Data predistortion processing method, device, equipment, medium and program product

Abstract

The present application discloses a data predistortion processing method, device, equipment, medium and program product, which relates to the field of signal processing technology. The disclosed data predistortion processing method includes: obtaining target input data of a power amplifier and associated input data within a preset time before the input moment of the target input data; constructing target graph data; performing predistortion processing on the target graph data using a predistortion model to obtain predistortion target graph data output by the predistortion model; converting the predistortion data in the predistortion target graph data into a predistortion signal, and inputting the predistortion signal into a power amplifier circuit; the predistortion signal is a digital signal. The present application can avoid the power amplifier from generating nonlinear characteristics when processing input data, thereby improving the accuracy of output data.

Description

Technical Field

[0001] The present application relates to the field of signal processing technology, and in particular to a data pre-distortion processing method, device, equipment, medium and program product. Background Art

[0002] In the field of wireless communication technology, a digital pre-distortion method based on an artificial neural network is used to pre-distort input data of a power amplifier to avoid distortion of output data caused by nonlinear characteristics of the power amplifier.

[0003] However, since the artificial neural network model cannot improve the accuracy of the pre-distortion model, the accuracy of the signal output still needs to be improved. Summary of the invention

[0004] The main purpose of the present application is to provide a data pre-distortion processing method, device, equipment, medium and program product, aiming to solve the technical problem in the related art that the accuracy of pre-distortion processing needs to be improved.

[0005] To achieve the above objectives, the present application proposes a data predistortion processing method, comprising:

[0006] Acquire target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data;

[0007] Construct target graph data; the nodes of the target graph data are target input data or associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data;

[0008] The target graph data is predistorted by using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on a nonlinear topological relationship between historical output data of a power amplifier; and the nodes of the predistortion target graph data store predistortion data corresponding to the target input data;

[0009] The predistortion data in the predistortion target image data is converted into a predistortion signal, and the predistortion signal is input into a power amplifier circuit; the predistortion signal is a digital signal.

[0010] In one embodiment, before the target image data is predistorted by using the predistortion model to obtain the predistorted target image data output by the predistortion model, the method further includes:

[0011] Acquire a historical output data sequence of the power amplifier within a preset time period;

[0012] Construct graph data; the nodes of the graph data are the historical output data in the historical output data sequence, and the edges of the graph data are the spatiotemporal correlations between the historical output data;

[0013] Based on the graph data, the initial pre-distortion model is obtained by inverse modeling;

[0014] An initial pre-distortion model is trained to obtain a pre-distortion model.

[0015] In one embodiment, the step of constructing graph data includes:

[0016] For each historical output data, any output data within a preset time period before the output time of the historical output data is used as the associated output data of the historical output data;

[0017] For each historical output data, determining a first real part and a first imaginary part of any one of the historical output data and all associated output data of the historical output data;

[0018] For any one of the historical output data and all the associated output data of the historical output data, obtain a plurality of first signal amplitude values ​​based on the first real part, the first imaginary part and the nonlinear order of the power amplifier;

[0019] For each historical output data, the set of the first real part, the first imaginary part and the first signal amplitude value of any one of the historical output data and all the associated output data is used as a node, and the spatiotemporal correlation relationship between the historical output data and all the associated output data is used as an edge between the nodes to construct the graph data.

[0020] In one embodiment, based on the graph data, inverse modeling is performed to obtain an initial predistortion model, including:

[0021] Build a learning framework based on all graph data:

[0022] The learning framework is: ;

[0023] in, represents the learning set, represents the i-th graph data, represents the node set in the i-th graph data, represents the edge set between nodes in the i-th graph data, represents the jth node in the i-th node set; represents the label value of the jth node in the i-th node set, Represents the set of all graph data. Represents a collection of tagged values;

[0024] Based on the learning framework, an initial pre-distortion model is obtained by inverse modeling; wherein the initial pre-distortion model includes a global transfer function and an output function;

[0025] The global conversion function is:

[0026] ,

[0027] represents the state vector of the current node, Represents the state vector of the adjacent nodes connected to the current node by edges, represents the node vector of the current node, represents the edge vector connecting the current node and the adjacent nodes, represents the node vector of the neighboring nodes, represents the global conversion function;

[0028] The output function is ,

[0029] represents the output feature vector of the current node, Represents the output function.

[0030] In one embodiment, the step of training the initial pre-distortion model to obtain the pre-distortion model includes:

[0031] Use graph data as training samples; each node in the graph data has a labeled value;

[0032] Inputting the graph data into the initial pre-distortion model to obtain predicted graph data output by the initial pre-distortion model;

[0033] Determining an error value based on the predicted graph data, the graph data, and the error function;

[0034] The error function is ;

[0035] represents the error value, To estimate the parameters, represents the predicted value of the jth node in the i-th graph data obtained using the estimated parameters, represents the number of training samples, Indicates the number of nodes;

[0036] If the error value is greater than the preset threshold, the error value is used for back propagation to update the weight parameters in the initial pre-distortion model, and the execution returns to input the image data into the initial pre-distortion model to obtain the predicted image data output by the initial pre-distortion model until the error value is less than the preset threshold and the pre-distortion model is obtained.

