] Optionally, the communication device includes a memory in which a computer program is stored.

[0106] In a twelfth aspect, the present application provides a communication device, comprising a processor, wherein the processor is configured to call and execute a computer program stored in a memory, so that the processor implements the fourth aspect or any one of the implementations of the fourth aspect.

[0107] Optionally, the communication device further includes a transceiver; the processor is further configured to control the transceiver to transmit and receive signals.

[0108] Optionally, the communication device includes a memory in which a computer program is stored.

[0109] The communication device described in the ninth to twelfth aspects above may be a device or a chip (system) in the device.

[0110] A thirteenth aspect of the present application provides a computer program product comprising instructions, characterized in that when the computer program product is run on a computer, the computer is caused to execute the first aspect or any one of the implementation methods of the first aspect.

[0111] The fourteenth aspect of the present application provides a computer program product comprising instructions, characterized in that when the computer program product is run on a computer, the computer is enabled to execute the second aspect or any one of the implementation methods of the second aspect.

[0112] A fifteenth aspect of the present application provides a computer program product comprising instructions, characterized in that when the computer program product is run on a computer, the computer is enabled to execute the third aspect or any one of the implementation methods of the third aspect.

[0113] The sixteenth aspect of the present application provides a computer program product comprising instructions, characterized in that when the computer program product is run on a computer, the computer is enabled to execute the fourth aspect or any one of the implementation methods of the fourth aspect.

[0114] In the seventeenth aspect, the present application provides a computer-readable storage medium comprising computer instructions. When the computer instructions are executed on a computer, the computer executes the first aspect or any one of the implementation methods of the first aspect.

[0115] In an eighteenth aspect, the present application provides a computer-readable storage medium comprising computer instructions. When the computer instructions are executed on a computer, the computer executes the second aspect or any one of the implementation methods of the second aspect.

[0116] The nineteenth aspect of the present application provides a computer-readable storage medium comprising computer instructions. When the computer instructions are executed on a computer, the computer executes the third aspect or any one of the implementation methods of the third aspect.

[0117] The twentieth aspect of the present application provides a computer-readable storage medium comprising computer instructions. When the computer instructions are executed on a computer, the computer executes the fourth aspect or any one of the implementation methods of the fourth aspect.

[0118] The twenty-first aspect of the present application provides a chip device, including a processor, which is used to be connected to a memory and call a program stored in the memory so that the processor executes the above-mentioned first aspect or any one of the implementation methods of the first aspect.

[0119] The twenty-second aspect of the present application provides a chip device, including a processor, which is used to be connected to a memory and call a program stored in the memory so that the processor executes the above-mentioned second aspect or any one of the implementation methods of the second aspect.

[0120] The twenty-third aspect of the present application provides a chip device, including a processor, which is used to be connected to a memory and call a program stored in the memory so that the processor executes the above-mentioned third aspect or any one of the implementation methods of the third aspect.

[0121] The twenty-fourth aspect of the present application provides a chip device, including a processor, which is used to be connected to a memory and call a program stored in the memory so that the processor executes the above-mentioned fourth aspect or any one of the implementation methods of the fourth aspect.

[0122] The twenty-fifth aspect of the present application provides a communication system, which includes a first device and a second device, and the first device is communicatively connected to the second device. The first device can be the communication device described in the fifth aspect, sixth aspect, seventh aspect, eighth aspect or any one of the implementation methods therein.

[0123] Regarding the technical effects of the above-mentioned fifth aspect and any one of the implementation methods of the fifth aspect, as well as the technical effects of the ninth, thirteenth, seventeenth and twenty-first aspects, they can be understood by referring to the technical effects of the above-mentioned first aspect and any one of the implementation methods of the first aspect.

[0124] The technical effects of the above-mentioned sixth aspect and any one of the implementation methods of the sixth aspect, as well as the technical effects of the tenth, fourteenth, eighteenth and twenty-second aspects, can be understood by referring to the technical effects of the above-mentioned second aspect and any one of the implementation methods of the second aspect.

[0125] The technical effects of the above-mentioned seventh aspect and any one of the implementation methods of the seventh aspect, as well as the technical effects of the eleventh, fifteenth, nineteenth and twenty-third aspects, can be understood by referring to the technical effects of the above-mentioned third aspect and any one of the implementation methods of the third aspect.

[0126] The technical effects of the above-mentioned eighth aspect and any one of the implementation methods of the eighth aspect, as well as the technical effects of the twelfth, sixteenth, twentieth and twenty-fourth aspects, can be understood by referring to the technical effects of the above-mentioned fourth aspect and any one of the implementation methods of the fourth aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0127] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0128] FIG1 is a schematic diagram of an architecture of a communication system provided in an embodiment of the present application;

[0129] FIG2 is another schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0130] FIG3A is another schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0131] FIG3B is another schematic diagram of the architecture of the communication system provided in an embodiment of the present application;

[0132] FIG4A is a schematic diagram of an embodiment of a method for model training provided in an embodiment of the present application;

[0133] FIG4B is a schematic diagram of another embodiment of the model training method provided in an embodiment of the present application;

[0134] FIG4C is a schematic diagram of another embodiment of the model training method provided in an embodiment of the present application;

[0135] FIG5A is a schematic diagram of an example of a reversible function provided in an embodiment of the present application;

[0136] FIG5B is another exemplary schematic diagram of a reversible function provided in an embodiment of the present application;

[0137] FIG5C is another exemplary schematic diagram of a reversible function provided in an embodiment of the present application;

[0138] FIG6A is a schematic diagram of an embodiment of a communication method provided in an embodiment of the present application;

[0139] FIG6B is a schematic diagram of another embodiment of the communication method provided in an embodiment of the present application;

[0140] FIG7A is another schematic diagram of the architecture of a communication system provided in an embodiment of the present application;

[0141] FIG7B is another schematic diagram of the architecture of the communication system provided in an embodiment of the present application;

[0142] FIG7C is another schematic diagram of the architecture of the communication system provided in an embodiment of the present application;

[0143] FIG7D is another schematic diagram of the architecture of the communication system provided in an embodiment of the present application;

[0144] FIG8 is a schematic structural diagram of a communication device provided in an embodiment of the present application;

[0145] FIG9 is a schematic structural diagram of a terminal device provided in an embodiment of the present application;

[0146] FIG10 is a schematic structural diagram of a network device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0147] Embodiments of the present application provide a communication method for enabling communication between transmitters and receivers from different manufacturers by adjusting a model. The present application also provides a model training method, a corresponding communication device, a computer-readable storage medium, and a computer program product. Each of these is described in detail below.

[0148] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0149] The technical solutions of the embodiments of the present application can be applied to various wireless communication systems, such as: fifth generation (5G) system or new radio (NR), long term evolution (LTE) system, LTE frequency division duplex (FDD) system, LTE time division duplex (TDD), universal mobile telecommunication system (UMTS), mobile communication systems after 5G network (for example, 6G mobile communication system), vehicle to everything (V2X) communication system, etc. The wireless communication system to which the present application is applicable includes terminal equipment and network equipment, and communication transmission can be carried out between the terminal equipment and the network equipment.

