Digital predistortion communication device and implementation method based on GRU network

Abstract

The present application provides a digital pre-distortion communication device and implementation method based on a GRU network, the method comprising: obtaining original data, and determining distorted digital signal data based on the original data. Preprocessing the data. Establishing a power amplifier model based on the preprocessed original data and the preprocessed digital signal data. Pre-training the power amplifier model to obtain a pre-trained power amplifier model. Establishing a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital pre-distortion model of the communication device. Training the digital pre-distortion model on the basis of fixed parameters of the pre-trained power amplifier model to obtain a trained digital pre-distortion model. Deploying the trained digital pre-distortion model to the baseband chip of the communication device, and offsetting the digital pre-distortion in the communication process of the communication device by the trained digital pre-distortion model. The method provided by the present application combines digital pre-distortion communication technology with actual implementation.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a digital predistortion communication device and implementation method based on a GRU network. Background Art

[0002] The advancement of communications technology, especially next-generation wireless systems like 6G and Wi-Fi 7, has placed higher demands on the linearization technology of broadband power amplifiers. Traditional DPD technology can no longer meet the growing demands for communication capacity and data accuracy. The development of deep learning technology has brought new breakthroughs in DPD technology. However, existing deep learning networks have only been studied from a network design perspective, without integrating them into actual communication systems. This makes it difficult to implement large models using existing technologies from the perspective of algorithm implementation, including training data acquisition, data preprocessing, model training, and model deployment.

[0003] Therefore, how to combine the digital predistortion communication technology with the actual communication system and put it into practical use has become a technical problem that those skilled in the art urgently need to solve. Summary of the Invention

[0004] Based on the above problems, the present application provides a digital predistortion communication device and implementation method based on a GRU network, the method comprising:

[0005] Acquiring original data, and determining distorted digital signal data based on the original data;

[0006] performing data preprocessing on the original data to obtain preprocessed original data, and performing preprocessing on the distorted digital signal data to obtain preprocessed digital signal data;

[0007] Establishing a power amplifier model based on the preprocessed raw data and the preprocessed digital signal data;

[0008] Pre-training the power amplifier model to obtain a pre-trained power amplifier model;

[0009] Establishing a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital predistortion model of the communication device;

[0010] Based on the radio frequency link model, the digital predistortion model is trained on the basis of fixing the parameters of the pre-trained power amplifier model to obtain a trained digital predistortion model;

[0011] The trained digital pre-distortion model is deployed in a baseband chip of a communication device, and the digital pre-distortion during the communication process of the communication device is offset by the trained digital pre-distortion model.

[0012] In a possible implementation, the obtaining of original data and determining distorted digital signal data based on the original data includes:

[0013] Generate an I / Q signal of the communication device baseband by the communication device, where I and Q represent two complex components in the communication process of the communication device, and the I / Q signal represents original data X, which is data to be sent in the communication device baseband;

[0014] Passing the original data X through a power amplifier of a communication device to obtain distorted original data;

[0015] The distorted original data is processed through a receiving link of a communication device to obtain distorted digital signal data Y.

[0016] In a possible implementation, preprocessing the original data to obtain preprocessed original data, and preprocessing the distorted digital signal data to obtain preprocessed digital signal data include:

[0017] Split the time series data of the original data X into frame data of the original data , and dividing the time series data of the distorted digital signal data Y into frame data of the distorted digital signal data , and The number of is T, and The step size is S, and >S;

[0018] Frame data based on raw data Sure Overlapping samples of the original data , t represents T The tth one in the t-th one, and the frame data based on the distorted digital signal data Sure Overlapping samples of distorted digital signal data , t represents T The tth one in ;

[0019] Determine amplitude data based on the frame data of the original data , phase sinusoidal component and phase cosine component .