[0037] In one embodiment, the step of constructing target graph data includes:

[0038] determining a second real part and a second imaginary part of any one of the target input data and all associated input data;

[0039] For any one of the target input data and all associated input data, obtain a plurality of second signal amplitude values ​​based on the second real part, the second imaginary part and the nonlinear order of the power amplifier;

[0040] The target input data and the second real part, the second imaginary part and the second signal amplitude value set of any one of all associated input data are taken as nodes, and the spatiotemporal correlation relationship between the target input data and all associated input data is taken as the edge between the nodes to construct the target graph data.

[0041] In addition, to achieve the above-mentioned purpose, the present application also proposes a data pre-distortion processing device, the data pre-distortion processing device comprising:

[0042] An acquisition module, used to acquire target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data;

[0043] A construction module is used to construct target graph data; the nodes of the target graph data are target input data or associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data;

[0044] An acquisition module is used to perform predistortion processing on the target graph data using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between historical output data of the power amplifier; and the nodes of the predistortion target graph data store predistortion data corresponding to the target input data;

[0045] The conversion module is used to convert the predistortion data in the predistortion target image data into a predistortion signal, and input the predistortion signal into the power amplifier circuit; the predistortion signal is a digital signal.

[0046] In addition, to achieve the above-mentioned purpose, the present application also proposes a data pre-distortion processing device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the data pre-distortion processing method as described above.

[0047] In addition, to achieve the above purpose, the present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the data pre-distortion processing method as described above are implemented.

[0048] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, wherein the computer program product includes a computer program, and when the computer program is executed by a processor, the steps of the data pre-distortion processing method as described above are implemented.

[0049] One or more technical solutions proposed in this application have at least the following technical effects:

[0050] In the present application, after obtaining the target input data and the associated input data within a preset time length before the input moment of the target input data, the target graph data is constructed using the target input data and the associated input data within a preset time length before the input moment of the target input data, that is, the nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data. Therefore, the spatiotemporal association relationships between the input data can be extracted from the target graph data, and then the target graph data is input into the pre-distortion model for pre-distortion processing.

[0051] Since the pre-distortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier, and since the graph neural network can reflect the spatiotemporal correlation between data at different times compared to the artificial neural network model, the pre-distortion model of the graph neural network structure can utilize the nonlinear topological relationship between the historical output data to more accurately predict the distortion that may be caused in the target input data, and give the target input new data a linear characteristic opposite to that of the power amplifier based on the distortion that may be caused in the target input data, that is, pre-distortion processing is performed on the input data, thereby improving the signal output accuracy of the output signal, and then realizing linear amplification of the target input data. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0054] Figure 1 A schematic diagram of a flow chart of a first embodiment of a data pre-distortion processing method of the present application;

[0055] Figure 2 This is a schematic diagram of the data for this application;

[0056] Figure 3 This is a schematic diagram of the relationship between the global conversion function and the output function of this application;

[0057] Figure 4 A schematic diagram of updating a node's state vector using a global conversion function in this application;

[0058] Figure 5 This is a schematic diagram of the module structure of the data pre-distortion processing device according to an embodiment of the present application;

[0059] Figure 6 Schematic diagram of the device structure of the hardware operating environment involved in the data pre-distortion processing method in the embodiment of the present application.

[0060] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0062] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0063] The main solution of the embodiment of the present application is: obtaining the target input data of the power amplifier and the associated input data within a preset time before the input moment of the target input data; constructing target graph data; the nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationship between the target input data and the associated input data; using a predistortion model to predistort the target graph data to obtain predistorted target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier; the nodes of the predistortion target graph data store the predistortion data corresponding to the target input data; converting the predistortion data in the predistortion target graph data into a predistortion signal, and inputting the predistortion signal into the power amplifier circuit; the predistortion signal is a digital signal.

[0064] With the rapid development of wireless communication technology, the power of signals is also increasing. As one of the key components of wireless communication systems, power amplifiers will operate in the saturation region due to higher input power, resulting in nonlinear characteristics and causing serious interference to communication quality.

[0065] In order to solve the problem that the nonlinear characteristics of the power amplifier interfere with the communication quality, in the related technology, a nonlinearity opposite to the distortion characteristics of the power amplifier is introduced to compensate for its distortion, and the self-learning and adaptive capabilities of the ANN (Artificial Neural Network) are used to adjust the pre-distortion algorithm, that is, the input data and output data of the power amplifier are used as training samples, and the artificial neural network model is trained to obtain a pre-distortion model, and the pre-distortion model is used to correct the output of the power amplifier to achieve linear output.

[0066] However, when the input data and output data of the power amplifier are used as training samples to train the artificial neural network model to obtain a pre-distortion model, the artificial neural network model cannot extract the spatiotemporal correlation between samples at different times, resulting in low accuracy of the pre-distortion model, which in turn leads to the signal output accuracy still needs to be improved.