[0150] The terminal device and network device of this application are introduced below.

[0151] The terminal device may be a wireless terminal device capable of receiving network device scheduling and instruction information. The wireless terminal device may be a device that provides voice and / or data connectivity to a user, or a handheld device with wireless connection function, or other processing device connected to a wireless modem.

[0152] Terminal devices, also known as user equipment (UE), mobile stations (MS), and mobile terminals (MT), are devices that include wireless communication capabilities (providing voice / data connectivity to users), such as handheld devices with wireless connection capabilities. Currently, some examples of terminal devices include: mobile phones, tablets, laptops, PDAs, notebook computers, wireless routers, mobile internet devices (MIDs), wearable devices, virtual reality (VR) devices, augmented reality (AR) devices, wireless terminals in industrial control, wireless terminals in self-driving cars, wireless terminals in the Internet of Vehicles (IoV), wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, and wireless terminals in smart homes. For example, wireless terminals in the IoV can be vehicle-mounted devices, complete vehicle equipment, vehicle-mounted modules, vehicles, etc. The wireless terminal in industrial control can be a robot, etc. For example, the wireless terminal in unmanned driving can be a drone.

[0153] A network device may be a device in a wireless network. For example, a network device may be a radio access network (RAN) node that connects a terminal device to a wireless network, and may also be referred to as a radio access network device. A radio access network device is a device deployed in a radio access network to provide wireless communication functions for a terminal device. A non-limiting example of a radio access network device is a base station, and a base station is various forms of macro base stations, micro base stations (also known as small stations), relay stations, access points (APs), wearable devices, vehicle-mounted devices, and the like. A base station may also be a transmission and reception point (TRP), a transmission measurement function (TMF), and the like. Exemplarily, the base station involved in the embodiments of the present application may be a base station in a new radio (NR). Among them, the base station in 5G NR can also be called a transmission reception point (TRP) or a transmission point (TP) or a next generation Node B (ngNB), or an evolutionary Node B (eNB or eNodeB) in the long term evolution (LTE) system. In a broad sense, it can also be a base band unit (BBU), a remote radio unit (RRU), an active antenna unit (AAU), a remote radio head (RRH), a centralized unit (CU), a distributed unit (DU), a positioning node, etc.

[0154] Figure 1 is a schematic diagram of a communication system provided in an embodiment of the present application. Please refer to Figure 1. The communication system shown in Figure 1 includes a network device and a terminal device. The communication system includes one or more network devices and one or more terminal devices. In the communication system, UE1 to UE6 can communicate with the network device. At the same time, UE4, UE5 and UE6 can also form a communication system. For example, the network device can send downlink information to UE5, and UE5 can send downlink information to UE4 or UE6. Optionally, the communication system shown in Figure 1 can be an LTE system, or a 5G mobile communication system, or a mobile communication system after the 5G network (for example, a 6G mobile communication system).

[0155] FIG2 is another schematic diagram of a communication system provided in an embodiment of the present application. Referring to FIG2 , the communication system shown in FIG2 includes a network device and a terminal device. The communication system includes one or more network devices and one or more terminal devices.

[0156] In this communication system, in one possible implementation, a network device can communicate with one or more terminal devices. For example, as shown in Figure 2, network device 1 communicates with terminal device 1 and terminal device 2, respectively. Network device 3 communicates with terminal device 2 and terminal device 3, respectively. In another possible implementation, multiple network devices can communicate with a single terminal device, meaning that multiple network devices can simultaneously provide services to a single terminal device. For example, network device 1, network device 2, and network device 3 can simultaneously provide services to terminal device 2.

[0157] The communication system provided in the embodiment of the present application will use an adjustment model in the process of communication between the terminal device and the network device in Figure 1 or Figure 2 above, so that the transmitter and receiver in the terminal device and network device that have not undergone joint training can adapt and communicate normally.

[0158] The communication system provided in the embodiment of the present application can also be understood by referring to Figures 3A and 3B.

[0159] 3A and 3B , the communication system includes a first device and a second device, wherein the first device includes a first transmitter 301 , a first receiver 302 , and a first adjustment model 303 , and the second device includes a second receiver 304 and a second transmitter 305 .

[0160] In Figure 3A, the first adjustment model 303 corresponds to the first transmitter 301. In Figure 3B, the first adjustment model 303 corresponds to the first receiver 302. The first adjustment models 303 in Figures 3A and 3B may be different.

[0161] 3A , the second device may further include a second adjustment model 306 corresponding to the second transmitter 305 . In FIG. 3B , the second device may further include a third adjustment model 307 corresponding to the second receiver 304 .

[0162] In the above Figures 3A and 3B, the first adjustment model 303 is set in the first device. In fact, the first adjustment model can be set in the first device in the form of a module / unit, or can be set independently of the first device. This application does not limit this.

[0163] In an embodiment of the present application, the functions of the first transmitter 301 / second transmitter 305 and the first receiver 302 / second receiver 304 may be relative, for example: the first transmitter 301 / second transmitter 305 may include an encoder, and the first receiver 302 / second receiver 304 may include a decoder; or, the first transmitter 301 / second transmitter 305 may include a modulator, and the first receiver 302 / second receiver 304 may include a demodulator; or, the first transmitter 301 / second transmitter 305 may include a compression unit, and the first receiver 302 / second receiver 304 may include a decompression unit.

[0164] It should be noted that the above-mentioned first device and second device can be devices or chips (systems) in devices. When the first device or the second device is a device, the first device can be a terminal device and the second device can be a network device or a terminal device; or, the first device can be a network device and the second device can be a terminal device or a network device. When the first device or the second device is a chip (system), the first device can be a chip (system) in a terminal device and the second device can be a chip (system) in a network device or a terminal device; or, the first device can be a chip (system) in a network device and the second device can be a chip (system) in a terminal device or a network device.

[0165] Based on the communication system shown in Figures 3A and 3B above, the embodiment of the present application can train an adjustment model and use the trained adjustment model for communication. The following first introduces the model training solution.

[0166] The model training scheme provided in the embodiment of the present application can be divided into a transmitter-side model training process and a receiver-side model training process. The transmitter-side model training process is based on the communication system with the structure shown in FIG3A above for model training, and the receiver-side model training process is based on the communication system with the structure shown in FIG3B above for model training. The following are introduced respectively:

[0167] 1. Transmitter-side model training scheme.

[0168] 1. Train according to the first training goal;

[0169] Referring to FIG. 4A , an embodiment of the model training method provided in the present application includes:

[0170] 401. The first device obtains a signal for model training, where the signal for model training includes a first signal and a second signal.

[0171] The first signal is a signal obtained by processing the third signal by the first transmitter and then processed by the neural network model participating in the training. The second signal is a signal obtained by processing the third signal by the second transmitter. The third signal is the source signal.

[0172] The source signal (x) can be understood as the signal before being processed by the first transmitter and the neural network model participating in the training (the first adjustment model before the training is completed). The source signal (x) can be provided by the first device or obtained from other channels. This application does not limit this.