[0020] In a possible implementation, establishing a power amplifier model based on the preprocessed raw data and the preprocessed digital signal data includes:

[0021] Will and As the label data of the input and output of the power amplifier model, Establishing a power amplifier model as a loss function of the power amplifier model;

[0022] = ;

[0023] in, represents the loss function of the power amplifier model, and The number of is T, The power amplifier model corresponds to the t-th input and the t-th input The output value obtained is express Overlapping samples of distorted digital signal data.

[0024] In a possible implementation, pre-training the power amplifier model to obtain a pre-trained power amplifier model includes:

[0025] Will in 、 and 、 and The data constitutes the input vector, and express The two complex components of ;

[0026] The input vector is used as the power amplifier model Input, As a power amplifier model The label for the power amplifier model Perform pre-training to obtain a pre-trained power amplifier model .

[0027] In a possible implementation, establishing a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital predistortion model of the communication device includes:

[0028] The RF link model It is expressed by the following formula:

[0029] + ;

[0030] in, represents the RF link model, represents the pre-trained power amplifier model, represents the digital predistortion model, The loss function is ;

[0031] The RF link model The output label data is , represents the gain of the desired output of the communication system, yes and Obtained by transformation;

[0032] described The loss function is It is expressed by the following formula:

[0033] ;

[0034] in, express The loss function is Indicates the tth The output data, Represents the output label data corresponding to the t-th output data.

[0035] In a possible implementation, the training of the digital predistortion model based on the radio frequency link model and on the basis of fixing the parameters of the pretrained power amplifier model to obtain a trained digital predistortion model includes:

[0036] Will in 、 and 、 and The data constitutes the input vector, and express The two complex components of ;

[0037] Will and Transformed into , and As Output label data of

[0038] It is transformed by the following formula:

[0039] ;

[0040] in, and express The two complex components of and express The two complex components of , t represents the t-th input;

[0041] train Model, in training When modeling, first fix The parameters of the model are only optimized The parameters of the model are trained to obtain a trained digital predistortion model.

[0042] In a possible implementation, the digital predistortion model of the communication device includes a GRU layer, a hidden layer, an activation layer, a connection layer, and an output layer;

[0043] In the GRU layer, input data 、 、 and is represented as a one-dimensional column vector , the number of input neurons M of the GRU layer is 5*T, and the hidden layer of the GRU layer is set to 5*T;

[0044] The hidden layer is a fully connected layer, and the input and output are set to the same parameters as the hidden layer of the GRU layer;

[0045] The activation layer uses the relu activation function, and the data after the hidden layer is ;

[0046] The connection layer will and The data is spliced ​​and input to the output layer. The splicing result It is expressed as follows:

[0047] );

[0048] The output layer is a fully connected layer, and the input data of the output layer is , the output data is represented as components and quantity .

[0049] The present application also provides a digital pre-distortion communication device based on a GRU network, wherein the baseband chip of the communication device is equipped with a trained digital pre-distortion model, and the digital pre-distortion during the communication process of the communication device is offset by the trained digital pre-distortion model.

[0050] The present application also provides an electronic device, comprising a processor and a memory:

[0051] The memory is used to store a computer program and transmit the computer program to the processor;

[0052] The processor is configured to execute the steps of the above-mentioned GRU network-based digital predistortion communication method according to the instructions in the computer program.

[0053] The present application also provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed by an electronic device, it implements the steps of the above-mentioned digital predistortion communication method based on the GRU network.

[0054] Compared with the prior art, this application has the following beneficial effects:

[0055] This paper provides a solution for implementing a digital pre-distortion (DPD) algorithm based on an end-to-end deep learning network in real-world communication devices. This solution includes data acquisition and preprocessing, power amplifier modeling, DPD modeling, model pre-training, and deployment. It also provides a DPD algorithm model framework based on a GRU network. This framework implements DPD with smaller network parameters, facilitating subsequent model deployment and laying the foundation for building DPD models based on deep learning networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0057] Figure 1 A flowchart of a digital predistortion communication method based on a GRU network provided in an embodiment of the present application;