[0067] The present application provides a solution, which utilizes the advantage of graph neural networks in reflecting the spatiotemporal correlation between data at different times compared to artificial neural network models, and constructs a pre-distortion model of a graph neural network structure, thereby improving the accuracy of pre-distortion processing.

[0068] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a data pre-distortion processing device, etc. The following takes the data pre-distortion processing device as an example to illustrate this embodiment and the following embodiments.

[0069] Based on this, the present application embodiment provides a data pre-distortion processing method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the data pre-distortion processing method of the present application.

[0070] In this embodiment, the data pre-distortion processing method includes steps S10 to S40:

[0071] Step S10: acquiring target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data.

[0072] In this embodiment, the target input data may be signal data that has not yet been input into the power amplifier for amplification processing. The associated input data is a group of input data that is sequentially sorted in time within a preset time length before the input moment of the target input data. The target input data and the associated input data within a preset time length before the input moment of the target input data may be stored locally in the data predistortion processing device. When executing step S10, the data predistortion processing device may directly call the target input data and the associated input data within a preset time length before the input moment of the target input data from the local.

[0073] Step S20, constructing target graph data.

[0074] The nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data.

[0075] In this embodiment, the edge in the graph data includes a spatiotemporal association relationship between two nodes, which may be a temporal association relationship and a spatial association relationship between the target input data and the associated input data. The temporal association relationship may be the influence information of the moment corresponding to the associated input data on the moment of the target input data, and the spatial association relationship may be the relationship between the target input data and the associated input data, such as the relationship between I (In-phase) and Q (Quadrature) between the target input data and the associated input data.

[0076] It can be understood that when constructing the target graph data using the target input data and the associated input data within a preset time before the input moment of the target input data, in order to improve the performance of the network, additional inputs may be added to the target graph data, such as the signal amplitude value of the target input data and the signal amplitude value of the associated input data. Further, as an optional implementation, step S20 specifically includes:

[0077] Step S201 : determining a second real part and a second imaginary part of any one of the target input data and all associated input data.

[0078] Step S202 : for any one of the target input data and all associated input data, obtain a plurality of second signal amplitude values ​​based on the second real part, the second imaginary part and the nonlinear order of the power amplifier.

[0079] Step S203, taking the second real part, the second imaginary part and the set of the second signal amplitude value of any one of the target input data and all associated input data as nodes, and taking the spatiotemporal correlation relationship between the target input data and all associated input data as edges between nodes, to construct the target graph data.

[0080] In this embodiment, the nonlinear order may be the order of distortion generated by the power amplifier in the output data, which is used to describe the severity of the nonlinear distortion. The target data and the associated input data may be expressed in a complex form, so that the power amplifier can more effectively perform signal processing and analysis on the input data, improve the performance and stability of the system, and meet the needs of complex signal processing.

[0081] Specifically, the nonlinear order of the power amplifier can be determined by performing spectrum analysis on the output data of the power amplifier, measuring the output characteristics of the power amplifier under different input data conditions, such as the relationship between the output data and the input data, performing circuit simulation using circuit simulation tools (such as SPICE software), and simulating the output characteristics of the power amplifier under different working conditions.

[0082] After determining the second real part and the second imaginary part of any one of the target input data and all the associated input data, the second signal amplitude values ​​of the target input data and each associated input data under different nonlinear orders can be determined according to the second real part and the second imaginary part of the target input data and each associated input data, to obtain multiple second signal amplitude values, that is, ;in, represents the amplitude value of the second signal under different nonlinear orders, represents the second real part, Represents the second imaginary part, h represents the nonlinear order of the power amplifier, h∈[1, k], k is the highest nonlinear order of the power amplifier.

[0083] After obtaining the second imaginary part, second real part and multiple second signal amplitude values ​​of the target input data, and the second imaginary part, second real part and multiple second signal amplitude values ​​of each associated input data, the second imaginary part, second real part and multiple second signal amplitude values ​​of the target input data can be used as nodes, and the second imaginary part, second real part and multiple second signal amplitude values ​​of each associated input data can be used as remaining nodes.

[0084] After obtaining multiple nodes, the second real part of the target input data and the second real part of all associated input data, and the second imaginary part of the target input data and the second imaginary part of all associated input data can be used to determine whether there is a spatiotemporal correlation relationship between the nodes.

[0085] Specifically, for any two nodes r and v that are not continuous in time, when the second real part in the node r and the second real part in the node v meet the following conditions:

[0086] ;

[0087] in, represents the second real part in the rth node, represents the second real part in the vth node, represents the second real part in the g-th node, r<v, and r<g<v.

[0088] It is determined that there is a temporal correlation relationship between node r and node v, and an edge is established between node r and node v at the real part level to connect node r and node v.

[0089] After determining whether there is time association information between any two non-adjacent nodes r and node v, the second imaginary parts of all nodes may also be subjected to the aforementioned operation, that is, when the second real part of node r and the second real part of node v meet the following conditions:

[0090] ;

[0091] in, represents the second imaginary part in the rth node, represents the second imaginary part in the vth node, represents the second imaginary part in the g-th node, r<v, and r<g<v.