[0173] The first signal is a signal (t1(f1(x))) obtained after being processed by the first transmitter and then by the neural network model participating in the training. Wherein, f1 can represent the first transmitter, and t1 represents the neural network model or the first adjustment model.

[0174] The second signal is a signal (f2(x)) obtained by the second transmitter processing the third signal, where f2 may represent the second transmitter.

[0175] 402. The first device trains a neural network model according to a first training objective to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal.

[0176] The first training objective can be understood as making the first and second signals as close as possible, or minimizing the error between them. Specifically, the signal (t1(f1(x))) processed by the first transmitter and the first adjustment model should be as close as possible to the signal (f2(x)) obtained by processing the same source signal by the second transmitter.

[0177] The loss function of the neural network model in this training scheme can be the difference between t1(f1(x)) and f2(x). For example, the mean square error (MSE) can be expressed as mse(t1(f1(x)),f2(x))=(t1(f1(x))-f2(x)). 2 .

[0178] The first device can calculate the loss function, update the weight of t1 through algorithms such as gradient descent, train the neural network model t1, and obtain the first adjustment model with the goal of minimizing the loss function.

[0179] By repeatedly executing the above-mentioned processes 401 and 402 through multiple source signals and iteratively updating the neural network model, a first adjustment model can be obtained.

[0180] Optionally, the above step 401 may include steps 401a, 401b and 401c.

[0181] 401a. The first device sends a third signal to the second device.

[0182] 401b. The first device receives a second signal from the second device.

[0183] 401c. The first device processes the third signal based on the first transmitter, and then uses the neural network model involved in the training to process the third signal processed by the first transmitter to obtain the first signal.

[0184] Optionally, step 403 and step 404 may be further included after step 402.

[0185] 403. The first device sends a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for a second transmitter.

[0186] The neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0187] 404. The second device determines a second adjustment model based on the first adjustment model.

[0188] The model training solution provided in the embodiment of the present application is that since the second receiver and the second transmitter are usually jointly trained, the second receiver can correctly receive the data processed by the second transmitter. If t1(f1(x)) is as close as possible to f2(x), the second receiver can also correctly receive t1(f1(x)). Therefore, the first device and the second device can communicate normally through this training process.

[0189] 2. Train according to the second training goal;

[0190] Referring to FIG. 4B , another embodiment of the model training method provided in the embodiment of the present application includes:

[0191] 411. The first device obtains a signal for model training, where the signal for model training includes a fourth signal and a fifth signal.

[0192] Among them, the fourth signal is the signal obtained by the second receiver processing the fifth signal, the fifth signal is the signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is the source signal.

[0193] The source signal (x) can be understood as the signal before being processed by the first transmitter and the first adjustment model. The source signal (x) can be provided by the first device or obtained from other channels, and this application does not limit this.

[0194] The fifth signal is a signal (t1(f1(x))) obtained by processing the source signal (x) by the first transmitter and then processed by the first adjustment model.

[0195] The fourth signal is a signal obtained by the second receiver processing the fifth signal (t1(f1(x))). The processing process may include first decoding, demodulating, etc. the fifth signal (t1(f1(x))) to obtain g2(t1(f1(x))), and then calculating the loss function L of g2(t1(f1(x))), such as mse(x,g2(t1(f1(x)))) or cross-entropy (CE)(x,g2(t1(f1(x)))), and performing derivation to obtain the gradient Among them, g2 can represent the second receiver.

[0196] 412. The first device trains the neural network model according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate that the loss function of the neural network model participating in the training is reduced according to the fourth signal.

[0197] The second training goal can be understood as the first device adjusting the weights of the neural network model participating in the training according to the gradient feedback from the second device, so that the loss function becomes as small as possible.

[0198] The first device can be Adjust the weights of the neural network model to update the neural network model.

[0199] By repeatedly executing the above-mentioned processes 411 and 412 through multiple source signals and iteratively updating the neural network model, a first adjustment model can be obtained.

[0200] Optionally, the above step 411 may include steps 411a, 411b and 411c.

[0201] 411a. The first device processes the sixth signal based on the first transmitter, and then uses the neural network model involved in the training to process the sixth signal processed by the first transmitter to obtain the fifth signal.

[0202] 411b. The first device sends a fifth signal to the second device.

[0203] 411c. The first device receives a fourth signal from the second device.

[0204] Optionally, step 413 and step 414 may be further included after step 412.

[0205] 413. The first device sends a first adjustment model to the second device, and the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

[0206] The neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0207] 414. The second device determines a second adjustment model based on the first adjustment model.

[0208] The model training solution provided in the embodiment of the present application makes the loss function as small as possible by jointly training the first device and the second device to obtain the first adjustment model, so the first adjustment model can enable the first transmitter and the second receiver to communicate normally.

[0209] 2. Receiver-side model training scheme.

[0210] Referring to FIG. 4C , another embodiment of the model training method provided in the embodiment of the present application includes:

[0211] 1. Train according to the third training goal;

[0212] 421. The first device obtains signals for model training, and the signals for model training include the seventh signal and the eighth signal.

[0213] Among them, the seventh signal is the signal obtained by the first receiver processing the ninth signal by the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal.

[0214] The source signal (x) can be understood as a signal before being processed by the second transmitter. The source signal (x) can be provided by the second device or obtained from other channels, and this application does not limit this.

[0215] The ninth signal is a signal (f2(x)) obtained by the second transmitter processing the ground source signal (x).

[0216] The eighth signal is the source signal (x) or the signal (g2(f2(x))) obtained by the second receiver processing the ninth signal.

[0217] The seventh signal is a signal (g1(t1(f2(x)))) obtained by further processing the signal (t1(f2(x))) obtained by the first receiver processing the ninth signal with the first adjustment model.

[0218] 422. The first device trains the neural network model according to the third training objective to obtain a first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal.

[0219] The third training goal can be understood as making the seventh signal and the eighth signal as close as possible, or making the error between the seventh signal and the eighth signal as small as possible, that is, the signal (g1(t1(f2(x)))) after joint processing by the first adjustment model and the first receiver can be as close as possible to the source signal (x) or the signal (g2(f2(x))) after processing the signal (f2(x)) by the second receiver.

[0220] By repeatedly executing the above-mentioned processes 421 and 422 through multiple source signals and iteratively updating the neural network model, a first adjustment model can be obtained.

[0221] Optionally, the above step 421 may include steps 421a and 421b.

[0222] 421a. The first device receives the eighth signal and the ninth signal from the second device.

[0223] 421b. The first device processes the ninth signal based on the neural network model involved in the training, and then processes the signal obtained by the first receiver through the processing of the ninth signal by the neural network to obtain the seventh signal.

[0224] Optionally, step 423 and step 424 may be included after step 422 .

[0225] 423. The first device sends a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0226] The neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0227] 424. The second device determines a third adjustment model based on the first adjustment model.

[0228] In the model training scheme provided in the embodiment of the present application, the result of the joint processing of the signal (f2(x)) encoded or modulated by the second transmitter by the first adjustment model and the first receiver is basically consistent with the signal (x) before encoding or modulation by the second transmitter or the processing result (g2(f2(x))) of the signal (f2(x)) by the second receiver. Because the second receiver and the second transmitter are usually jointly trained, the data processed by the second transmitter can be correctly received by the second receiver. If g1(t1(f2(x))) is as close as possible to x or g2(f2(x)), the first receiver can also correctly receive the signal (f2(x)) sent by the second transmitter. Therefore, the first device and the second device can communicate normally through this training process.