[0058] Figure 2 A system principle block diagram of a digital predistortion communication device and implementation method based on a GRU network provided in an embodiment of the present application;

[0059] Figure 3 A network structure diagram of a digital predistortion model provided in an embodiment of the present application;

[0060] Figure 4 A schematic structural diagram of a digital predistortion communication device based on a GRU network provided in an embodiment of the present application. DETAILED DESCRIPTION

[0061] As previously mentioned, the advancement of communications technology, especially the development of next-generation wireless systems such as 6G and Wi-Fi 7, has placed higher demands on the linearization technology of broadband power amplifiers. Traditional DPD technology can no longer meet the growing demands for communication capacity and data accuracy. The development of deep learning technology has brought new breakthroughs in DPD technology. However, existing deep learning networks have only been studied from the perspective of network design, without integrating them into actual communication systems. This makes it difficult to implement large models using existing technologies from the perspective of algorithm implementation, including training data collection, data preprocessing, model training, and model deployment.

[0062] To address the above issues, the present invention proposes a GRU-based digital pre-distortion communication device and implementation method. This invention provides an end-to-end deep learning network-based digital pre-distortion solution for broadband power amplifier modeling, digital pre-distortion model modeling, model pre-training, and deployment. This solution includes data acquisition and pre-processing, power amplifier modeling, digital pre-distortion model modeling, model pre-training and deployment, and provides a GRU-based digital pre-distortion algorithm model framework. This framework implements digital pre-distortion functionality with smaller network parameters, facilitating subsequent model deployment and laying the foundation for building digital pre-distortion models based on deep learning networks.

[0063] In order to help those skilled in the art better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0064] It is understandable that the method provided in the present application can be applied to a processing device that can obtain original data and determine distorted digital signal data based on the original data, such as a terminal device or server that can obtain original data and determine distorted digital signal data based on the original data. The method provided in the present application can be executed independently by a terminal device or a server, or can be applied to a network scenario in which a terminal device and a server communicate, and can be executed in cooperation with the terminal device and the server. Among them, the terminal device can be a computer, a mobile phone, or other device. The server can be understood as an application server or a web server. In actual deployment, the server can be a standalone server or a cluster server.

[0065] Figure 1 This is a flowchart of a digital predistortion communication method based on a GRU network provided in this application, the method comprising the following steps:

[0066] S101: Acquire original data, and determine distorted digital signal data based on the original data.

[0067] The processing device acquires original data and determines distorted digital signal data based on the original data.

[0068] In one possible implementation, the processing device may generate an I / Q signal of the baseband of the communication device through the communication device, where I and Q represent two complex components in the communication process of the communication device, and the I / Q signal represents the original data X, which is the data to be sent in the baseband of the communication device.

[0069] In a possible implementation, the original data X can be expressed by the following formula:

[0070] ;

[0071] in, Represents the nth original data, Represents the two components of the nth original data, n∈0, 1, 2, ..., N-1. For complex data, express The imaginary part of .

[0072] Raw data in the baseband of communication equipment After passing through the transmit chain's digital-to-analog converter (DAC), frequency conversion, power amplifier (PA), filtering circuits, and other RF components, the data is transmitted through the antenna. After passing through the PA, the data is distorted, resulting in the distorted original data. This distorted original data then needs to be fed into the communication system's receive chain through feedback circuitry within the communication device.

[0073] The distorted original data passes through the receiving link's filtering circuit, low-noise amplifier (LAN), frequency conversion, and analog-to-digital conversion (ADC) before reaching the baseband of the communication device. The signal at this point is the distorted digital signal data Y, which can be expressed by the following formula:

[0074] ;

[0075] in, Represents the nth distorted digital signal data, Represents the two components of the nth distorted digital signal data, n∈0, 1, 2, ..., N-1. For complex data, express The imaginary part of .