[0092] Then, an edge is established between node r and node v at the imaginary level to connect node r and node v.

[0093] For any two discontinuous nodes, after determining the connection relationship between the real layer and the imaginary layer, the spatial association relationship between any two discontinuous nodes can be determined according to the connection relationship between the real layer and the imaginary layer, such as the similarity between the nodes or other rules can be used as the spatial association relationship between the nodes. For example, when node r and node v have edges in both the real layer and the imaginary layer, the weight of the edge is determined to be 2; when node r and node v have edges in the real layer and have no edges in the imaginary layer, or when node r and node v have no edges in the real layer and have edges in the imaginary layer, the weight of the edge is determined to be 1; when node r and node v have no edges in both the real layer and the imaginary layer, the weight of the edge is determined to be 0. When the weight of the edge is 1 or 2, it is determined that there is both a temporal association relationship and a spatiotemporal association relationship between node r and node v. When the weight of the edge is 0, it is determined that there is no spatial association relationship and no temporal association relationship between node r and node v.

[0094] Finally, after determining the temporal correlation and spatial correlation between any two discontinuous nodes, the edges in the real layer and the edges in the imaginary layer are used as edges in the target graph data, and the spatial correlation is stored in the corresponding edges to obtain the target graph data.

[0095] For better understanding, a specific example is provided below.

[0096] The input time of the target input data is m, the historical input data between time m and time m-10 is taken as the associated input data, and the second real part of the associated input data between time m-10 and time m-1 is recorded as , ,……, ; The second imaginary part of the associated input data between time m-10 and time m-1 is recorded as , ,……, , the second real part of the target input data is recorded as , the second imaginary part of the target input data is recorded as .

[0097] For the target input data, based on the above operations, multiple second signal amplitude values ​​of the target input data can be calculated and obtained as follows: , ,……, , the multiple second signal amplitude values ​​of the associated input data at time m-10 are respectively , ,……, , and so on, to determine multiple second signal amplitude values ​​for each associated input data.

[0098] After determining the target input data and multiple second signal amplitude values ​​of each associated input data, all second signal amplitude values, second imaginary parts and second real parts between time m-10 and time m can be recorded as: .

[0099] You can , , ,……, As a node, , , , ,……, The set of is taken as a node, and so on. After all the nodes are determined, the two nodes that need to be connected can be determined according to the spatiotemporal correlation between the target input data and all the associated input data, as well as the spatiotemporal correlation between the associated input data, so as to obtain the target graph data.

[0100] In this embodiment, when constructing the target graph data, the second real part, the second imaginary part and multiple second signal amplitude values ​​of the target input data, and the second real part, the second imaginary part and multiple second signal amplitude values ​​of each associated input data are used as additional inputs of the target graph data. This can provide more information, help the pre-distortion model to make predictions more accurately, and thus improve the network performance of the pre-distortion model.

[0101] Step S30, performing predistortion processing on the target image data using the predistortion model to obtain predistorted target image data output by the predistortion model.

[0102] Among them, the pre-distortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier; the nodes of the pre-distortion target graph data store pre-distortion data corresponding to the target input data.

[0103] It should be noted that after obtaining the target graph data of the target input data, in order to facilitate the subsequent rapid determination of the predistortion signal based on the target graph data and improve work efficiency, further, as an optional implementation, before step S30, it also includes:

[0104] Step S26, obtaining a historical output data sequence of the power amplifier within a preset time period;

[0105] In this embodiment, the historical output data sequence may include multiple historical output data arranged according to output time. The historical output data sequence may be stored locally in the data predistortion processing device, and when executing step S26, the data predistortion processing device may directly call the historical output data sequence locally.

[0106] Step S27, construct graph data.

[0107] The nodes of the graph data are the historical output data in the historical output data sequence, and the edges of the graph data are the spatiotemporal correlations between the historical output data.

[0108] It should be noted that when constructing graph data using historical output data sequences, due to the memory effect of the power amplifier, that is, due to the nonlinear effect of the power amplifier, the characteristics of the output data will be affected by the previous signal state. Therefore, when constructing graph data, in order to better reflect the data characteristics between historical output data at different output moments in the graph data sample, further, as an optional implementation, step S27 specifically includes:

[0109] Step S271 , for each historical output data, any output data within a preset time period before the output time of the historical output data is used as the associated output data of the historical output data.

[0110] Step S272: for each historical output data, determine the first real part and the first imaginary part of any one of the historical output data and all associated output data of the historical output data.

[0111] Step S273 , for any one of the historical output data and all associated output data of the historical output data, obtain a plurality of first signal amplitude values ​​based on the first real part, the first imaginary part and the nonlinear order of the power amplifier.