[0229] The above three model training schemes all include a scheme in which the first device sends a first adjustment model to the second device, and the second device determines a second adjustment model or a third adjustment model based on the first adjustment model.

[0230] Whether determining the second adjustment model based on the first adjustment model or determining the third adjustment model based on the first adjustment model, the second adjustment model or the third adjustment model can be an inverse neural network of the neural network of the first adjustment model. The process of the second device determining the second adjustment model or the third adjustment model can be understood by referring to the following process.

[0231] If the function of the first adjustment model is expressed as: y = t1 (x, W), then the function of the second adjustment model or the third adjustment model can be expressed as: x = t2 (y, W), t2 = t1 -1 , where W is the weight of the model. The adjustment model can be composed of multiple reversible functions, that is, t1(x)=t1 K (t1 K-1 ·(·(t11(x)))), Where K represents the number of reversible functions. Reversible functions include: reversible activation functions, such as Linear operations, such as t1(x)=Wx+b, t2(y)=t1 -1 (y) = W -1 (yb). The reversible neural network is shown in FIG5A , where v1(., θ1) represents a function v1 with parameter θ1, u1(., φ1) represents a function u1 with parameter φ1, v2(., θ2) represents a function v2 with parameter θ2, v2(., θ2) represents a function v2 with parameter θ2, u2(., φ2) represents a function u2 with parameter φ2, and t1 is z1=x1*v1(x2, θ 1)+u1(x2, φ1), z2=x2*v2(z1, θ2)+u2(z1, φ2), the corresponding t2 is x2=(z2-u2(z1, θ2)) / v2(z1, φ2), x1=(z1-u1(x2, θ1)) / v1(x2, φ1), among which only one of u1, v1, u2, v2 needs to have a parameter; as shown in Figure 5B, t1 is z1=x1+

[0232] u1(x2, φ1), z2=x2+u2(z1, φ2), the corresponding t2 is x2=z2-u2(z1, θ2), x1=z1-u1(x2, θ1); as shown in Figure 5C, t1 is z1=x1*v1(x2, θ1)+u1(x2, φ1), z2=x2, the corresponding t2 is x2=z2, x1=(z1-u1(x2, θ1)) / v1(x2, φ1).

[0233] Figures 5A to 5C illustrate the process of deriving the second or third adjustment model from the first adjustment model. This eliminates the need for the second device to train a second adjustment model for the second transmitter or a third adjustment model for the second receiver. The second or third adjustment model can be derived from the first adjustment model, thereby reducing computational complexity and the overhead of acquiring the second or third adjustment model.

[0234] In addition, after the above-mentioned adjustment model is trained, an association relationship between the information of the first transmitter, the information of the second receiver and the first adjustment model can be established, such as expressing the association relationship in the form of a first identification group, and the first identification group can be expressed as (Tx01, M01, Re02), for example. Among them, Tx01 represents the identification of the first transmitter, Re02 represents the identification of the second receiver, and M01 represents the identification of the first adjustment model. Similarly, an association relationship between the first receiver, the second transmitter and the first adjustment model can also be established, such as expressing the association relationship in the form of a second identification group, and the second identification group can be expressed as (Tx02, M02, Re01), for example. Among them, Tx02 represents the identification of the second transmitter, Re01 represents the identification of the first receiver, and M02 represents the identification of the first adjustment model corresponding to the first receiver and the second transmitter. Of course, the forms of each identifier in the first identification group and the second identification group listed here are only examples, and the specific form of the identifier is not limited in this application.

[0235] It should be noted that the signals appearing in this application can also be expressed as data, symbols or bits. The only difference is the way of expression. In essence, they are all data signals.

[0236] The above describes the model training scheme. After the model is trained, it can be used to adjust the model for communication. The following describes the communication process based on the adjustment model with reference to the accompanying figure.

[0237] Based on the communication system shown in FIG. 3A , referring to FIG. 6A , an embodiment of a communication method provided in an embodiment of the present application includes:

[0238] 601. The first device receives first information from the second device.

[0239] The first information may be an identifier of the first adjustment model or associated information that can be used to determine the first adjustment model.

[0240] 602. The first device determines a first adjustment model based on the first information, where the first adjustment model is associated with the information of the first transmitter and the information of the second receiver.

[0241] The information of the first transmitter may be an identifier of the first transmitter or other information that may indicate the first transmitter, and the information of the second receiver may be an identifier of the second receiver or other information that may indicate the second receiver.

[0242] The first adjustment model is associated with the information of the first transmitter and the information of the second receiver, which means that the first adjustment model can be determined (eg, found or calculated) through the information of the first transmitter and the information of the second receiver.

[0243] 603. The first device communicates with the second receiver of the second device based on the first transmitter and the first adjustment model.

[0244] Optionally, before the above step 601, step 601a may be further included.

[0245] 601a. ​​The first device sends information of the first transmitter to the second device.

[0246] The second device may determine the identifier of the first adjustment model based on the information of the first transmitter and the information of the second receiver used to communicate with the first transmitter, and then return the identifier of the first adjustment model to the first device.

[0247] Alternatively, the second device may return the information of the second receiver to the first device based on the information of the first transmitter and the information of the second receiver used to communicate with the first transmitter, and the first device may determine the identifier of the first adjustment model based on the information of the first transmitter and the information of the second receiver.

[0248] Both the first device and the second device may include a first identification group, and both the first device and the second device may determine the identification of the first adjustment model from the first identification group based on the identification of the first transmitter and the identification of the second receiver. The first identification group includes the identification of the first transmitter, the identification of the second receiver, and the identification of the first adjustment model.

[0249] The content of the first identification group can be understood by referring to the previous introduction, and will not be repeated here.

[0250] Optionally, the above step 603 may specifically include: the first device processes the signal to be transmitted based on the first transmitter, and then processes the signal processed by the first transmitter based on the first adjustment model; the first device sends the signal processed by the first adjustment model to the second device.

[0251] In this optional embodiment, after the first transmitter processes the signal to be transmitted (x), the first adjustment model processes the signal processed by the first transmitter (f1(x)), and then sends the signal processed by the first adjustment model (t1(f1(x))). Because the second device participates in the training of the first adjustment model, the signal processed by the first adjustment model can be correctly received by the second receiver, thereby realizing effective communication between the first transmitter and the second receiver.

[0252] It should be noted that the above-mentioned first transmitter first processes the signal to be sent (x), and the first adjustment model then processes the signal processed by the first transmitter (f1(x)). This is only a possible implementation method. Other processing orders can also be adopted, such as: the first adjustment model first processes the signal to be sent, and then the first transmitter processes the signal processed by the first adjustment model. This application does not limit this.

[0253] In the embodiments provided herein, a first device may determine a first adjustment model and then communicate with a second receiver of a second device based on the first transmitter and the first adjustment model. In this manner, even if the first transmitter and the second receiver are not jointly trained, communication between the first and second devices can be achieved using the first adjustment model.