[0076] S102: performing data preprocessing on the original data to obtain preprocessed original data, and performing preprocessing on the distorted digital signal data to obtain preprocessed digital signal data.

[0077] The processing device performs data preprocessing on the original data to obtain preprocessed original data, and preprocesses the distorted digital signal data to obtain preprocessed digital signal data.

[0078] In a possible implementation, the processing device may divide the time series data of the original data X into frame data of the original data , and, dividing the time series data of the distorted digital signal data Y into frame data of the distorted digital signal data , and The number of is T, and The step size is S, and >S.

[0079] Frame data based on raw data Sure Overlapping samples of the original data , t represents T The tth one in the t-th one, and the frame data based on the distorted digital signal data Sure Overlapping samples of distorted digital signal data , t represents T The tth one in ;

[0080] [t]= ;

[0081] in, [t] represents the overlapping sample of the t-th original data, s represents the s-th in the step size S, Indicates the sth The weight, Indicates the sth Another component of .

[0082] Determine amplitude data based on frame data of raw data , phase sinusoidal component and phase cosine component .

[0083] ;

[0084] ;

[0085] ;

[0086] in, and express Corresponding to the two components of t.

[0087] S103: Establishing a power amplifier model based on the preprocessed raw data and the preprocessed digital signal data.

[0088] The processing device establishes a power amplifier model based on the preprocessed raw data and the preprocessed digital signal data.

[0089] In one possible implementation, the processing device may and As label data for the input and output of the power amplifier model, As input, As the output label data, Establishing a power amplifier model as a loss function of the power amplifier model;

[0090] = ;

[0091] in, represents the loss function of the power amplifier model, and The number of is T, The power amplifier model corresponds to the t-th input and the t-th input The output value obtained is express Overlapping samples of distorted digital signal data.

[0092] S104: Pre-training the power amplifier model to obtain a pre-trained power amplifier model.

[0093] The processing device pre-trains the power amplifier model to obtain a pre-trained power amplifier model.

[0094] In one possible implementation, the processing device may in 、 and 、 and The data constitutes the input vector, and express The two complex components of ;

[0095] The input vector is used as the power amplifier model Input, As a power amplifier model The label for the power amplifier model Perform pre-training to obtain a pre-trained power amplifier model .

[0096] S105: Establishing a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital predistortion model of the communication device.

[0097] The processing device establishes a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital predistortion model of the communication device.

[0098] In one possible implementation, the RF link model It is expressed by the following formula:

[0099] + ;

[0100] in, represents the RF link model, represents the pre-trained power amplifier model, represents the digital predistortion model, The loss function is ;

[0101] The RF link model The output label data is , represents the gain of the desired output of the communication system, yes and Obtained by transformation;

[0102] described The loss function is It is expressed by the following formula:

[0103] ;

[0104] in, express The loss function is Indicates the tth The output data, Represents the output label data corresponding to the t-th output data.

[0105] In one possible implementation, a digital predistortion model of a communication device includes a GRU layer, a hidden layer, an activation layer, a connection layer, and an output layer;

[0106] In the GRU layer, input data 、 、 and is represented as a one-dimensional column vector , the number of input neurons M of the GRU layer is 5*T, and the hidden layer of the GRU layer is set to 5*T;

[0107] The hidden layer is a fully connected layer, and the input and output are set to the same parameters as the hidden layer of the GRU layer;

[0108] The activation layer uses the relu activation function, and the data after the hidden layer is ;

[0109] The connection layer will and The data is spliced ​​and input to the output layer. The splicing result It is expressed as follows:

[0110] );

[0111] The output layer is a fully connected layer, and the input data of the output layer is , the output data is represented as components and quantity .

[0112] S106: Based on the radio frequency link model, the digital pre-distortion model is trained on the basis of fixing the parameters of the pre-trained power amplifier model to obtain a trained digital pre-distortion model.

[0113] The processing device trains the digital predistortion model based on the radio frequency link model and on the basis of fixing the parameters of the pretrained power amplifier model to obtain a trained digital predistortion model.