[0112] Step S274, for each historical output data, take the first real part, the first imaginary part and the first signal amplitude value of any one of the historical output data and all the associated output data as a node, and take the spatiotemporal correlation relationship between the historical output data and all the associated output data as an edge between the nodes, to construct the graph data.

[0113] Specifically, after obtaining the historical output data sequence, for each historical output data, any output data within a preset time period before the historical output data moment can be used as the associated output data of the historical output data. Based on the above, since the input data of the power amplifier is expressed in a plural form, and in order to enable the output data to maintain the integrity and accuracy of the signal, the output data of the power amplifier is also expressed in a plural form.

[0114] For each historical output data, after all associated output data of the historical output data are determined, the first real part and the first imaginary part of the historical output data can be extracted from the historical output data, and the first real part and the first imaginary part of each associated output data can be extracted from each associated output data. After determining the first real part and the first imaginary part of the historical output data, and the first real part and the first imaginary part of each associated output data, the first signal amplitude values ​​of the historical output data and each associated output data under different nonlinear orders can be determined by referring to the above method, and multiple first signal amplitude values ​​can be obtained, that is, , represents the amplitude value of the first signal under different nonlinear orders, Represents the first real part of the output data, Represents the first imaginary part of the output data, h represents the nonlinear order of the power amplifier, h∈[1, k], k is the highest nonlinear order of the power amplifier.

[0115] After obtaining the first imaginary part, the first real part and multiple first signal amplitude values ​​of the historical output data, and the first imaginary part, the first real part and multiple first signal amplitude values ​​of each associated output data, the first imaginary part, the first real part and multiple first signal amplitude values ​​of the historical output data can be taken as a node, and the first imaginary part, the first real part and multiple first signal amplitude values ​​of each associated output data can be taken as the remaining nodes, and the spatiotemporal correlation relationship between the historical output data and all the associated output data is used as the edge between the nodes, that is, referring to the above operation, all the first real parts and all the first imaginary parts are used to determine the time correlation relationship and the spatial correlation relationship between the nodes, and the graph data is constructed, referring to Figure 2 As shown, Figure 2 is a schematic diagram of graph data, where , , are nodes for storing the first real part, the first imaginary part and a set of multiple first signal amplitude values ​​of each associated output data, respectively. is a node for storing a first real part, a first imaginary part and a set of a plurality of first signal amplitude values ​​of historical output data, Representation Node and nodes The edge vector between Representation Node and nodes The edge vector between Representation Node and nodes The edge vector between Representation Node and nodes The edge vector between .

[0116] It can be seen that each historical output data can be constructed into a graph data.

[0117] In this embodiment, when constructing graph data, any output data within a preset time period before the output moment of historical output data is used as associated output data of the historical output data, and the historical output data and all associated output data are used to construct graph data, so that the constructed graph data can well reflect the degree to which the characteristics of the historical output data are affected by the previous signal state, thereby allowing the graph data to better reflect the data characteristics between the historical output data at different output moments.

[0118] Step S28: Based on the graph data, inverse modeling is performed to obtain an initial pre-distortion model.

[0119] After obtaining the graph data, in order to enable the initial pre-distortion model to adaptively learn and update the state vector of the node in the process of training the initial pre-distortion model using the graph data, thereby improving the expressiveness and generalization ability of the model, a learning framework can be constructed using all the graph data, and the initial pre-distortion model can be obtained by inverse modeling using the learning framework, that is, a learning framework is constructed based on all the graph data; the learning framework is: ;in, represents the learning set, represents the i-th graph data, represents the node set in the i-th graph data, represents the edge set between nodes in the i-th graph data, represents the jth node in the i-th node set; represents the label value of the jth node in the i-th node set, Represents the set of all graph data. Represents a set of labeled values. Based on the learning framework, the initial pre-distortion model is obtained by inverse modeling. The initial pre-distortion model includes a global conversion function and an output function. The global conversion function is: , represents the state vector of the current node, Represents the state vector of the adjacent nodes connected to the current node by edges, represents the node vector of the current node, represents the edge vector connecting the current node and the adjacent nodes, represents the node vector of the neighboring nodes, represents the global conversion function. The output function is , represents the output feature vector of the current node, Represents the output function.

[0120] Step S29: training an initial pre-distortion model to obtain a pre-distortion model.

[0121] It should be noted that in the original graph data, the information of each node only includes the information of itself, adjacent edges and adjacent nodes, but non-adjacent nodes may also affect the node. Since the state vector of a node in the graph data can only reflect the local information of the node, rather than the information of the node in the entire graph data, in the process of using the graph data to train the initial pre-distortion model, the state vector of the current node can be calculated using the state vector of the adjacent nodes adjacent to the current node, so as to obtain the state vector of the current node in the graph data in this recursive way. Since the function of the state vector of the current node is a recursive formula and it is difficult to calculate directly, the GNN (Graph Neural Network) can calculate the state vector of the current node with the help of the iterative properties of the fixed point theory, that is, to ensure that the state vector x of the current node finally converges to a fixed point after t iterations. After the state vector of the current node converges to a fixed point, the node vector of the current node and the converged state vector can be converted to obtain the model output, and then the gradient is calculated using the loss back propagation, and finally the parameters of the initial pre-distortion model are updated using the gradient descent method.