[0254] Based on the communication system shown in FIG. 3B , referring to FIG. 6B , another embodiment of the communication method provided in the embodiment of the present application includes:

[0255] 611. The first device receives first information from the second device.

[0256] The first information may be an identifier of the first adjustment model or associated information that can be used to determine the first adjustment model.

[0257] 612. The first device determines a first adjustment model based on the first information, and the first adjustment model is associated with the information of the first receiver and the information of the second transmitter.

[0258] The information of the first receiver may be an identifier of the first receiver or other information that can indicate the first receiver. The information of the second transmitter may be an identifier of the second transmitter or other information that can indicate the second transmitter.

[0259] The first adjustment model being associated with the information of the first receiver and the information of the second transmitter means that the first adjustment model can be determined (eg, obtained by searching or calculating) through the information of the first receiver and the information of the second transmitter.

[0260] 613. The first device communicates with the second transmitter of the second device based on the first receiver and the first adjustment model.

[0261] In the embodiments provided herein, a first device can determine a first adjustment model and then communicate with a second transmitter of a second device based on the first receiver and the first adjustment model. This allows communication between the first and second devices using the first adjustment model, even if the first receiver and second transmitter are from different manufacturers and have not undergone joint training.

[0262] Optionally, before the above step 611, step 611a may be further included.

[0263] 611a. The first device sends the information of the first receiver to the second device.

[0264] The second device may determine the identifier of the first adjustment model based on the information of the first receiver and the information of the second transmitter used to communicate with the first receiver, and then return the identifier of the first adjustment model to the first device.

[0265] Alternatively, the second device may return the information of the second transmitter to the first device based on the information of the first receiver and the information of the second transmitter used to communicate with the first receiver, and the first device may determine the identifier of the first adjustment model based on the information of the second transmitter and the information of the first receiver.

[0266] The first device and the second device may both include a second identification group, the second identification group including an identification of the second transmitter, an identification of the first receiver, and an identification of the first adjustment model. The first device and the second device may both determine the identification of the first adjustment model from the second identification group based on the identification of the second transmitter and the identification of the first receiver.

[0267] The content of the second identification group can be understood by referring to the previous introduction, and will not be repeated here.

[0268] Optionally, the above step 613 may specifically include: the first device processes the signal from the second transmitter of the second device based on the first adjustment model; and the first device processes the signal processed by the first adjustment model based on the first receiver.

[0269] In this possible embodiment, the first adjustment model processes the signal f2(x) from the second transmitter of the second device to obtain t1(f2(x)), and the first receiver processes t1(f2(x)) to obtain g1(t1(f2(x))). Because the second device participated in the training of the first adjustment model, the first receiver can correctly receive the signal processed by the first adjustment model, thereby achieving effective communication between the second transmitter and the first receiver.

[0270] It should be noted that the above-mentioned first adjustment model first processes the signal f2(x) sent by the second transmitter, and then the first receiver processes the signal t1(f2(x)) processed by the first adjustment model. This is only a possible implementation method. Other processing orders can also be adopted, such as: the first receiver first processes the signal sent by the second transmitter, and then the first adjustment model processes the signal processed by the first receiver. This application does not limit this.

[0271] It should be noted that during the communication process, the model training process described in FIG. 4A to FIG. 4C may also be performed, so that the model can be continuously updated and adjusted to improve the communication quality.

[0272] The communication method introduced above introduces a first adjustment model. In fact, there can be multiple first adjustment models, and multiple first adjustment models can be connected in parallel or in series. They are introduced below respectively.

[0273] 1. A plurality of first adjustment models are connected in parallel, and each of the first adjustment models connected in parallel is used for a first transmitter to communicate with a different second receiver.

[0274] Please refer to FIG7A . In FIG7A , there are three adjustment models corresponding to the transmitter 1 , which are represented by adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 , respectively. The adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 are connected in parallel.

[0275] Among them, the adjustment model t1-1 is used for communication between transmitter 1 and receiver 2, the adjustment model t1-2 is used for communication between transmitter 1 and receiver 3, and the adjustment model t1-3 is used for communication between transmitter 1 and receiver 4. It should be noted that the embodiment of the present application can also realize communication between other different transmitters and receivers by combining existing adjustment models t1-1, t1-2, and t1-3, such as: the inverse neural network of the adjustment model t1-1 and the adjustment model t1-2 are connected in series to realize communication between transmitter 2 and receiver 3.

[0276] 2. A plurality of first adjustment models are connected in parallel, and each of the first adjustment models in parallel is used for the first receiver to communicate with a different second transmitter.

[0277] Please refer to FIG. 7B . In FIG. 7B , there are three adjustment models corresponding to the receiver 1 , which are represented by adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 , respectively. The adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 are connected in parallel.

[0278] The adjustment model t1 - 1 is used for communication between receiver 1 and transmitter 2 , the adjustment model t1 - 2 is used for communication between receiver 1 and transmitter 3 , and the adjustment model t1 - 3 is used for communication between receiver 1 and transmitter 4 .

[0279] The multiple parallel adjustment models provided in the embodiments of the present application are used for a transmitter to communicate with different receivers, or for a receiver to communicate with different transmitters, which can improve the flexibility of communication.

[0280] 3. A plurality of first adjustment models are connected in series, and at least two directly connected first adjustment models in series are used for a first transmitter to communicate with different second receivers.

[0281] Please refer to Figure 7C. There are three adjustment models corresponding to the transmitter 1 in Figure 7C, which are represented by adjustment model t1-1, adjustment model t1-2 and adjustment model t1-3 respectively. The adjustment model t1-1, adjustment model t1-2 and adjustment model t1-3 are connected in series.

[0282] Among them, the adjustment model t1-1 is used for communication between transmitter 1 and receiver 2, the adjustment model t1-1 and the adjustment model t1-2 are connected in series and used for communication between transmitter 1 and receiver 3, and the adjustment model t1-1, the adjustment model t1-2 and the adjustment model t1-3 are connected in series and used for communication between transmitter 1 and receiver 4. It should be noted that the embodiment of the present application can also realize communication between other different transmitters and receivers by combining the existing adjustment models t1-1, t1-2, and t1-3, such as: the adjustment model t1-2 and the adjustment model t1-3 are connected in series to realize communication between transmitter 2 and receiver 4.

[0283] 4. A plurality of first adjustment models are connected in series, and at least two directly connected first adjustment models in series are used for a first receiver to communicate with different second transmitters.

[0284] Please refer to FIG. 7D . In FIG. 7D , there are three adjustment models corresponding to the receiver 1 , which are represented by adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 , respectively. The adjustment model t1 - 1 , adjustment model t1 - 2 and adjustment model t1 - 3 are connected in series.

[0285] Among them, the adjustment model t1-1 is used for communication between receiver 1 and transmitter 2, the adjustment model t1-1 and the adjustment model t1-2 are connected in series and used for communication between receiver 1 and transmitter 3, and the adjustment model t1-1, the adjustment model t1-2 and the adjustment model t1-3 are connected in series and used for communication between receiver 1 and transmitter 4.