[0114] In one possible implementation, the processing device may in 、 and 、 and The data constitutes the input vector, and express The two complex components of ;

[0115] Will and Transformed into , and As Output label data of

[0116] It is transformed by the following formula:

[0117] ;

[0118] in, and express The two complex components of and express The two complex components of , t represents the t-th input;

[0119] train Model, in training When modeling, first fix The parameters of the model are only optimized The parameters of the model are trained to obtain a trained digital predistortion model.

[0120] S107: Deploy the trained digital pre-distortion model to the baseband chip of the communication device, and offset the digital pre-distortion during the communication process of the communication device through the trained digital pre-distortion model.

[0121] The processing device deploys the trained digital pre-distortion model to the baseband chip of the communication device, and offsets the digital pre-distortion during the communication process of the communication device through the trained digital pre-distortion model.

[0122] The method provided in this application includes data acquisition and preprocessing, power amplifier model modeling, digital predistortion model modeling, model pretraining and deployment, and provides a digital predistortion algorithm model framework based on the GRU network. It realizes the digital predistortion function through smaller network parameters, facilitates subsequent model deployment, and lays the foundation for building a digital predistortion model based on a deep learning network.

[0123] In actual model deployment, the actual application only needs to use the digital predistortion model of the communication equipment The model is deployed in the baseband chip of the communication equipment to offset the digital pre-distortion of the communication equipment and realize digital pre-distortion based on deep learning network.

[0124] Next, we will further explain it in combination with practical applications. Figure 2 This is a system principle block diagram of a digital predistortion communication device and implementation method based on a GRU network provided in an embodiment of the present application.

[0125] Figure 2 It is divided into two parts: model training server and communication equipment.

[0126] The model training server includes five modules: data acquisition, data preprocessing, power amplifier model construction, digital pre-distortion model construction, model pre-training and deployment. Data acquisition and model deployment require data interaction with communication equipment, while the functions of other modules are completed in the model training server.

[0127] In order to realize the digital pre-distortion function in communication equipment, it is necessary to add corresponding data acquisition and model deployment modules to the application layer and baseband layer of the communication equipment. In actual application, the data acquisition module of the model training server initiates the data acquisition process, controls the communication equipment to complete the acquisition of original transmission data and distorted data, and then performs data preprocessing, power amplifier model construction, and digital pre-distortion model construction in sequence until the digital pre-distortion model training is completed. The trained model parameters are deployed to the digital pre-distortion model of the device baseband layer through the pre-training model deployment module.

[0128] The overall processing steps of the system are as follows:

[0129] Step 1: The data acquisition module in the model training server initiates the data acquisition process. The data acquisition module of the communication device starts the data acquisition process and sends the original data sent by the baseband layer. And the received data The data collection is completed for the model training server;

[0130] Step 2: The data preprocessing module in the model training server processes the data into , amplitude data , phase sinusoidal component , phase cosine component ;

[0131] Step 3: Build the power amplifier model in the model training server;

[0132] Step 4: Build a digital pre-distortion model in the model training server;

[0133] Step 5: Use the model pre-training and deployment module in the model training server to pre-train the power amplifier model. Then, based on the trained power amplifier model, train the digital pre-distortion model.

[0134] Step 6: Through model pre-training and deployment in the model training server, the digital pre-distortion model parameters are configured to the baseband layer of the communication device to implement the digital pre-distortion function in the baseband layer.

[0135] Figure 3 This is a network structure diagram of a digital predistortion model provided in an embodiment of the present application. During actual training or use, the processing device can process the input data into 、 、 、 The input data is input to the GRU layer to complete the time series feature extraction. The time series feature data is input to the hidden layer and the activation layer for further feature extraction and nonlinear processing. The data input to the activation layer is spliced ​​with the original data to obtain richer information and then input to the output layer. The output layer outputs the digital pre-distorted 、 data.