[0122] Therefore, further, in order to improve the training efficiency of the initial pre-distortion model, as an optional implementation, step S29 specifically includes:

[0123] Step S291, using graph data as training samples; each node in the graph data has a labeled value.

[0124] Step S292: input the image data into the initial pre-distortion model to obtain the predicted image data output by the initial pre-distortion model.

[0125] Step S293, determining the error value based on the predicted graph data, graph data and error function.

[0126] The error function is ;

[0127] represents the error value, To estimate the parameters, represents the predicted value of the jth node in the i-th graph data obtained using the estimated parameters, represents the number of training samples, Indicates the number of nodes.

[0128] Step S294, if the error value is greater than the preset threshold, the error value is used for back propagation to update the weight parameters in the initial pre-distortion model, and the execution returns to input the image data into the initial pre-distortion model to obtain the predicted image data output by the initial pre-distortion model, until the error value is less than the preset threshold, and the pre-distortion model is obtained.

[0129] Specifically, after the graph data is used as a training sample and input into the initial pre-distortion model, the initial pre-distortion model can use a certain node in the graph data as the current node, for example, use the node storing the first real part, the first imaginary part and multiple first signal amplitude values ​​of the historical output data as the current node. For the current node, the global conversion function can be used to calculate the state vector of the current node, that is, the state vector of the adjacent nodes connected to the current node by edges, the node vector of the current node, the edge vector connecting the current node and the adjacent nodes, and the sum of the node vectors of the adjacent nodes are used as the intermediate vector of the current node, and the global conversion function is used to convert the intermediate vector to obtain the state vector of the current node. For details, refer to Figure 3 As shown, Figure 3 Schematic diagram of the relationship between the global conversion function and the output function. The calculation method of the state vector of the adjacent node is the same as the calculation method of the state vector of the current node, which will not be repeated here. When the only adjacent node of the adjacent node is the current node, the state vector of the adjacent node is composed of the node vector of the adjacent node and the edge connected to the adjacent node.

[0130] After obtaining the state vector of the current node, the output function can be used to transform the state vector of the current node and the node vector of the current node, and the output result of the conversion function is used as the prediction graph data of the initial pre-distortion model. The state vector of the current node can be iterated multiple times, that is, , recorded as formula 1, where represents the t-th iteration of the state vector x of the current node, Can be thought of as a global conversion function Therefore, the output of the state vector and the output function can be calculated iteratively, that is, , recorded as formula 2, , recorded as formula 3, and the minimization of the error function is expressed as a learning task, that is, , recorded as formula 4, where the error value must be smaller than the preset threshold.

[0131] When the error value is not less than the preset threshold, the weight parameters in the initial pre-distortion model can be updated by back propagation. , that is, the equation after defining the stable fixed point iteration t times is: , recorded as formula 5, Represents the state vector of the current node after t iterations. After obtaining the equation of the stable fixed point, the stable fixed point can be determined according to the equation of the stable fixed point, that is, , recorded as formula 6, Represents the a-dimensional identity matrix.

[0132] It should be noted that the error value depends on the output of the initial predistortion model , therefore, the gradient can be calculated using the differential chain rule, that is, , recorded as formula 7, formula 6 and formula 7 are combined to get formula 8, where formula 8 is , the first term of formula eight represents the contribution of the output function to the gradient, and the second term represents the contribution of the conversion function to the gradient. Further, formula five is further derived to obtain formula nine, and the relationship between the state vector after the Tth iteration and the state vector after the tth iteration is obtained, that is, formula nine is: .

[0133] Assumptions ,in, , When , recorded as formula 10, where T is the starting iteration number, To terminate the iteration number, refer to Figure 4 As shown, Figure 4 This is a schematic diagram of using the global conversion function to update the state vector of the node. Formula 10 represents the relationship between the fixed point iteration equation and the error. When back propagation is performed to update the weight parameters, the calculation If it is less than the set value, it is considered that the convergence point of the global transfer function and the output function has been found.

[0134] In this embodiment, in the process of training the pre-distortion processing model using graph data, the state vector of the adjacent nodes adjacent to the current node is used to calculate the state of the current node, so as to obtain the state vector of the current node in the graph data in this recursive manner. After determining the state vector of the current node, the state vector of the current node is calculated with the help of the iterative property of the fixed point theory, that is, it is ensured that the state vector x of the current node finally converges to a fixed point after t iterations. After the state vector of the current node converges to a fixed point, the node vector of the current node and the converged state vector can be converted to obtain the model output, and then the gradient is calculated using the loss back propagation, and finally the parameters of the initial pre-distortion model are updated using the gradient descent method, which can avoid the situation where the state vector of the current node has too many iterations, resulting in low training efficiency, thereby improving the training efficiency of the initial pre-distortion model.