[0286] The multiple series-connected adjustment models provided in the embodiments of the present application are used for a transmitter to communicate with different receivers, or for a receiver to communicate with different transmitters, which can improve the flexibility of communication.

[0287] It should be noted that in the embodiments of the present application, in the parallel and series adjustment model schemes introduced in Figures 7A to 7D above, the trained adjustment models in any of the structures can be migrated to other devices for use, such as: the adjustment model t1-2 in Figure 7C above can be used for communication between transmitter 2 and receiver 3, and the adjustment model t1-3 can be used for communication between transmitter 3 and receiver 4.

[0288] The above introduces the model training method and the communication method based on the adjustment model. The following introduces the communication device.

[0289] Please refer to FIG8 , an embodiment of a communication device 800 provided in an embodiment of the present application includes: a processing module 801 and a transceiver module 802; wherein,

[0290] Processing module 801 is used to determine a first adjustment model based on first information from the second device, the first adjustment model is associated with information of the first transmitter and information of the second receiver, or the first adjustment model is associated with information of the first receiver and information of the second transmitter, wherein the first transmitter and the first receiver correspond to the first device, and the second receiver and the second transmitter correspond to the second device.

[0291] The processing module 801 is further configured to communicate with the second receiver of the second device based on the first transmitter and the first adjustment model; or the first device communicates with the second transmitter of the second device based on the first receiver and the first adjustment model.

[0292] Optionally, the processing module 801 is specifically configured to process the signal to be transmitted based on the first transmitter, and then process the signal processed by the first transmitter based on the first adjustment model.

[0293] The transceiver module 802 is configured to send the signal processed by the first adjustment model to the second device.

[0294] Optionally, the processing module 801 is specifically configured to process a signal from a second transmitter of the second device based on the first adjustment model, and process the signal processed by the first adjustment model based on the first receiver.

[0295] Optionally, the first information includes an identifier of the first adjustment model, and the identifier of the first adjustment model is used to determine the first adjustment model.

[0296] Optionally, the first information includes an identifier of the second receiver or an identifier of the second transmitter; the identifier of the second receiver or the identifier of the second transmitter is used to determine an identifier of the first adjustment model.

[0297] Optionally, the identifier of the first adjustment model is determined by the first device from a first identifier group based on the identifier of the first transmitter and the identifier of the second receiver, and the first identifier group includes the identifier of the first transmitter, the identifier of the second receiver, and the identifier of the first adjustment model; or, the identifier of the first adjustment model is determined by the first device from a second identifier group based on the identifier of the first receiver and the identifier of the second transmitter, and the second identifier group includes the identifier of the second transmitter, the identifier of the first receiver, and the identifier of the first adjustment model.

[0298] Optionally, the processing module 801 is also used to train the neural network model according to the first training objective to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal, the first signal is a signal obtained by the first transmitter processing the third signal and then processing it by the neural network model participating in the training, the second signal is a signal obtained by the second transmitter processing the third signal, and the third signal is the source signal.

[0299] Optionally, the processing module 801 is also used to train the neural network model according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate that the loss function of the neural network model participating in the training is reduced according to the fourth signal, the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal.

[0300] Optionally, the processing module 801 is also used to train the neural network model according to a third training objective to obtain a first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal, the seventh signal is a signal obtained by the first receiver processing the ninth signal with the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal.

[0301] Optionally, the transceiver module 802 is further configured to send the first adjustment model to the second apparatus, where the first adjustment model is used by the second apparatus to determine a second adjustment model for the second transmitter.

[0302] Optionally, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0303] Optionally, the transceiver module 802 is further configured to send a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0304] Optionally, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0305] Optionally, if there are multiple first adjustment models, and the multiple first adjustment models are connected in parallel; each parallel first adjustment model is used for the first transmitter to communicate with a different second receiver; or, each parallel first adjustment model is used for the first receiver to communicate with a different second transmitter.

[0306] Optionally, if there are multiple first adjustment models, and the multiple first adjustment models are connected in series; at least two directly connected first adjustment models in series are used for the first transmitter to communicate with different second receivers; or, at least two directly connected first adjustment models in series are used for the first receiver to communicate with different second transmitters.

[0307] When the communication device 800 is used in a model training process:

[0308] Processing module 801 is used to obtain signals for model training, and the signals for model training include a first signal and a second signal; wherein the first signal is a signal obtained by processing a third signal by a first transmitter and then processed by a neural network model participating in the training, and the second signal is a signal obtained by processing the third signal by a second transmitter, and the third signal is a source signal; according to a first training objective, the neural network model is trained to obtain a first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal.

[0309] Optionally, the transceiver module 802 is configured to send a first adjustment model to a second device, where the first adjustment model is used by the second device to determine a second adjustment model for a second transmitter.

[0310] Optionally, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0311] When the communication device 800 is used in another model training process:

[0312] Processing module 801 is used to obtain signals for model training, and the signals for model training include a fourth signal and a fifth signal; wherein the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal; the neural network model is trained according to the second training objective to obtain a first adjustment model, wherein the second training objective is used to indicate how to reduce the loss function of the neural network model participating in the training according to the fourth signal.

[0313] Optionally, the transceiver module 802 is configured to send a first adjustment model to a second device, where the first adjustment model is used by the second device to determine a second adjustment model for a second transmitter.

[0314] Optionally, the neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0315] When the communication device 800 is used in another model training process:

[0316] Processing module 801 is used to obtain signals for model training, and the signals for model training include a seventh signal and an eighth signal; wherein the seventh signal is a signal obtained by the first receiver processing the ninth signal of the neural network model participating in the training, and then processing the ninth signal; the ninth signal is obtained by the second transmitter processing the source signal; and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal; the neural network model is trained according to the third training objective to obtain a first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal.

[0317] Optionally, the transceiver module 802 is configured to send a first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

[0318] Optionally, the neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

[0319] The communication device introduced in the embodiment of the present application can be understood by referring to the functions of the first device in the previous Figures 3A to 7D. The communication device can be a terminal device or a network device.

[0320] Figure 9 shows a simplified schematic diagram of the structure of a terminal device, where the communication device is a terminal device. For ease of understanding and illustration, Figure 9 uses a mobile phone as an example of a terminal device. As shown in Figure 9, the terminal device includes a processor, memory, radio frequency circuitry, an antenna, and input / output devices.

[0321] The processor is mainly used to process communication protocols and communication data, as well as control terminal devices, execute software programs, process software program data, etc.

[0322] Memory is mainly used to store software programs and data.

[0323] Radio frequency circuits are mainly used for conversion between baseband signals and radio frequency signals and for processing radio frequency signals.

[0324] Antennas are mainly used to send and receive radio frequency signals in the form of electromagnetic waves.

[0325] Input and output devices, such as touch screens, display screens, keyboards, etc., are mainly used to receive data input by users and output data to users.

[0326] It should be noted that some types of terminal devices may not have input and output devices.