[0136] This application also provides a Figure 4 The structure diagram of the digital predistortion communication device based on the GRU network is shown, and the digital predistortion communication device 400 includes:

[0137] An acquisition module 401 is configured to acquire original data and determine distorted digital signal data based on the original data;

[0138] A preprocessing module 402 is configured to perform data preprocessing on the original data to obtain preprocessed original data, and to preprocess the distorted digital signal data to obtain preprocessed digital signal data;

[0139] A first establishing module 403 is configured to establish a power amplifier model based on the pre-processed raw data and the pre-processed digital signal data;

[0140] A pre-training module 404 is configured to pre-train the power amplifier model to obtain a pre-trained power amplifier model;

[0141] A second establishing module 405 is configured to establish a radio frequency link model of the communication device based on the pre-trained power amplifier model and the digital predistortion model of the communication device;

[0142] A training module 406 is configured to train the digital predistortion model based on the radio frequency link model and on the basis that the parameters of the pre-trained power amplifier model are fixed to obtain a trained digital predistortion model;

[0143] The deployment module 407 is configured to deploy the trained digital pre-distortion model to a baseband chip of a communication device, and to offset the digital pre-distortion during the communication process of the communication device by using the trained digital pre-distortion model.

[0144] This paper provides a solution for implementing a digital pre-distortion (DPD) algorithm based on an end-to-end deep learning network in real-world communication devices. This solution includes data acquisition and preprocessing, power amplifier modeling, DPD modeling, model pre-training, and deployment. It also provides a DPD algorithm model framework based on a GRU network. This framework implements DPD with smaller network parameters, facilitating subsequent model deployment and laying the foundation for building DPD models based on deep learning networks.

[0145] An embodiment of the present application further provides a digital predistortion communication device based on a GRU network, wherein the device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes so that the device performs the steps of the digital predistortion communication method based on the GRU network described in any embodiment of the present application.

[0146] In practical applications, the computer-readable storage medium may be any combination of one or more computer-readable media, which may be a computer-readable signal medium or a computer-readable storage medium.

[0147] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0148] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0149] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0150] Computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0151] It should be noted that the various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0152] The above is merely one specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A digital predistortion communication method based on a GRU network, characterized in that: include: Acquiring original data, and determining distorted digital signal data based on the original data, including: Generate an I / Q signal of the communication device baseband by the communication device, where I and Q represent two complex components in the communication process of the communication device, and the I / Q signal represents original data X, which is data to be sent in the communication device baseband; Passing the original data X through a power amplifier of a communication device to obtain distorted original data; Processing the distorted original data through a receiving link of a communication device to obtain distorted digital signal data Y; performing data preprocessing on the original data to obtain preprocessed original data, and performing preprocessing on the distorted digital signal data to obtain preprocessed digital signal data, wherein the method includes: Splitting the time series data of the original data X into frame data X′ of the original data, and splitting the time series data of the distorted digital signal data Y into frame data Y′ of the distorted digital signal data, wherein the number of X′ and Y′ is T, the step length of X′ and Y′ is S, and T>S; Determine TS overlapping samples x′[t] of the original data based on the frame data X′ of the original data, where t represents the tth sample among the T X′, and determine TS overlapping samples y′[t] of the distorted digital signal data based on the frame data Y′ of the distorted digital signal data, where t represents the tth sample among the T Y′; Determine the amplitude data AMP' and the phase sinusoidal component P' based on the frame data of the original data sin and the phase cosine component P′ cos ; Establishing a power amplifier model based on the preprocessed raw data and the preprocessed digital signal data; Pre-training the power amplifier model to obtain a pre-trained power amplifier model; A radio frequency link model of the communication device is established based on the pre-trained power amplifier model and the digital predistortion model of the communication device, wherein the radio frequency link model is expressed by the following formula: Among them, M ser represents the RF link model, represents the pre-trained power amplifier model, M dpd Denotes the digital predistortion model, M ser The loss function is The output label data of the radio frequency link model is G′, where G′ represents the gain of the communication system's expected output, and G′ is obtained by transforming X′ and Y′; The M ser The loss function is It is expressed by the following formula: in, Indicates M ser The loss function, y′ ser [t] represents the tth M ser The output data of , G′[t] represents the output label data corresponding to the t-th output data; Based on the radio frequency link model, the digital predistortion model is trained on the basis of fixing the parameters of the pre-trained power amplifier model to obtain a trained digital predistortion model; The trained digital pre-distortion model is deployed in a baseband chip of a communication device, and the digital pre-distortion during the communication process of the communication device is offset by the trained digital pre-distortion model.