[0135] Step S40, converting the predistortion data in the predistortion target image data into a predistortion signal, and inputting the predistortion signal into a power amplifier circuit; the predistortion signal is a digital signal.

[0136] Specifically, after obtaining the predistortion target graph data, the predistortion data in the target graph data can be converted into a signal amplitude value with nonlinear order information after predistortion, and further converted into a predistortion signal in complex form according to the conversion relationship between the signal amplitude value with nonlinear order information and the real part and the imaginary part. After obtaining the predistortion signal, the predistortion signal can be input into a digital-to-analog converter to convert the predistortion signal of the digital signal into an analog signal, and finally the analog signal is input into a power amplifier, so that the output data of the power amplifier achieves a linear amplification effect.

[0137] In this embodiment, since the pre-distortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier, and since the graph neural network can reflect the spatiotemporal correlation between data at different times compared to the artificial neural network model, the pre-distortion model of the graph neural network structure can utilize the nonlinear topological relationship between the historical output data to more accurately predict the distortion that may be caused in the target input data, and give the target input data a linear characteristic opposite to that of the power amplifier according to the distortion that may be caused in the target input data, that is, pre-distortion processing is performed on the target input data, thereby improving the signal output accuracy of the output signal, and then realizing linear amplification of the target input data.

[0138] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the data pre-distortion processing method of the present application. More simple transformations based on this technical concept are all within the protection scope of the present application.

[0139] This application also provides a data pre-distortion processing device, please refer to Figure 5 , the data pre-distortion processing device comprises:

[0140] An acquisition module, used to acquire target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data;

[0141] A construction module is used to construct target graph data; the nodes of the target graph data are target input data or associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data;

[0142] An acquisition module is used to perform predistortion processing on the target graph data using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between historical output data of the power amplifier; and the nodes of the predistortion target graph data store predistortion data corresponding to the target input data;

[0143] The conversion module is used to convert the predistortion data in the predistortion target image data into a predistortion signal, and input the predistortion signal into the power amplifier circuit; the predistortion signal is a digital signal.

[0144] The data predistortion processing device provided by the present application adopts the data predistortion processing method in the above embodiment, which can solve the technical problem that the predistortion processing accuracy in the related art needs to be improved. Compared with the related art, the beneficial effects of the data predistortion processing device provided by the present application are the same as the beneficial effects of the data predistortion processing method provided by the above embodiment, and other technical features in the data predistortion processing device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0145] The present application provides a data pre-distortion processing device, the data pre-distortion processing device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data pre-distortion processing method in the above-mentioned embodiment 1.

[0146] Reference below Figure 6, which shows a schematic diagram of the structure of a data pre-distortion processing device suitable for implementing an embodiment of the present application. The data pre-distortion processing device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The data pre-distortion processing device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0147] like Figure 6 As shown, the data pre-distortion processing device may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the data pre-distortion processing device are also stored. The processing device 1001, ROM1002 and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the data pre-distortion processing device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows a data pre-distortion processing device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or provided instead.

[0148] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0149] The data pre-distortion processing device provided by the present application adopts the data pre-distortion processing method in the above embodiment, which can solve the technical problem that the accuracy of pre-distortion processing in the related art needs to be improved. Compared with the related art, the beneficial effects of the data pre-distortion processing device provided by the present application are the same as the beneficial effects of the data pre-distortion processing method provided by the above embodiment, and other technical features in the data pre-distortion processing device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0150] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0151] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0152] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the data pre-distortion processing method in the above-mentioned embodiment.

[0153] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0154] The computer-readable storage medium may be included in the data pre-distortion processing device; or may exist independently without being assembled into the data pre-distortion processing device.

[0155] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the data predistortion processing device, the data predistortion processing device: obtains the target input data of the power amplifier and the associated input data within a preset time length before the input moment of the target input data; constructs target graph data; the nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationship between the target input data and the associated input data; performs predistortion processing on the target graph data using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier; the nodes of the predistortion target graph data store the predistortion data corresponding to the target input data; converts the predistortion data in the predistortion target graph data into a predistortion signal, and inputs the predistortion signal into the power amplifier circuit; the predistortion signal is a digital signal.

[0156] Computer program code for performing the operations of the present application 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 separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of 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., via the Internet using an Internet service provider).

[0157] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0158] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0159] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned data pre-distortion processing method, and can solve the technical problem that the accuracy of pre-distortion processing in the related art needs to be improved. Compared with the related art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the data pre-distortion processing method provided in the above-mentioned embodiment, and will not be repeated here.

[0160] The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the data pre-distortion processing method as described above are implemented.

[0161] The computer program product provided by the present application can solve the technical problem that the accuracy of pre-distortion processing in the related art needs to be improved. Compared with the related art, the beneficial effects of the computer program product provided by the present application are the same as the beneficial effects of the data pre-distortion processing method provided by the above embodiment, which will not be repeated here.