[0327] When data needs to be sent, the processor performs baseband processing on the data to be transmitted and outputs the baseband signal to the RF circuit. The RF circuit then performs RF processing on the baseband signal and transmits it via the antenna in the form of electromagnetic waves. When data is sent to a terminal device, the RF circuit receives the RF signal via the antenna, converts it into a baseband signal, and outputs the baseband signal to the processor, which converts the baseband signal into data and processes it.

[0328] For ease of explanation, Figure 9 shows only one memory and processor. In actual terminal device products, one or more processors and one or more memories may exist. Memory may also be referred to as a storage medium or storage device. The memory may be provided independently of the processor or integrated with the processor, and this embodiment of the application does not impose any restrictions on this.

[0329] In the embodiment of the present application, the antenna and radio frequency circuit with transceiver functions can be regarded as the transceiver unit of the terminal device, and the processor with processing function can be regarded as the processing unit of the terminal device.

[0330] As shown in Figure 9, the terminal device includes a transceiver unit 910 and a processing unit 920. The transceiver unit may also be called a transceiver, a transceiver, a transceiver device, etc. The processing unit may also be called a processor, a processing board, a processing module, a processing device, etc.

[0331] Alternatively, the device in the transceiver unit 910 that implements the receiving function may be considered a receiving unit, and the device in the transceiver unit 910 that implements the transmitting function may be considered a transmitting unit. That is, the transceiver unit 910 includes a receiving unit and a transmitting unit. The transceiver unit may also be sometimes referred to as a transceiver, a transceiver, or a transceiver circuit. The receiving unit may also be sometimes referred to as a receiver, a receiver, or a receiving circuit. The transmitting unit may also be sometimes referred to as a transmitter, a transmitter, or a transmitting circuit.

[0332] It should be understood that the transceiver unit 910 is used to perform the sending operation and the receiving operation of the first device in the above method embodiment, and the processing unit 920 is used to perform other operations except the sending and receiving operation on the first device in the above method embodiment.

[0333] When the terminal device is a chip, the chip includes a transceiver unit and a processing unit. The transceiver unit may be an input / output circuit or a communication interface; the processing unit may be a processor, microprocessor, integrated circuit, or logic circuit integrated on the chip.

[0334] When the communication device is a network device, please refer to Figure 10, which is a schematic diagram of the structure of a network device 1000 provided in an embodiment of the present application. The network device 1000 can be applied to the system shown in Figure 1 or Figure 2. For example, the network device 1000 can be a network device in the system of Figure 1 or Figure 2, and is used to perform the functions of the first device in the above method embodiment. It should be understood that the following is only an example, and in future communication systems, network devices may have other forms and structures.

[0335] For example, in a 5G communication system, the network device 1000 may include a CU, a DU, and an AAU. Compared to a network device in an LTE communication system, which consists of one or more radio frequency units, such as an RRU, and one or more BBUs:

[0336] The non-real-time portion of the original BBU will be separated and redefined as a CU, responsible for handling non-real-time protocols and services. Some of the BBU's physical layer processing functions will be merged with the original RRU and passive antennas into the AAU. The remaining BBU functions will be redefined as a DU, responsible for handling physical layer protocols and real-time services. In short, the CU and DU are differentiated by the real-time nature of their processing, and the AAU is a combination of the RRU and antenna.

[0337] The CU, DU, and AAU can be deployed separately or together, resulting in a variety of network deployment configurations. One possible deployment configuration, shown in Figure 10, is consistent with traditional 4G network equipment, with the CU and DU deployed on shared hardware. It should be understood that Figure 10 is merely an example and does not limit the scope of protection of this application. For example, the deployment configuration may also include the DU being deployed in the BBU room, the CU being deployed centrally, or the DU being deployed centrally, with the CU being centralized at a higher level.

[0338] The AAU 1100 can implement transceiver functions and is called a transceiver unit 1100, corresponding to the transceiver module 802 in Figure 8. Optionally, the transceiver unit 1100 can also be called a transceiver, a transceiver circuit, or a transceiver, and can include at least one antenna 1101 and a radio frequency unit 1102. Optionally, the transceiver unit 1100 can include a receiving unit and a transmitting unit. The receiving unit can correspond to a receiver (or a receiver, a receiving circuit), and the transmitting unit can correspond to a transmitter (or a transmitter, a transmitting circuit). The CU and DU 1200 can implement internal processing functions and is called a processing unit 1200, corresponding to the processing module 801 in Figure 8. Optionally, the processing unit 1200 can control network devices and can be called a controller. The AAU, CU, and DU can be physically located together or physically separated.

[0339] In addition, the network equipment is not limited to the form shown in Figure 10, but can also be other forms: for example: including a BBU and an adaptive radio unit (ARU), or including a BBU and an AAU; it can also be customer premises equipment (CPE), or it can be other forms, which are not limited in this application.

[0340] In one example, the processing unit 1200 can be composed of one or more single boards, and multiple single boards can jointly support a wireless access network with a single access standard (such as an LTE network), or can respectively support wireless access networks with different access standards (such as an LTE network, a 5G network, a future network or other networks). The CU and DU1200 also include a memory 1201 and a processor 1202. The memory 1201 is used to store necessary instructions and data. The processor 1202 is used to control the network device to perform necessary actions, such as controlling the network device to execute the operation process of the network device in the above method embodiment. The memory 1201 and the processor 1202 can serve one or more single boards. That is, a memory and a processor can be set separately on each single board. Alternatively, multiple single boards can share the same memory and processor. In addition, necessary circuits can also be set on each single board.

[0341] It should be understood that the network device 1000 shown in Figure 10 is capable of implementing the first device functions involved in the method embodiments of Figures 4A to 7D. The operations and / or functions of the various units in the network device 1000 are respectively for implementing the corresponding processes performed by the first device in the method embodiments of the present application. To avoid repetition, detailed descriptions are appropriately omitted here. The structure of the network device illustrated in Figure 10 is only one possible form and should not constitute any limitation on the embodiments of the present application. The present application does not exclude the possibility of other forms of network device structures that may appear in the future.

[0342] The CU and DU 1200 can be used to perform the actions described in the previous method embodiments, implemented within the first device. The AAU 1100 can be used to perform the actions described in the previous method embodiments, where the first device sends data to or receives data from the second device. For details, please refer to the description in the previous method embodiments and will not be repeated here.

[0343] An embodiment of the present application also provides a computer program product including instructions, which, when executed on a computer, enables the computer to execute the method of the embodiment shown in FIG. 4A to FIG. 7D above.

[0344] An embodiment of the present application further provides a computer-readable storage medium comprising computer instructions. When the computer instructions are executed on a computer, the computer executes the method of the embodiment shown in FIG. 4A to FIG. 7D .

[0345] An embodiment of the present application also provides a chip device, including a processor, which is connected to a memory and calls a program stored in the memory so that the processor executes the method of the embodiment shown in Figures 4A to 7D above.

[0346] The processor mentioned in any of the above may be a general-purpose central processing unit, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the method of the embodiment shown in Figures 4A to 4C. The memory mentioned in any of the above may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM), etc.

[0347] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0348] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0349] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0350] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0351] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the part that essentially contributes to the technical solution of the present application or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, ROM, RAM, a magnetic disk, or an optical disk.