2. The method according to claim 1, characterized in that The establishing of a power amplifier model based on the pre-processed raw data and the pre-processed digital signal data comprises: Let X′ be the input of the power amplifier model, and Y′ be the label data output by the power amplifier model. Establishing a power amplifier model as a loss function of the power amplifier model; in, Denotes the loss function of the power amplifier model, the number of X′ and Y′ is T, y′ pa [t] represents the output value of the power amplifier model corresponding to the t-th input X′ and the t-th input Y′, and y′[t] represents the overlapping samples of TS distorted digital signal data.

3. The method according to claim 2, characterized in that The pre-training the power amplifier model to obtain a pre-trained power amplifier model includes: The I in X′ (x) , Q (x) With AMP′, P′ sin and P′ cos The data constitutes the input vector, I (x) and Q (x) represents the two complex components of X′; The input vector is used as the power amplifier model M pa Input, Y′ as the power amplifier model M pa The label for the power amplifier model M pa Perform pre-training to obtain a pre-trained power amplifier model 4. The method according to claim 3, characterized in that The step of training the digital predistortion model based on the radio frequency link model and on the basis of fixing the parameters of the pre-trained power amplifier model to obtain a trained digital predistortion model includes: The I in X′ (x) , Q (x) With AMP′, P′ sin and P′ cos The data constitutes the input vector, I (x) and Q (x) represents the two complex components of X′; Transform X′ and Y′ into G′, and use G′ as M ser Output label data of G′ is transformed by the following formula: Among them, I (x) and Q (x) represents the two complex components of X′, I (y) and Q (y) represents the two complex components of Y′, and t represents the t-th input; Training M ser Model, in training M ser When modeling, first fix The parameters of the model, only M is optimized dpd The parameters of the model are trained to obtain a trained digital predistortion model.

5. The method according to claim 3, characterized in that The digital predistortion model of the communication device includes a GRU layer, a hidden layer, an activation layer, a connection layer and an output layer; In the GRU layer, input data I (x) , Q (x) ,AMP′,P′ sin and P′ cos is represented as a one-dimensional column vector x input , the number of input neurons M of the GRU layer is 5*T, and the hidden layer of the GRU layer is set to 5*T; The hidden layer is a fully connected layer, and the input and output are set to the same parameters as the hidden layer of the GRU layer; The activation layer uses the relu activation function, and the data after the hidden layer is H x ; The connection layer converts x input and H x The data is spliced ​​and input to the output layer, and the splicing result F cat It is expressed as follows: F cat =cat(x input ,H x ); The output layer is a fully connected layer, and the input data of the output layer is F cat , the output data is represented as component I (out) and component Q (out) .

6. A digital predistortion communication device based on a GRU network, used to execute the digital predistortion communication method based on a GRU network according to claims 1 to 5, characterized in that: include: The baseband chip of the communication device is equipped with a trained digital pre-distortion model, and the trained digital pre-distortion model is used to offset the digital pre-distortion during the communication process of the communication device.

7. A computer-readable storage medium, characterized in that Used to store a computer program, wherein when the computer program is executed by a processor, the digital predistortion communication method based on the GRU network according to any one of claims 1 to 5 is implemented.

Images