[0162] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A data predistortion processing method, characterized in that: The method includes: Acquire target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data; Constructing target graph data; the nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationships between the target input data and the associated input data; The target graph data is predistorted by using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on a nonlinear topological relationship between historical output data of the power amplifier; and the nodes of the predistortion target graph data store predistortion data corresponding to the target input data; The predistortion data in the predistortion target image data is converted into a predistortion signal, and the predistortion signal is input into a power amplifier circuit; the predistortion signal is a digital signal.

2. The data predistortion processing method according to claim 1, characterized in that: Before the pre-distortion processing is performed on the target image data by using the pre-distortion model to obtain the pre-distortion target image data output by the pre-distortion model, the method further includes: Acquire a historical output data sequence of the power amplifier within a preset time period; Constructing graph data; the nodes of the graph data are the historical output data in the historical output data sequence, and the edges of the graph data are the spatiotemporal associations between the historical output data; Based on the graph data, an initial pre-distortion model is obtained by inverse modeling; The initial pre-distortion model is trained to obtain the pre-distortion model.

3. The data predistortion processing method according to claim 2, characterized in that: The step of constructing graph data includes: For each of the historical output data, any output data within the preset time period before the output time of the historical output data is used as the associated output data of the historical output data; For each of the historical output data, determining a first real part and a first imaginary part of the historical output data and any one of all the associated output data of the historical output data; For any one of the historical output data and all the associated output data of the historical output data, obtain a plurality of first signal amplitude values ​​based on the first real part, the first imaginary part and the nonlinear order of the power amplifier; For each of the historical output data, the set of the first real part, the first imaginary part and the first signal amplitude value of any one of the historical output data and all the associated output data is used as a node, and the spatiotemporal correlation relationship between the historical output data and all the associated output data is used as an edge between the nodes to construct the graph data.

4. The data predistortion processing method according to claim 3, characterized in that: The inverse modeling based on the graph data to obtain an initial pre-distortion model includes: Based on all the graph data, a learning framework is constructed: The learning framework is: ; in, represents the learning set, represents the i-th graph data, represents the node set in the i-th graph data, represents the edge set between nodes in the i-th graph data, represents the jth node in the i-th node set; represents the label value of the jth node in the i-th node set, represents the set of all the graph data, represents a set of said annotation values; Based on the learning framework, an initial pre-distortion model is obtained by inverse modeling; wherein the initial pre-distortion model includes a global conversion function and an output function; The global conversion function is: ， represents the state vector of the current node, Represents the state vector of the adjacent nodes connected to the current node by edges, represents the node vector of the current node, represents the edge vector connecting the current node and the adjacent nodes, represents the node vector of the neighboring nodes, represents the global conversion function; The output function is , represents the output feature vector of the current node, represents the output function.

5. The data predistortion processing method according to claim 3, characterized in that: The step of training the initial pre-distortion model to obtain the pre-distortion model comprises: The graph data is used as a training sample; each node in the graph data has a labeled value; Inputting the graph data into the initial pre-distortion model to obtain predicted graph data output by the initial pre-distortion model; determining an error value based on the predicted graph data, the graph data and an error function; The error function is ; represents the error value, To estimate the parameters, represents the predicted value of the jth node in the i-th graph data obtained using the estimated parameters, represents the number of training samples, Indicates the number of nodes; If the error value is greater than a preset threshold, back propagation is performed using the error value to update the weight parameters in the initial pre-distortion model, and the process returns to input the graph data into the initial pre-distortion model to obtain predicted graph data output by the initial pre-distortion model, until the error value is less than a preset threshold to obtain the pre-distortion model.

6. The data predistortion processing method according to claim 1, characterized in that: The step of constructing the target graph data comprises: determining a second real part and a second imaginary part of any one of the target input data and all the associated input data; For any one of the target input data and all the associated input data, obtain a plurality of second signal amplitude values ​​based on the second real part, the second imaginary part and the nonlinear order of the power amplifier; The target graph data is constructed by taking the second real part, the second imaginary part and the second signal amplitude value of any one of the target input data and all the associated input data as nodes, and taking the spatiotemporal correlation relationship between the target input data and all the associated input data as edges between nodes.

7. A data predistortion processing device, characterized in that: The device comprises: An acquisition module, used to acquire target input data of the power amplifier and associated input data within a preset time period before the input moment of the target input data; A construction module, used for constructing target graph data; the nodes of the target graph data are the target input data or the associated input data, and the edges of the target graph data are the spatiotemporal association relationship between the target input data and the associated input data; An acquisition module, used to perform predistortion processing on the target graph data using a predistortion model to obtain predistortion target graph data output by the predistortion model; wherein the predistortion model is a graph neural network obtained by inverse modeling based on the nonlinear topological relationship between the historical output data of the power amplifier; the nodes of the predistortion target graph data store predistortion data corresponding to the target input data; The conversion module is used to convert the predistortion data in the predistortion target image data into a predistortion signal, and input the predistortion signal into a power amplifier circuit; the predistortion signal is a digital signal.

8. A data predistortion processing device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the data predistortion processing method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the data pre-distortion processing method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the data predistortion processing method according to any one of claims 1 to 6 are implemented.

Images