[0352] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A communication method, characterized in that: include: The first device determines a first adjustment model based on first information from the second device, wherein the first adjustment model is associated with information of a first transmitter and information of a second receiver, or the first adjustment model is associated with information of the first receiver and information of the second transmitter, wherein the first transmitter and the first receiver correspond to the first device, and the second receiver and the second transmitter correspond to the second device; The first device communicates with a second receiver of the second device based on the first transmitter and the first adjustment model; or, The first device communicates with a second transmitter of the second device based on the first receiver and the first adjustment model.

2. The method according to claim 1, characterized in that The first device communicates with a second receiver of the second device based on the first transmitter and the first adjustment model, including: The first device processes the signal to be transmitted based on the first transmitter, and then processes the signal processed by the first transmitter based on the first adjustment model; The first device sends the signal processed by the first adjustment model to the second device.

3. The method according to claim 1 or 2, characterized in that The first device communicates with a second transmitter of the second device based on the first receiver and the first adjustment model, including: The first device processes a signal from a second transmitter of the second device based on the first adjustment model; The first device processes the signal processed by the first adjustment model based on the first receiver.

4. The method according to any one of claims 1 to 3, characterized in that The first information includes an identifier of the first adjustment model, and the identifier of the first adjustment model is used to determine the first adjustment model.

5. The method according to any one of claims 1 to 3, characterized in that The first information includes an identifier of the second receiver or an identifier of the second transmitter; The identifier of the second receiver or the identifier of the second transmitter is used to determine the identifier of the first adjustment model.

6. The method according to claim 5, characterized in that The identifier of the first adjustment model is determined by the first apparatus from a first identifier group according to the identifier of the first transmitter and the identifier of the second receiver, the first identifier group including the identifier of the first transmitter, the identifier of the second receiver, and the identifier of the first adjustment model; or The identifier of the first adjustment model is determined by the first device from a second identifier group based on the identifier of the first receiver and the identifier of the second transmitter, and the second identifier group includes the identifier of the second transmitter, the identifier of the first receiver, and the identifier of the first adjustment model.

7. The communication method according to any one of claims 1 to 6, characterized in that: Before the first device determines the first adjustment model according to the first information from the second device, the method further includes: The first device trains a neural network model according to a first training objective to obtain the first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal, the first signal is a signal obtained by the first transmitter processing the third signal and then processing it by the neural network model participating in the training, the second signal is a signal obtained by the second transmitter processing the third signal, and the third signal is a source signal.

8. The communication method according to any one of claims 1 to 6, characterized in that: Before the first device determines the first adjustment model according to the first information from the second device, the method further includes: The first device trains a neural network model according to a second training objective to obtain the first adjustment model, wherein the second training objective is used to indicate reducing the loss function of the neural network model participating in the training according to a fourth signal, the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal.

9. The communication method according to any one of claims 1 to 6, characterized in that: Before the first device determines the first adjustment model according to the first information from the second device, the method further includes: The first device trains a neural network model according to a third training objective to obtain the first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal, the seventh signal is a signal obtained by the first receiver processing the ninth signal by the neural network model participating in the training, the ninth signal is obtained by the second transmitter processing the source signal, and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal.

10. The communication method according to claim 7 or 8, characterized in that: The method further comprises: The first device sends the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

11. The method according to claim 10, characterized in that The neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

12. The communication method according to claim 9, wherein: The method further comprises: The first device sends the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

13. The communication method according to claim 12, wherein: The neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

14. The communication method according to any one of claims 1 to 13, characterized in that: If there are multiple first adjustment models, and the multiple first adjustment models are connected in parallel; Each first adjustment model connected in parallel is used for the first transmitter to communicate with a different second receiver; or, Each first adjustment model in parallel is used for the first receiver to communicate with a different second transmitter.

15. The communication method according to claims 1 to 13, characterized in that: If there are multiple first adjustment models, and the multiple first adjustment models are connected in series; At least two directly connected first adjustment models connected in series are used for the first transmitter to communicate with different second receivers; or, At least two directly connected first adjustment patterns are connected in series for the first receiver to communicate with different second transmitters.

16. A method for model training, characterized in that: include: The first device obtains a signal for model training, wherein the signal for model training includes a first signal and a second signal; wherein the first signal is a signal obtained by the first transmitter processing a third signal and then processed by the neural network model participating in the training; the second signal is a signal obtained by the second transmitter processing the third signal; and the third signal is a source signal; The first device trains a neural network model according to a first training objective to obtain the first adjustment model, wherein the first training objective is used to indicate reducing the error between the first signal and the second signal.

17. The method according to claim 16, characterized in that The method further comprises: The first device sends the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

18. The method according to claim 17, characterized in that The neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

19. A method for model training, characterized in that: include: The first device obtains a signal for model training, the signal for model training including a fourth signal and a fifth signal; wherein the fourth signal is a signal obtained by the second receiver processing the fifth signal, the fifth signal is a signal obtained by the first transmitter processing the sixth signal and then processed by the neural network model participating in the training, and the sixth signal is a source signal; The first device trains a neural network model according to a second training objective to obtain the first adjustment model, wherein the second training objective is used to indicate reducing a loss function of the neural network model participating in the training according to a fourth signal.

20. The method according to claim 19, characterized in that The method further comprises: The first device sends the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a second adjustment model for the second transmitter.

21. The method according to claim 20, characterized in that The neural network of the second adjustment model is an inverse neural network of the neural network of the first adjustment model.

22. A method for model training, characterized in that: include: The first device obtains a signal for model training, the signal for model training including a seventh signal and an eighth signal; wherein the seventh signal is a signal obtained by the first receiver processing the ninth signal by the neural network model participating in the training, and then processing the ninth signal; the ninth signal is obtained by the second transmitter processing the source signal; and the eighth signal is obtained by the source signal or the second receiver processing the ninth signal; The first device trains a neural network model according to a third training objective to obtain the first adjustment model, wherein the third training objective is used to indicate reducing the error between the seventh signal and the eighth signal.

23. The method according to claim 22, characterized in that The method further comprises: The first device sends the first adjustment model to the second device, where the first adjustment model is used by the second device to determine a third adjustment model for the second receiver.

24. The method according to claim 23, wherein The neural network of the third adjustment model is an inverse neural network of the neural network of the first adjustment model.

25. A communication device, characterized in that: include: Transceiver module and processing module, The transceiver module is used to perform the sending step or the receiving step in the method according to any one of claims 1 to 24; The processing module is used to execute the steps except the sending step and the receiving step in the method according to any one of claims 1 to 24.

26. A communication device, characterized in that: comprising at least one processor coupled to a memory; The memory is used to store programs or instructions; The at least one processor is configured to execute the program or instructions so as to enable the apparatus to implement the method according to any one of claims 1 to 24.

27. A computer program product comprising program instructions, characterized in that When the program instructions are executed on a computer, the computer is caused to perform the method according to any one of claims 1 to 24.

28. A computer-readable storage medium, characterized in that The computer-readable storage medium stores program instructions, and when the program instructions are executed, the method according to any one of claims 1 to 24 is executed.

Images